Disclosure of Invention
The invention overcomes the defects of the prior art and provides a gun-ball integrated solar hunting camera management method based on intelligent identification.
The technical scheme adopted by the invention for achieving the purpose is as follows:
The invention discloses a gun-ball integrated solar hunting camera management method based on intelligent identification, which comprises the following steps:
Acquiring a site characteristic three-dimensional model map of a target area, dividing the site characteristic three-dimensional model map into a plurality of sub-site characteristic three-dimensional model maps, and analyzing the terrain, sunlight and vegetation conditions of each sub-site characteristic three-dimensional model map to obtain a plurality of target installation sub-areas;
Installing the target hunting camera into the target installation subarea to complete the installation layout process of the target hunting camera; acquiring infrared radiation signal information in each preset area through a PIR sensor in the target hunting camera, and converting the infrared radiation signal information into preset type electric signals;
If the threshold value of the preset type electric signal is larger than the preset threshold value, the PIR sensor generates a trigger signal, a target hunting camera is started according to the trigger signal, video frame information of a preset area is obtained through the target hunting camera, and recognition processing is carried out on the video frame information to obtain a recognition result;
if the identification result is a first identification result, indicating that the target hunting object does not exist in the preset area, and controlling the target hunting camera to be closed; if the identification result is the second identification result, indicating that the target hunting object exists in the preset area, generating alarm information and pushing the alarm information to a preset user terminal.
Further, in a preferred embodiment of the present invention, a site feature three-dimensional model map of a target area is obtained, the site feature three-dimensional model map is divided into a plurality of sub-site feature three-dimensional model maps, and terrain, solar radiation and vegetation condition analysis is performed on each sub-site feature three-dimensional model map to obtain a plurality of target installation sub-areas, which specifically includes:
acquiring remote sensing image information of a target area, constructing a site characteristic three-dimensional model map of the target area based on the remote sensing image information, and dividing the site characteristic three-dimensional model map into a plurality of sub-areas to obtain a plurality of sub-site characteristic three-dimensional model maps;
Performing terrain feature analysis on the plurality of sub-site feature three-dimensional model diagrams, and marking the sub-site feature three-dimensional model diagrams with the terrain features meeting the preset terrain feature requirements to obtain marked sub-site feature three-dimensional model diagrams; marking a subarea corresponding to the marked subarea characteristic three-dimensional model map as a first installation subarea; wherein the topographical features include mountain ranges, rivers, and woodlands;
Acquiring product specification information of a target hunting camera, determining total power consumption information and solar panel efficiency information of the target hunting camera according to the product specification information, and calculating daily average solar radiation amount required by the target hunting camera according to the total power consumption information and the solar panel efficiency information of the target hunting camera;
Acquiring average sunlight information of each first installation sub-region, and determining the daily average solar radiation quantity which can be achieved by each first installation sub-region according to the average sunlight information; comparing the daily average solar radiation amount which can be achieved by each first installation subarea with the daily average solar radiation amount required by the target hunting camera;
And further calibrating the first mounting subarea corresponding to the daily average solar radiation quantity which can be achieved and is larger than or equal to the daily average solar radiation quantity required by the target hunting camera as a second sub-mounting area.
Further, in a preferred embodiment of the present invention, the method further comprises the steps of:
Acquiring a subfield characteristic three-dimensional model diagram of each second subfield installation area and the achievable average sunlight information; the method comprises the steps of obtaining outline dimension information of a target hunting camera, and constructing an outline feature three-dimensional model diagram of the target hunting camera according to the outline dimension information;
according to the three-dimensional model map of the appearance characteristic and the three-dimensional model map of the sub-field characteristic of the second sub-installation area, simulating and installing the target hunting camera into each second sub-installation area to obtain a plurality of simulated installation model maps;
analyzing vegetation distribution conditions of each simulated installation model diagram so as to analyze whether vegetation exists in a preset range area of the target hunting camera; if not, the corresponding second sub-installation area is further marked as a target installation sub-area;
If the vegetation type information and the real-time growth condition information in the preset range area in the corresponding second sub-installation area exist, a search tag is generated based on the vegetation type information and the real-time growth condition information in the preset range area, and the predicted growth rate of the vegetation in the preset range area is obtained by searching in a big data network based on the search tag;
acquiring preset working time length information of a target hunting camera, and constructing a predicted growth model diagram of vegetation in a preset range area after the preset working time length information according to the preset working time length information of the target hunting camera, the real-time growth condition information of the vegetation in the preset range area and the predicted growth rate;
fitting a predicted growth model diagram of vegetation in the preset range area after the preset working time information into a corresponding simulated installation model diagram to obtain a second simulated installation model diagram of the corresponding second sub-installation area after the preset working time information;
Respectively carrying out sunlight shielding simulation analysis on each second simulation installation model graph according to the average sunlight information which can be achieved by the corresponding second sub-installation area, and further obtaining the simulated daily average solar radiation quantity of the corresponding second sub-installation area after the preset working time information; the simulated daily average solar radiation of the corresponding second sub-mount area is compared with the daily average solar radiation required by the target hunting camera,
Further marking a second sub-installation area corresponding to the simulated daily average solar radiation amount which is greater than or equal to the daily average solar radiation amount required by the target hunting camera as a target installation sub-area;
obtaining the geographic position information of all the target installation sub-areas, and outputting the geographic position information of all the target installation sub-areas to a preset platform.
Further, in a preferred embodiment of the present invention, the video frame information is identified, so as to obtain an identification result, which specifically is:
Acquiring various morphological image information of the target hunting object through a big data network, and constructing various morphological feature model diagrams of the target hunting object according to the various morphological image information; constructing a knowledge graph, and importing various morphological feature model graphs of the target hunting object into the knowledge graph;
Acquiring video frame information of a preset area, extracting images of a plurality of specific frames from the video frame information, and storing the images as a picture format to obtain a plurality of pieces of real-time picture information of the preset area; dividing the real-time picture information based on a region growing algorithm to obtain region-of-interest image information;
Constructing a living animal three-dimensional model diagram in a preset area according to the image information of the region of interest; introducing an ICP algorithm, and calculating the similarity between the three-dimensional model diagram of the living animal and each morphological feature model diagram in the knowledge graph based on the ICP algorithm to obtain a plurality of similarities; comparing the similarity with a preset similarity one by one;
if the similarity is not greater than the preset similarity, indicating that the living animal in the preset area is not a target prey, and generating a first identification result;
If at least one of the similarities is larger than the preset similarity, the living animal in the preset area is the target prey, and a second identification result is generated.
Further, in a preferred embodiment of the present invention, a three-dimensional model map of a living animal in a preset area is constructed according to the image information of the region of interest, specifically:
Carrying out feature extraction processing on the image information of a plurality of regions of interest based on a SURF algorithm to obtain a plurality of initial feature points in the image information of each region of interest;
introducing a RANSAC algorithm, and carrying out feature pairing on initial feature points in each piece of region-of-interest image information to obtain a plurality of initial feature point pairs; randomly selecting a preset number of initial characteristic point pairs from a plurality of initial characteristic point pairs to serve as sample point pairs;
Estimating a transformation model according to the randomly selected sample point pairs to obtain an initial transformation model; predicting the positions of all initial feature point pairs by using the initial transformation model to obtain the predicted positions of all initial feature point pairs;
Comparing the predicted position and the actual position of each initial characteristic point pair one by one; if the difference between the predicted position and the actual position of a certain initial feature point pair is smaller than a preset threshold value, marking the initial feature point pair as an inner point, otherwise marking the initial feature point pair as an outer point; counting the number of all the inner points;
Repeating the steps until the iteration times are greater than the preset times, selecting an initial transformation model with the maximum number of inner points as an optimal transformation model, and re-estimating the optimal transformation model by using all inner points of the optimal transformation model to obtain a final transformation model;
Inputting all initial characteristic point pairs into the final transformation model, and predicting the positions of all initial characteristic point pairs by using the final transformation model to obtain a second predicted position;
Comparing the second predicted position of each initial feature point pair with the actual position one by one; if the difference between the second predicted position and the actual position of a certain initial feature point pair is smaller than a preset threshold value, marking the initial feature point pair as an inner point, otherwise marking the initial feature point pair as an outer point; removing the initial characteristic point pair marked as the outer point, and reserving the initial characteristic point pair marked as the inner point to obtain a final characteristic point pair;
Generating a living animal characteristic point model in a preset area based on the final characteristic point pairs, acquiring relative coordinates among the characteristic points based on the living animal characteristic point model, and generating a characteristic point relative coordinate set based on the relative coordinates;
And importing the characteristic point relative coordinate set into three-dimensional modeling software for model construction to obtain a living animal three-dimensional model diagram in a preset area.
Further, in a preferred embodiment of the present invention, if the identification result is the second identification result, which indicates that the target hunting object exists in the preset area, alarm information is generated, and the alarm information is pushed to the preset user terminal, specifically:
If the identification result is a second identification result, indicating that the target hunting object exists in the preset area, acquiring real-time activity characteristic information of the target hunting object; the real-time activity characteristic information comprises real-time activity time, real-time activity speed, real-time acquisition position, real-time activity direction and real-time activity topography characteristics;
acquiring a large amount of target hunting object historical activity data through a big data network, and dividing the target hunting object historical activity data into training set data and test set data; wherein the historical activity data includes time, location coordinates, speed, direction, and preference activity terrain;
Constructing a polynomial regression model based on a deep learning network, and dividing the training set data into a feature set X and a corresponding target value set Y; wherein X comprises time, position coordinates, speed and direction characteristics; y contains the corresponding prey location or trajectory;
Converting the feature set X into a new feature set containing polynomial features using a polynomial feature converter, including higher order terms and interaction terms of the added features; fitting the converted feature set X and the corresponding target value set Y by using a polynomial regression model, so that the polynomial regression model continuously searches for a proper polynomial function to fit the relation between the feature and the target;
fitting a polynomial regression model by a least square method to find the optimal polynomial coefficient, so that the polynomial regression model can be best fit to training data; the fitting degree of the polynomial regression model is evaluated by using the mean square error or the coefficient index on the test set data, and an evaluation result is obtained;
According to the evaluation result, preventing the polynomial regression model from being overfitted by adjusting the order of the polynomial or using a regularization technology; the generalization capability of the polynomial regression model is verified through a cross verification method, so that the polynomial regression model can be accurately predicted on unseen data; repeating the steps until the model meets the preset requirement, and storing the trained polynomial regression model;
The real-time activity characteristic information of the target hunting object is imported into the trained polynomial regression model to carry out reasoning prediction so as to predict the activity track of the target hunting object in real time; and generating alarm information based on the target hunting activity track and the real-time activity characteristic information, and pushing the alarm information to a preset user terminal.
The invention discloses a gun-ball integrated solar hunting camera management system based on intelligent recognition, which comprises a memory and a processor, wherein a gun-ball integrated solar hunting camera management method program is stored in the memory, and when the gun-ball integrated solar hunting camera management method program is executed by the processor, the following steps are realized:
Acquiring a site characteristic three-dimensional model map of a target area, dividing the site characteristic three-dimensional model map into a plurality of sub-site characteristic three-dimensional model maps, and analyzing the terrain, sunlight and vegetation conditions of each sub-site characteristic three-dimensional model map to obtain a plurality of target installation sub-areas;
Installing the target hunting camera into the target installation subarea to complete the installation layout process of the target hunting camera; acquiring infrared radiation signal information in each preset area through a PIR sensor in the target hunting camera, and converting the infrared radiation signal information into preset type electric signals;
If the threshold value of the preset type electric signal is larger than the preset threshold value, the PIR sensor generates a trigger signal, a target hunting camera is started according to the trigger signal, video frame information of a preset area is obtained through the target hunting camera, and recognition processing is carried out on the video frame information to obtain a recognition result;
if the identification result is a first identification result, indicating that the target hunting object does not exist in the preset area, and controlling the target hunting camera to be closed; if the identification result is the second identification result, indicating that the target hunting object exists in the preset area, generating alarm information and pushing the alarm information to a preset user terminal.
The invention solves the technical defects existing in the background technology, and has the following beneficial effects: according to the invention, the optimal mounting layout points of the hunting cameras are automatically screened out by combining terrain and sunlight analysis, more scientific and accurate mounting point selection can be provided, and the durability and stability of the system can be effectively enhanced; the intelligent recognition of the target prey can be rapidly carried out, the response speed of the system is high, the situation of misjudgment due to missed judgment can be effectively avoided, and the reliability of the system is improved; and the method can predict the moving track of the target hunting object, has high prediction precision, provides hunting object data with high reliability for users, and can help the users to better perform hunting activities.
Detailed Description
In order that the above-recited objects, features and advantages of the present application will be more clearly understood, a more particular description of the application will be rendered by reference to the appended drawings and appended detailed description. It should be noted that, without conflict, the embodiments of the present application and features in the embodiments may be combined with each other.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways than those described herein, and therefore the scope of the present invention is not limited to the specific embodiments disclosed below.
As shown in fig. 1, the first aspect of the present invention discloses a gun-ball integrated solar hunting camera management method based on intelligent recognition, comprising the following steps:
S102: acquiring a site characteristic three-dimensional model map of a target area, dividing the site characteristic three-dimensional model map into a plurality of sub-site characteristic three-dimensional model maps, and analyzing the terrain, sunlight and vegetation conditions of each sub-site characteristic three-dimensional model map to obtain a plurality of target installation sub-areas;
S104: installing the target hunting camera into the target installation subarea to complete the installation layout process of the target hunting camera; acquiring infrared radiation signal information in each preset area through a PIR sensor in the target hunting camera, and converting the infrared radiation signal information into preset type electric signals;
s106: if the threshold value of the preset type electric signal is larger than the preset threshold value, the PIR sensor generates a trigger signal, a target hunting camera is started according to the trigger signal, video frame information of a preset area is obtained through the target hunting camera, and recognition processing is carried out on the video frame information to obtain a recognition result;
S108: if the identification result is a first identification result, indicating that the target hunting object does not exist in the preset area, and controlling the target hunting camera to be closed; if the identification result is the second identification result, indicating that the target hunting object exists in the preset area, generating alarm information and pushing the alarm information to a preset user terminal.
Further, in a preferred embodiment of the present invention, a site feature three-dimensional model map of a target area is obtained, the site feature three-dimensional model map is divided into a plurality of sub-site feature three-dimensional model maps, and terrain, solar radiation and vegetation condition analysis is performed on each sub-site feature three-dimensional model map to obtain a plurality of target installation sub-areas, which specifically includes:
acquiring remote sensing image information of a target area, constructing a site characteristic three-dimensional model map of the target area based on the remote sensing image information, and dividing the site characteristic three-dimensional model map into a plurality of sub-areas to obtain a plurality of sub-site characteristic three-dimensional model maps;
Performing terrain feature analysis on the plurality of sub-site feature three-dimensional model diagrams, and marking the sub-site feature three-dimensional model diagrams with the terrain features meeting the preset terrain feature requirements to obtain marked sub-site feature three-dimensional model diagrams; marking a subarea corresponding to the marked subarea characteristic three-dimensional model map as a first installation subarea; wherein the topographical features include mountain ranges, rivers, and woodlands;
Acquiring product specification information of a target hunting camera, determining total power consumption information and solar panel efficiency information of the target hunting camera according to the product specification information, and calculating daily average solar radiation amount required by the target hunting camera according to the total power consumption information and the solar panel efficiency information of the target hunting camera;
Acquiring average sunlight information of each first installation sub-region, and determining the daily average solar radiation quantity which can be achieved by each first installation sub-region according to the average sunlight information; comparing the daily average solar radiation amount which can be achieved by each first installation subarea with the daily average solar radiation amount required by the target hunting camera;
And further calibrating the first mounting subarea corresponding to the daily average solar radiation quantity which can be achieved and is larger than or equal to the daily average solar radiation quantity required by the target hunting camera as a second sub-mounting area.
The target area is a hunting area, and the remote sensing image of the target area can be obtained through equipment such as an unmanned aerial vehicle, a satellite and the like, so that a site characteristic three-dimensional model image of the target area is constructed according to the remote sensing image, wherein the site characteristic three-dimensional model image is a real-time scene condition image of the target area, such as characteristic information including mountains, rivers, forests and the like of the target area. The three-dimensional model map of the sub-field features is marked by carrying out terrain feature analysis on the three-dimensional model map of each sub-field feature, so that a first installation sub-region is obtained, and the monitoring effect of the camera is prevented from being influenced by big trees, hills or other obstacles if the terrain features meet the wide view; such as terrain features, select terrain installations where the target prey is frequently bought or left or passed, such as near water sources, animal habitats or migration paths, etc., according to the range of motion and habit of the target prey. And the first installation subarea corresponding to the daily average solar radiation quantity required by the target hunting camera is further calibrated to be a second sub-installation area so as to ensure that sufficient light exists in the selected installation position area and ensure that the installed camera can effectively continue to voyage.
Further, in a preferred embodiment of the present invention, the method further comprises the steps of:
Acquiring a subfield characteristic three-dimensional model diagram of each second subfield installation area and the achievable average sunlight information; the method comprises the steps of obtaining outline dimension information of a target hunting camera, and constructing an outline feature three-dimensional model diagram of the target hunting camera according to the outline dimension information;
according to the three-dimensional model map of the appearance characteristic and the three-dimensional model map of the sub-field characteristic of the second sub-installation area, simulating and installing the target hunting camera into each second sub-installation area to obtain a plurality of simulated installation model maps;
analyzing vegetation distribution conditions of each simulated installation model diagram so as to analyze whether vegetation exists in a preset range area of the target hunting camera; if not, the corresponding second sub-installation area is further marked as a target installation sub-area;
If the vegetation type information and the real-time growth condition information in the preset range area in the corresponding second sub-installation area exist, a search tag is generated based on the vegetation type information and the real-time growth condition information in the preset range area, and the predicted growth rate of the vegetation in the preset range area is obtained by searching in a big data network based on the search tag;
acquiring preset working time length information of a target hunting camera, and constructing a predicted growth model diagram of vegetation in a preset range area after the preset working time length information according to the preset working time length information of the target hunting camera, the real-time growth condition information of the vegetation in the preset range area and the predicted growth rate;
fitting a predicted growth model diagram of vegetation in the preset range area after the preset working time information into a corresponding simulated installation model diagram to obtain a second simulated installation model diagram of the corresponding second sub-installation area after the preset working time information;
Respectively carrying out sunlight shielding simulation analysis on each second simulation installation model graph according to the average sunlight information which can be achieved by the corresponding second sub-installation area, and further obtaining the simulated daily average solar radiation quantity of the corresponding second sub-installation area after the preset working time information; the simulated daily average solar radiation of the corresponding second sub-mount area is compared with the daily average solar radiation required by the target hunting camera,
Further marking a second sub-installation area corresponding to the simulated daily average solar radiation amount which is greater than or equal to the daily average solar radiation amount required by the target hunting camera as a target installation sub-area;
obtaining the geographic position information of all the target installation sub-areas, and outputting the geographic position information of all the target installation sub-areas to a preset platform.
After the three-dimensional model map of the external feature of the target hunting camera and the three-dimensional model map of the sub-field feature of the second sub-installation area are obtained, the hunting camera is simulated and installed in each second sub-installation area by software such as SolidWorks, CAD to obtain a simulated installation model map, vegetation distribution condition analysis is carried out on the simulated installation model map to analyze whether vegetation exists in a preset range area of the target hunting camera, if the vegetation does not exist, the situation that after the hunting camera is installed in the second sub-installation area, the probability that the hunting camera is blocked by trees growing in the nearby area because long-time work is needed is extremely low is indicated, the corresponding second sub-installation area is further marked as the target installation sub-area, and the area is indicated to be a preferable installation place of the camera.
If vegetation exists in a preset range area of a target hunting camera, it is indicated that after the camera is installed for a period of time, tree growth in the area near the camera may cause shielding of the camera along with environmental changes, power supply duration deficiency conditions of the camera in the area may occur, at this time, whether the tree in the area can affect the camera needs to be further judged, specifically, according to preset working time length information of the target hunting camera, real-time growth condition information of vegetation in the preset range area and predicted growth rate, the tree growth conditions in the range area are predicted, thereby predicting the predicted growth conditions in the working time of the hunting camera, a predicted growth model diagram of the range area is obtained by utilizing three-dimensional software according to the predicted growth conditions, and the predicted growth model diagram is fitted into a corresponding simulated installation model diagram in three-dimensional software, namely, scene diagram information of the area after the tree in the range area grows is obtained, and the solar radiation average time length of the solar panel is further equal to the solar radiation average time length of the solar panel after the solar panel is further analyzed in the preset working time range, the solar radiation average area is obtained after the solar panel is further subjected to the average radiation of the target is further indicated, if the simulated daily average solar radiation is less than the daily average solar radiation required by the target hunting camera, it is indicated that after the trees grow in the area of the range after the preset working time, the trees can cause shielding influence on the camera, the solar cell panel in the camera cannot be irradiated by enough light, and the area is not suitable for installing the camera.
In summary, the optimal installation layout points of the hunting cameras are automatically screened out by combining topography and sunlight analysis, subjectivity and errors in manual layout point selection can be reduced, more scientific and accurate installation point selection is provided, optimal sunlight quantity received by a solar panel can be ensured, solar charging efficiency is improved, long-term effective operation of equipment is ensured, monitoring range and image quality of the cameras can be improved, later maintenance cost and potential migration cost can be reduced by reasonably planning installation of the layout cameras, and durability and stability of a system are enhanced.
As shown in fig. 2, in a further preferred embodiment of the present invention, the video frame information is identified, so as to obtain an identification result, which is specifically:
S202: acquiring various morphological image information of the target hunting object through a big data network, and constructing various morphological feature model diagrams of the target hunting object according to the various morphological image information; constructing a knowledge graph, and importing various morphological feature model graphs of the target hunting object into the knowledge graph;
S204: acquiring video frame information of a preset area, extracting images of a plurality of specific frames from the video frame information, and storing the images as a picture format to obtain a plurality of pieces of real-time picture information of the preset area; dividing the real-time picture information based on a region growing algorithm to obtain region-of-interest image information;
S206: constructing a living animal three-dimensional model diagram in a preset area according to the image information of the region of interest; introducing an ICP algorithm, and calculating the similarity between the three-dimensional model diagram of the living animal and each morphological feature model diagram in the knowledge graph based on the ICP algorithm to obtain a plurality of similarities; comparing the similarity with a preset similarity one by one;
S208: if the similarity is not greater than the preset similarity, indicating that the living animal in the preset area is not a target prey, and generating a first identification result;
s210: if at least one of the similarities is larger than the preset similarity, the living animal in the preset area is the target prey, and a second identification result is generated.
The image information of the region of interest is the moving range region of the target hunting object, and various morphological image information of the target hunting object, such as morphological image information of the target hunting object during walking, eating, rest, and the like, is obtained through a big data network, so that a morphological feature model diagram is constructed. And calculating the similarity between the three-dimensional model diagram of the living animal and each morphological feature model diagram in the knowledge graph based on an ICP algorithm, and if a plurality of similarities are not larger than a preset similarity, indicating that the living animal in the preset area is not a target hunting object, generating a first recognition result, and controlling the target hunting camera to be closed. If at least one of the similarities is larger than the preset similarity, the living animal in the preset area is the target hunting object, a second recognition result is generated, and continuous tracking is carried out on the target hunting object. The method can be used for rapidly identifying the target hunting object, improves the response speed of the system, can effectively avoid the situation of missed judgment and misjudgment, and improves the reliability of the system.
Further, in a preferred embodiment of the present invention, a three-dimensional model map of a living animal in a preset area is constructed according to the image information of the region of interest, specifically:
Carrying out feature extraction processing on the image information of a plurality of regions of interest based on a SURF algorithm to obtain a plurality of initial feature points in the image information of each region of interest;
introducing a RANSAC algorithm, and carrying out feature pairing on initial feature points in each piece of region-of-interest image information to obtain a plurality of initial feature point pairs; randomly selecting a preset number of initial characteristic point pairs from a plurality of initial characteristic point pairs to serve as sample point pairs;
Estimating a transformation model according to the randomly selected sample point pairs to obtain an initial transformation model; predicting the positions of all initial feature point pairs by using the initial transformation model to obtain the predicted positions of all initial feature point pairs;
Comparing the predicted position and the actual position of each initial characteristic point pair one by one; if the difference between the predicted position and the actual position of a certain initial feature point pair is smaller than a preset threshold value, marking the initial feature point pair as an inner point, otherwise marking the initial feature point pair as an outer point; counting the number of all the inner points;
Repeating the steps until the iteration times are greater than the preset times, selecting an initial transformation model with the maximum number of inner points as an optimal transformation model, and re-estimating the optimal transformation model by using all inner points of the optimal transformation model to obtain a final transformation model;
Inputting all initial characteristic point pairs into the final transformation model, and predicting the positions of all initial characteristic point pairs by using the final transformation model to obtain a second predicted position;
Comparing the second predicted position of each initial feature point pair with the actual position one by one; if the difference between the second predicted position and the actual position of a certain initial feature point pair is smaller than a preset threshold value, marking the initial feature point pair as an inner point, otherwise marking the initial feature point pair as an outer point; removing the initial characteristic point pair marked as the outer point, and reserving the initial characteristic point pair marked as the inner point to obtain a final characteristic point pair;
Generating a living animal characteristic point model in a preset area based on the final characteristic point pairs, acquiring relative coordinates among the characteristic points based on the living animal characteristic point model, and generating a characteristic point relative coordinate set based on the relative coordinates;
And importing the characteristic point relative coordinate set into three-dimensional modeling software for model construction to obtain a living animal three-dimensional model diagram in a preset area.
After a plurality of initial feature points in each piece of region-of-interest image information are obtained, the initial feature points among different images are paired to obtain feature point pairs, so that the feature points among different images are paired and fused to obtain a living animal feature point model, and a living animal three-dimensional model diagram is constructed by utilizing relative coordinates among the feature points in the living animal feature point model through three-dimensional software. However, due to factors such as noise, occlusion, deformation, etc. that may exist in the image, the feature point matching result often includes a certain number of false feature points (also referred to as outliers or outliers). The purpose of eliminating the false feature points is to improve the accuracy and robustness of matching, so that the final feature point matching result is more reliable. False feature points are removed, false matching can be reduced, accuracy of the corresponding relation is improved, subsequent image processing and analysis tasks are better supported, and accuracy of subsequent modeling is improved. In view of the above, in the invention, false feature points are removed by introducing a RANSAC algorithm, so that the quality and reliability of feature point matching are improved, a living animal three-dimensional model diagram with higher precision is constructed, the misjudgment of a system is avoided, and the accuracy and robustness of the system are improved.
Further, in a preferred embodiment of the present invention, if the identification result is the second identification result, which indicates that the target hunting object exists in the preset area, alarm information is generated, and the alarm information is pushed to the preset user terminal, specifically:
If the identification result is a second identification result, indicating that the target hunting object exists in the preset area, acquiring real-time activity characteristic information of the target hunting object; the real-time activity characteristic information comprises real-time activity time, real-time activity speed, real-time acquisition position, real-time activity direction and real-time activity topography characteristics;
acquiring a large amount of target hunting object historical activity data through a big data network, and dividing the target hunting object historical activity data into training set data and test set data; wherein the historical activity data includes time, location coordinates, speed, direction, and preference activity terrain;
Constructing a polynomial regression model based on a deep learning network, and dividing the training set data into a feature set X and a corresponding target value set Y; wherein X comprises time, position coordinates, speed and direction characteristics; y contains the corresponding prey location or trajectory;
Converting the feature set X into a new feature set containing polynomial features using a polynomial feature converter, including higher order terms and interaction terms of the added features; fitting the converted feature set X and the corresponding target value set Y by using a polynomial regression model, so that the polynomial regression model continuously searches for a proper polynomial function to fit the relation between the feature and the target;
fitting a polynomial regression model by a least square method to find the optimal polynomial coefficient, so that the polynomial regression model can be best fit to training data; the fitting degree of the polynomial regression model is evaluated by using the mean square error or the coefficient index on the test set data, and an evaluation result is obtained;
According to the evaluation result, preventing the polynomial regression model from being overfitted by adjusting the order of the polynomial or using a regularization technology; the generalization capability of the polynomial regression model is verified through a cross verification method, so that the polynomial regression model can be accurately predicted on unseen data; repeating the steps until the model meets the preset requirement, and storing the trained polynomial regression model;
The real-time activity characteristic information of the target hunting object is imported into the trained polynomial regression model to carry out reasoning prediction so as to predict the activity track of the target hunting object in real time; and generating alarm information based on the target hunting activity track and the real-time activity characteristic information, and pushing the alarm information to a preset user terminal.
It should be noted that, firstly, historical activity data of the prey needs to be collected, including information such as time, position coordinates, speed, direction and the like; then preprocessing the data, including cleaning the data, removing abnormal values and noise, filling in missing values and the like; the data set is divided into a training set for constructing the model and a testing set for evaluating the performance of the model. A polynomial feature converter (e.g., polynomialFeatures in scikit-learn) is used to convert feature set X into a new feature set containing polynomial features. This includes adding high-order items and interactive items of the feature. For example, if the original feature set X contains time and position coordinates, the polynomial feature transformation may add a temporal higher order term and an interactive term of position coordinates. A polynomial regression model is used to fit the transformed feature set X and the corresponding target value set Y. The polynomial regression model will try to find the appropriate polynomial function to fit the relationship between the feature and the target. The model is fitted by minimizing a loss function (e.g., least squares) to find the optimal polynomial coefficients so that the model fits the training data best. The performance of the model is evaluated using an index such as Mean Square Error (MSE) or decision coefficient (R) over the training set. Depending on the evaluation, the order of the polynomial is adjusted or regularization techniques (such as ridge regression or Lasso) are used to prevent overfitting. Future activity trajectories of the prey are predicted using a trained and optimized polynomial regression model in combination with current prey activity data. The method can predict the moving track of the target hunting object, has high prediction precision, provides hunting object data with high reliability for users, and can help the users to better perform hunting activities.
Furthermore, the method comprises the following steps:
Acquiring operation log information of a target solar hunting camera, and extracting fault records and maintenance records of all devices in the target solar hunting camera and fault characteristic data corresponding to fault periods according to the operation log information;
analyzing the working state, the fault time and the environmental factors causing faults of each device according to the fault records and the maintenance records; defining the state of the system according to the working state of each device, the fault time and the environmental factors causing the fault;
introducing a Markov chain, and calculating transition probability among states by combining fault characteristic data corresponding to the fault period of each device;
Constructing a hidden Markov model, importing the transition probability among the states into the hidden Markov model, training the hidden Markov model by utilizing the transition probability among the states and based on a Baum-Welch algorithm, and optimizing model parameters to maximize the possibility of observing a data sequence until the model parameters are converged to a preset value, so as to obtain a trained hidden Markov model;
Acquiring real-time operation parameters of a target solar hunting camera, importing the real-time operation parameters of the target solar hunting camera into the trained hidden Markov model, and predicting the most probable state sequence of each device in the target solar hunting camera through a forward algorithm or a Viterbi algorithm;
generating a fault prediction result according to the predicted state sequence, and pushing the fault prediction result to a preset user terminal.
It should be noted that, by acquiring the historical maintenance of the camera, fault records and fault characteristic data corresponding to the fault period, the working state and fault time of each device and environmental factors (such as temperature, humidity, rainwater invasion, etc.) possibly causing the fault are recorded; the state of the system is then defined, e.g., for a device, possible states are "normal operation", "performance degradation", "failure", etc. Calculating transition probabilities between states according to fault characteristic data, which means that the probability that a certain device transitions to other states at the next time point under the given current state needs to be calculated; a hidden Markov chain model is constructed using the collected data and the calculated transition probabilities, and this model may be a state transition matrix in which each element of the matrix represents a probability from one state to another. The current state or the most likely state sequence is predicted by a forward algorithm (Forward Algorithm) or a viterbi algorithm (Viterbi Algorithm), from which it can be determined whether the device is likely to fail or the severity of the failure. The method can utilize the historical data and the current observation data of the device to predict the future state of the camera, thereby identifying possible faults in advance, being beneficial to timely maintenance or replacement, reducing accidental interruption and loss, and being capable of implementing more effective preventive maintenance planning and avoiding unnecessary maintenance and cost expenditure by accurately predicting the possibility of the faults.
Furthermore, the method comprises the following steps:
Judging the fault state of the PIR sensor in the target solar hunting camera according to the fault prediction result, and controlling the target solar hunting camera to start if the fault state of the PIR sensor is a preset fault state;
acquiring real-time scene image information of a preset area at a preset time node through a target solar hunting camera, and constructing a first scene model diagram according to the real-time scene image information;
Acquiring historical scene image information when no living animal exists in a preset area based on the operation log information of the target solar hunting camera, and constructing a second scene model diagram according to the historical scene image information;
calculating the coincidence degree between the first scene model diagram and the second scene model diagram based on an ICP algorithm, and comparing the coincidence degree with a preset coincidence degree;
If the overlap ratio is greater than the preset overlap ratio, indicating that no living animal exists in the preset area of the current preset time node;
if the coincidence degree is not greater than the preset coincidence degree, indicating that living animals exist in a preset area of the current preset time node, aligning the first scene model image with the second scene model image based on a FAST algorithm, removing model parts overlapped with the first scene model image and the second scene model image after alignment, and reserving model parts which are not overlapped to obtain a living animal three-dimensional model image in the preset area of the current preset time node;
calculating the similarity between the three-dimensional model diagram of the living animal and each morphological feature model diagram in the knowledge graph based on an ICP algorithm to obtain a plurality of similarities; comparing the similarity with a preset similarity one by one;
if the similarity is not greater than the preset similarity, indicating that the living animal in the preset area is not a target prey, and generating a first identification result;
If at least one of the similarities is larger than the preset similarity, the living animal in the preset area is the target hunting object, a second identification result is generated, and continuous tracking is carried out on the target hunting object.
It should be noted that, if the fault state of the PIR sensor is a preset fault state, namely a complete failure fault state, after the PIR sensor fails, the camera cannot automatically shoot or record when an animal passes, so that the camera fails, and as the camera is located outdoors, a certain time is required for an maintainer to go to overhaul, and in this time, the camera loses the function of tracking and identifying the prey. In order to overcome this technical problem, another method for starting a hunting-object-recognition camera is proposed in this step, which is used for disabling devices in PIR sensors, so as to ensure that the camera can effectively perform hunting-object recognition tracking, and improve the reliability of the camera.
As shown in fig. 3, the second aspect of the present invention discloses a gun-ball integrated solar hunting camera management system based on intelligent recognition, which includes a memory 20 and a processor 60, wherein a gun-ball integrated solar hunting camera management method program is stored in the memory 20, and when the gun-ball integrated solar hunting camera management method program is executed by the processor 60, the following steps are implemented:
Acquiring a site characteristic three-dimensional model map of a target area, dividing the site characteristic three-dimensional model map into a plurality of sub-site characteristic three-dimensional model maps, and analyzing the terrain, sunlight and vegetation conditions of each sub-site characteristic three-dimensional model map to obtain a plurality of target installation sub-areas;
Installing the target hunting camera into the target installation subarea to complete the installation layout process of the target hunting camera; acquiring infrared radiation signal information in each preset area through a PIR sensor in the target hunting camera, and converting the infrared radiation signal information into preset type electric signals;
If the threshold value of the preset type electric signal is larger than the preset threshold value, the PIR sensor generates a trigger signal, a target hunting camera is started according to the trigger signal, video frame information of a preset area is obtained through the target hunting camera, and recognition processing is carried out on the video frame information to obtain a recognition result;
if the identification result is a first identification result, indicating that the target hunting object does not exist in the preset area, and controlling the target hunting camera to be closed; if the identification result is the second identification result, indicating that the target hunting object exists in the preset area, generating alarm information and pushing the alarm information to a preset user terminal.
In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. The above described device embodiments are only illustrative, e.g. the division of the units is only one logical function division, and there may be other divisions in practice, such as: multiple units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. In addition, the various components shown or discussed may be coupled or directly coupled or communicatively coupled to each other via some interface, whether indirectly coupled or communicatively coupled to devices or units, whether electrically, mechanically, or otherwise.
The units described above as separate components may or may not be physically separate, and components shown as units may or may not be physical units; can be located in one place or distributed to a plurality of network units; some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in each embodiment of the present invention may be integrated in one processing unit, or each unit may be separately used as one unit, or two or more units may be integrated in one unit; the integrated units may be implemented in hardware or in hardware plus software functional units.
Those of ordinary skill in the art will appreciate that: all or part of the steps for implementing the above method embodiments may be implemented by hardware related to program instructions, and the foregoing program may be stored in a computer readable storage medium, where the program, when executed, performs steps including the above method embodiments; and the aforementioned storage medium includes: a mobile storage device, a read-only memory (ROM), a random access memory (RAM, random Access Memory), a magnetic disk or optical disk, or the like, which can store program codes.
Or the above-described integrated units of the invention may be stored in a computer-readable storage medium if implemented in the form of software functional modules and sold or used as separate products. Based on such understanding, the technical solutions of the embodiments of the present invention may be embodied in essence or a part contributing to the prior art in the form of a software product stored in a storage medium, including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the methods of the embodiments of the present invention. And the aforementioned storage medium includes: a removable storage device, ROM, RAM, magnetic or optical disk, or other medium capable of storing program code.
The foregoing is merely illustrative embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily think about variations or substitutions within the technical scope of the present invention, and the invention should be covered. Therefore, the protection scope of the invention is subject to the protection scope of the claims.