Summary of the invention
A kind of roads recognition method and device are provided in the embodiment of the present invention, cannot be continued accurately with solving the prior artThe problem of identifying road area in image.
In order to solve the above-mentioned technical problem, the embodiment of the invention discloses following technical solutions:
A kind of roads recognition method, which comprises
Obtain the reference picture in front of vehicle movement；
Receive the virtual controlling instruction of user's input, virtual controlling instruction includes at least: front-wheel pivot angle, throttle amount andOne of braking amount；
According to the driving trace line of the virtual controlling branch prediction vehicle；
The driving trace line is added on the reference picture；
Choose the pixel of multiple reflection roadway characteristics in the reference picture according to the driving trace line；
Regard each pixel of selection as a road sample, obtain include multiple road samples road sampleCollection；
Utilize the road area in reference picture described in all road specimen discernings in the road sample set.
Optionally,
The driving trace line includes two wheel trajectories lines；
The first predeterminable area is chosen in the interval region of two wheel trajectories lines；
Two the second predeterminable areas, track of vehicle line described in each corresponding one are taken according to two wheel trajectories line selectionsA second predeterminable area, and each second predeterminable area is using the corresponding wheel trajectories line as center line；
The road is chosen in the reference picture according to first predeterminable area and/or second predeterminable areaSample.
Optionally, the road in reference picture described in all road specimen discernings using in the road sample setRegion, comprising:
Default road model is established using all road samples in the road sample set；
Differentiate that the attribute of each pixel in the reference picture, the attribute are road using the default road modelRoad or non-rice habitats；
The picture of all reflection roadway characteristics in the reference picture is obtained according to the differentiation result of the default road modelVegetarian refreshments；
The roadway area in the reference picture is identified using the pixel of reflection roadway characteristics all in the reference pictureDomain.
Optionally, the method also includes:
The pixel of multiple reflection non-rice habitats features is chosen in the non-rice habitats predeterminable area of the reference picture；
Regard each pixel of selection as a non-rice habitats sample, obtain include multiple non-rice habitats samples non-roadRoad sample set；
Choose multiple pixels undetermined in the reference picture according to preset strategy；
Determine the category of each pixel undetermined respectively according to the road sample set and the non-rice habitats sample setProperty, the attribute is road or non-rice habitats；
The pixel reflection undetermined of each in the reference picture is differentiated respectively using the default road modelThe attribute；
For pixel undetermined described in each, if the differentiation result of the default road model and the pixel undeterminedThe attribute of point is consistent, determines that the differentiation result of the default road model is correct；
If the correctly corresponding pixel quantity undetermined of the differentiation result, with the pixel quantity undeterminedRatio is less than preset threshold, utilizes the incorrect corresponding pixel amendment undetermined of the differentiation result default roadModel.
Optionally,
The pixel of multiple reflection non-rice habitats features is chosen in the non-rice habitats predeterminable area of the reference picture；
Regard each pixel of selection as a non-rice habitats sample, obtain include multiple non-rice habitats samples non-roadRoad sample set；
Choose multiple pixels undetermined in the reference picture according to preset strategy；
Determine the category of each pixel undetermined respectively according to the road sample set and the non-rice habitats sample setProperty, the attribute is road or non-rice habitats；
According to all road samples in the road sample set, all non-rice habitats samples in the non-rice habitats sample setThe default road model is established with all pixels undetermined.
Optionally, the picture that multiple reflection non-rice habitats features are chosen in the non-rice habitats predeterminable area of the reference pictureVegetarian refreshments, comprising:
It is described non-in the selection of the top, the upper left corner and/or the upper right corner of the reference picture according to virtual controlling instructionRoad predeterminable area；
Choose pixel of multiple pixels as reflection non-rice habitats feature in the non-rice habitats predeterminable area.
Optionally, described to determine that each is described undetermined respectively according to the road sample set and the non-rice habitats sample setThe attribute of pixel, comprising:
Obtain the pixel characteristic of road sample described in each of described road sample set, in the non-rice habitats sample setEach non-rice habitats sample pixel characteristic, and, the pixel characteristic of each pixel undetermined；
The similar pixel of pixel characteristic is formed into a similar set, the similar set has multiple, any two instituteThe pixel characteristic for stating pixel in similar set is dissimilar, includes: that at least one is described undetermined in each similar setPixel, and, at least one pixel in the road sample set and the non-rice habitats sample set；
Judge whether the quantity of road sample described in each described similar set is greater than the non-rice habitats sample respectivelyQuantity；
If the quantity of the road sample is greater than the quantity of the non-rice habitats sample, the institute in the similar set is determinedHaving the pixel undetermined is the pixel for reflecting roadway characteristic；If the quantity of the road sample is less than the non-rice habitatsThe quantity of sample determines that all pixels undetermined in the similar set are the pixel for reflecting non-rice habitats feature.
Optionally,
Obtain the setting parameter for shooting the video camera of the reference picture；
Using video camera setting gain of parameter subject picture position in the reference picture in realityThe matching relationship between spatial position in space；
The driving trace line is superimposed upon on the reference picture according to the matching relationship.
A kind of road Identification device, applied to Vehicular body front installation camera vehicle, described device include acquiring unit,Trajectory unit, superpositing unit, sample unit and recognition unit:
The acquiring unit connects with the camera and the control equipment for the virtual controlling instruction for sending user's input respectivelyIt connects, for obtaining the reference picture in front of vehicle movement, and receives the virtual controlling instruction, the virtual controlling instruction is at leastIt include: one of front-wheel pivot angle and throttle amount, braking amount；
The trajectory unit is connect with the acquiring unit, for the traveling according to the virtual controlling branch prediction vehicleTrajectory line；
The superpositing unit is connect with the trajectory unit and the acquiring unit respectively, is used for the driving trace lineIt is added on the reference picture；
The sample unit is connect with the superpositing unit and the acquiring unit respectively, for according to the driving traceLine chooses the pixel of multiple reflection roadway characteristics in the reference picture；
Regard each pixel of selection as a road sample, obtain include multiple road samples road sampleCollection；
The recognition unit is connect with the sample unit, for utilizing all road samples in the road sample setIdentify the road area in the reference picture.
Optionally,
The trajectory unit includes wheel trajectories unit, and the wheel trajectories unit is for obtaining two wheel trajectories lines；
The sample unit includes the first predeterminable area selection unit, and the first predeterminable area selection unit is used for twoThe first predeterminable area is chosen in the interval region of wheel trajectories line described in item；
The sample unit further includes the second predeterminable area selection unit, and the second predeterminable area selection unit is used for rootTwo the second predeterminable areas are taken according to two wheel trajectories line selections, corresponding one second, track of vehicle line described in each defaultRegion, and each second predeterminable area is using the corresponding wheel trajectories line as center line；
The sample unit further includes choosing list with the first predeterminable area selection unit and the second predeterminable area respectivelyThe road sample selection unit of member connection, the road sample selection unit are used for according to first predeterminable area and/or instituteIt states the second predeterminable area and chooses the road sample in the reference picture.
Optionally, the recognition unit includes model unit, template(-let), road pixel point acquiring unit and road IdentificationUnit；
The model unit is used to establish default road model using all road samples in the road sample set；
The template(-let) is used to differentiate each pixel in the reference picture using the default road modelAttribute, the attribute are road or non-rice habitats；
The road pixel point acquiring unit is connect with the template(-let), for sentencing according to the default road modelOther result obtains the pixel of all reflection roadway characteristics in the reference picture；
The road Identification unit is connect with the road pixel point acquiring unit, for utilizing institute in the reference pictureThere is the pixel of reflection roadway characteristic to identify the road area in the reference picture.
Optionally,
The sample unit includes non-rice habitats sample unit, and the non-rice habitats sample unit is used in the reference pictureThe pixel of multiple reflection non-rice habitats features is chosen in non-rice habitats predeterminable area；By each pixel of selection be used as one it is non-Road sample, obtain include multiple non-rice habitats samples non-rice habitats sample set；
The recognition unit includes unit, determination unit, judgement unit, differentiation result determination unit and amendment list undeterminedMember；
The unit undetermined for choosing multiple pixels undetermined according to preset strategy in the reference picture；
The determination unit is connect with the non-rice habitats sample unit and the unit undetermined respectively, for according to the roadRoad sample set and the non-rice habitats sample set determine the attribute of each pixel undetermined respectively, the attribute be road orNon-rice habitats；
The judgement unit is connect with the template(-let), for differentiating the ginseng respectively using the default road modelExamine the attribute of the pixel reflection undetermined of each in image；
The differentiation result unit is connect with the determination unit and the judgement unit respectively, for being directed to each institutePixel undetermined is stated, when the differentiation result of the default road model is consistent with the attribute of the pixel undetermined, determines instituteThe differentiation result for stating default road model is correct；
The amending unit is connect with the differentiation result determination unit, for corresponding in the correct differentiation resultThe pixel quantity undetermined, when being less than preset threshold with the ratio of the pixel quantity undetermined, using incorrect describedDifferentiate that the corresponding pixel undetermined of result corrects the default road model.
Optionally,
The sample unit includes non-rice habitats sample unit, and the non-rice habitats sample unit is used in the reference pictureThe pixel of multiple reflection non-rice habitats features is chosen in non-rice habitats predeterminable area；
Regard each pixel of selection as a non-rice habitats sample, obtain include multiple non-rice habitats samples non-roadRoad sample set；
The recognition unit includes unit, determination unit and road model unit undetermined；
The unit undetermined for choosing multiple pixels undetermined according to preset strategy in the reference picture；
The determination unit is connect with the non-rice habitats sample unit and the unit undetermined respectively, for according to the roadRoad sample set and the non-rice habitats sample set determine the attribute of each pixel undetermined respectively, the attribute be road orNon-rice habitats；
The road model unit is connect with the non-rice habitats sample unit and the determination unit, for according to the roadAll non-rice habitats samples in all road samples, the non-rice habitats sample set and all pixels undetermined in the sample set of roadPoint establishes the default road model.
Optionally, the non-rice habitats sample unit includes non-rice habitats region selection unit；
Non-rice habitats region selection unit is used to be instructed according to the virtual controlling in the road image reference pictureTop, the upper left corner and/or the upper right corner choose the non-rice habitats predeterminable area；It chooses multiple in the non-rice habitats predeterminable areaPixel of the pixel as reflection non-rice habitats feature.
Optionally, the determination unit includes that pixel characteristic unit, set component units, judging unit and feature determine listMember；
The pixel characteristic unit is connect with the sample unit and the unit undetermined respectively, for obtaining the roadNon-rice habitats sample described in each of the pixel characteristic of road sample described in each of sample set, described non-rice habitats sample setThis pixel characteristic, and, the pixel characteristic of each pixel undetermined；
The set component units are connect with the pixel characteristic unit, for forming the similar pixel of pixel characteristicOne similar set, the similar set have multiple, the pixel characteristic dissmilarity of pixel in any two similar set, eachIt include: at least one described pixel undetermined in the similar set, and, the road sample set and the non-rice habitats sampleAt least one pixel in this collection；
The judging unit is connect with the set component units, for judging institute in each described similar set respectivelyWhether the quantity for stating road sample is greater than the quantity of the non-rice habitats sample；
The characteristics determining unit is connect with the judging unit, is greater than for the quantity in the road sample described non-When the quantity of road sample, determine that all pixels undetermined in the similar set are the pixel for reflecting roadway characteristicThe attributive character of point reflection is roadway characteristic；It is less than the quantity of the non-rice habitats sample in the quantity of the road sampleWhen, determine that all pixels undetermined in the similar set are the pixel for reflecting non-rice habitats feature.
Optionally, the superpositing unit includes camera parameter unit, matching relationship unit and image superimposition unit；
The camera parameter unit is used to obtain the setting parameter for the video camera for shooting the reference picture；
The matching relationship unit is connect with the camera parameter unit, for the setting gain of parameter quilt using video cameraShoot the object picture position of object and the matching relationship between the spatial position in real space in the reference picture；
Described image superpositing unit is connect with the matching relationship unit, for according to the matching relationship by the travelingTrajectory line is superimposed upon on the reference picture.
By above technical scheme as it can be seen that a kind of roads recognition method provided in an embodiment of the present invention and device, pass through acquisitionVehicle front reference picture and virtual vehicle target front-wheel pivot angle and vehicle target acceleration, obtain the dummy row of vehicleTrajectory line is sailed, and on a reference by the superposition of driving trace line, the row in vehicle future is intuitively known using reference picture as backgroundSail track.The pixel for choosing multiple reflection roadway characteristics in a reference image according to driving trace line is road sample, is utilizedRoad in the road specimen discerning reference picture of selection makes to travel in the road that road vehicle is identified referring to reference pictureRoad continues on road advance, and the embodiment of the present invention can constantly updated according to the virtual vehicle driving trace line of real-time updateReference picture in effectively choose reliable road sample, to continue to identify the road in updated reference picture.Due toRoad sample is chosen according to the real-time driving trace line of vehicle and the real-time reference picture of vehicle front, therefore, constantlyThe road sample of update can adapt to the continually changing road of vehicle front, to accurately know in the reference picture of vehicle frontOther road, with guiding vehicle according to be correctly oriented advance.
Specific embodiment
Technical solution in order to enable those skilled in the art to better understand the present invention, below in conjunction with of the invention realThe attached drawing in example is applied, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described implementationExample is only a part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, this field is commonTechnical staff's every other embodiment obtained without making creative work, all should belong to protection of the present inventionRange.
Fig. 1 is a kind of roads recognition method flow diagram provided in an embodiment of the present invention, is mainly comprised the steps that
Step S101: obtaining the reference picture of vehicle front, and, receive the virtual controlling instruction of user's input, whereinVirtual controlling instruction includes at least one of front-wheel pivot angle, throttle amount and braking amount, and certainly, virtual controlling instruction also can wrapInclude other instructions.
The embodiment of the present invention is applied to car body front end and is provided with the vehicle of video camera, such as unmanned vehicle, in following embodimentsBy taking unmanned vehicle as an example.Obtain the reference picture in front of the unmanned vehicle direction of motion of shot by camera.
Reference picture in whole embodiments of the present invention is the data of reference picture, by the pixel of reflection object featuresComposition, each pixel has pixel characteristic, for example, the pixel characteristic of each pixel is by coloration, brightness or other numerical tabularsShow.In an embodiment of the present invention, the pixel in reference picture can be divided into the pixel of reflection roadway characteristic and reflect non-roadThe pixel of road feature, wherein the attribute for reflecting the pixel of roadway characteristic is road, reflects the pixel of non-rice habitats featureAttribute is non-rice habitats.
In one particular embodiment of the present invention, virtual controlling instruction can pass through human-computer interaction device or other equipmentIt obtains, is that operator instructs according to the virtual controlling that trend of road is arranged, virtual controlling instruction does not directly control unmanned vehicleMotion state, and be only realize in following embodiments predict vehicle driving trace line.Wherein, human-computer interaction device can beSteering wheel, gas pedal and the brake pedal of emulation, other controlling equipments being also possible to based on mouse-keyboard, so as toIt is inputted by user, obtains the instruction of the virtual controlling such as front-wheel pivot angle, throttle amount and braking amount.
Step S102: according to the driving trace line of virtual controlling branch prediction vehicle.
On the basis of step S101 has obtained virtual controlling instruction, in conjunction with the default ginseng such as the distance between each wheelNumber obtains multiple tracing points of the unmanned vehicle within following a period of time according to vehicle dynamic model, and by multiple tracing pointIt is linked in sequence and constitutes the driving trace line of unmanned vehicle.
Step S103: on a reference by the superposition of driving trace line.
The driving trace line that step S102 is obtained is superimposed upon on the reference picture of step S101 acquisition, in order to subsequentRoad sample is chosen in step, and manipulates human-computer interaction device for operator and reference basis is provided.
In one particular embodiment of the present invention, firstly, obtaining the setting parameter of the video camera of shooting reference picture, exampleMounting height, the field range parameter of such as video camera, the setting parameter of acquired video camera can be the reality of video cameraParameter, or the parameter of user's input；Then, using the setting gain of parameter subject of video camera in reference pictureIn matching relationship between the spatial position in real space of picture position and subject.
Finally, driving trace line is superimposed on a reference according to matching relationship.
After obtaining matching relationship, the conversion of driving trace line is superimposed on a reference according to matching relationship.For example, such asShown in dotted line in Fig. 2 on road, when the driving trace using display display reference picture and superposition on a referenceWhen line, the driving trace line of unmanned vehicle can be intuitively shown on reference picture, reflected within unmanned vehicle following a period of timeRunning track on corresponding reference picture.
In one particular embodiment of the present invention, system passes the reference picture of superposition traveling trajectory line by wireless communicationIt transports to remote control computer and shows.Superimposed image of the operator towards real-time reception, according to the road in reference pictureThe vehicle driving trace line of environment and current predictive, learn current driving trace line whether with road area in reference pictureIt is consistent, and virtual controlling instruction is sent by human-computer interaction device when the two is inconsistent, makes to instruct based on virtual controllingThe driving trace line of generation can always be consistent with the road area in reference picture, i.e. traveling trajectory line can be always superimposedOn road area in a reference image.
Step S104: the pixel of multiple reflection roadway characteristics is chosen in a reference image according to driving trace line, and willChoose each pixel be used as a road sample, obtain include multiple road samples road sample set.
After being superimposed due to driving trace line with reference picture, the picture position of driving trace line in a reference image usually withReflect the pixel intersection of roadway characteristic in reference picture, the present embodiment is chosen multiple in a reference image according to driving trace lineReflect the pixel of roadway characteristic, and regard each pixel of selection as a road sample, obtains including multipleThe road sample set of road sample.
In one particular embodiment of the present invention, as shown in figure 3, driving trace line includes wheel trajectories line (vehicle in figureTwo preceding solid lines), wheel trajectories line is the two lines for reflecting front-wheel motion profile on vehicle.Between two wheel trajectories linesIt is chosen in septal area domain the first predeterminable area (in figure shown in A), also, chooses the second preset areas respectively according to two wheel trajectories linesDomain (in figure shown in B), for the second predeterminable area using corresponding wheel trajectories line as center line, every wheel trajectories line is one correspondingSecond predeterminable area.
First predeterminable area and the second predeterminable area are the region obtained by the physical location of object, according to above-mentioned objectPicture position and the matching relationship that is calculated of spatial position, by the physical location of the first predeterminable area and the second predeterminable areaIt is converted into the picture position in reference picture, and the first predeterminable area after superposition conversion and the second preset areas on a referenceDomain, this obtains the corresponding region (in such as figure shown in A ' and B ') of the first predeterminable area and the second predeterminable area in a reference image,Pixel of the reference picture in the corresponding region is essentially the pixel for reflecting roadway characteristic, therefore, in the corresponding regionSelected pixels point is reliable and accurate as the mode of road sample.
In addition, when practical application the present embodiment identifies road area in a reference image, it can be according to the first predeterminable areaRoad sample is chosen with the second predeterminable area, or chooses road sample according only to the first predeterminable area or the second predeterminable area, ifUnmanned vehicle will run the narrow passages that the road of process is slightly larger than wheel width for two, such as the transfer bridge temporarily built, orWhen road middle section grows the road of a large amount of weeds due to rolling for a long time without wheel, as shown in Figure 4, it is evident that by referring toSelected pixels point is that a large amount of non-rice habitats samples will necessarily be chosen for road sample by the mode of road sample in fixed area in imageThis, causes identification road mistake occur, therefore, in one particular embodiment of the present invention, can be default by only choosing secondRegion effectively avoids the occurrence of above-mentioned identification mistake without choosing the mode of the first predeterminable area, to guarantee the standard of road sampleTrue property.
Step S105: the road area in all road specimen discerning reference pictures in road sample set is utilized.
The road sample set obtained using step S104, it is determining with road sample in all pixels point in a reference imageThis similar pixel of pixel characteristic, these pixels similar with the pixel characteristic of road sample are anti-in reference pictureThe pixel for reflecting roadway characteristic just identifies reference in determining reference picture after the pixel of all reflection roadway characteristicsRoad area in image, unmanned vehicle can adjust operating parameter according to the road area identified, and continue to move.
In one particular embodiment of the present invention, operator can utilize human-computer interaction device according to the road identifiedVirtual controlling instruction is adjusted, to guarantee that the driving trace line of vehicle is always consistent with the road in reference picture.For example, ifOperator sees that the road in reference picture extends afield straight, then operator sets 0 degree for front-wheel pivot angle and increases oilMen Liang is straight line according to the driving trace line that front-wheel pivot angle and throttle amount obtain and extends farther out, makes to be added to reference to figure in this wayAs the trend of road in upper driving trace line and reference picture preferably coincide, convenient for the correctness of subsequent road sample collection.
The present embodiment using the virtual controlling branch prediction vehicle of user's input driving trace line, and by driving trace lineIt is superimposed upon on the reference picture in vehicle operation front, road sample set is chosen according to driving trace line on a reference, finallyAccording to the road area in each of road sample set road specimen discerning reference picture.It is chosen in this embodiment schemeRoad sample set is always chosen according to the actual conditions of road area in reference picture, and no matter vehicle is traveling in linear roadOn upper or crankcase ventilaton, road sample set can flexibly, be accurately chosen, to effectively avoid the occurrence of using selection with reference to figureSample is easy to falsely drop non-rice habitats sample for road sample as the prior art of road sample set in fixed area as inBad result.
In another embodiment of the present invention, as shown in figure 5, realizing that step S105 utilizes road sample in above-described embodimentRoad area in all road specimen discerning reference pictures of this concentration, including the following steps:
Step S501: default road model is established using all road samples in road sample set.
Since all road samples in road sample set are the sample for reflecting roadway characteristic, with these roadsOn the basis of the pixel point feature of sample, can there will be the pixel of similar pixel point feature to regard as with road sample in imageThe pixel for reflecting roadway characteristic, the biggish pixel of pixel feature difference in image with road sample is regarded as reflectingThe pixel of non-rice habitats feature.
In conjunction with existing image classification model, and using all road samples in road sample set as classification referring to basePlinth can be obtained the default road model in the present embodiment, preset road model by this, may recognize that all reflections in imageThe pixel of roadway characteristic and the pixel of reflection non-rice habitats feature.
Step S502: the attribute of each pixel in reference picture is differentiated using default road model, attribute is roadOr non-rice habitats.
Differentiate in reference picture that each pixel is reflection roadway characteristic or anti-using preset road modelNon-rice habitats feature is reflected, if certain pixel is the pixel for reflecting roadway characteristic, the attribute of the pixel is road；If certainPixel is the pixel for reflecting non-rice habitats feature, then the attribute of the pixel is non-rice habitats.
Step S503: the pixel that all properties in reference picture are road is obtained according to the differentiation result of default road modelPoint.
Step S504: the roadway area in reference picture is identified using the pixel that all properties in reference picture are roadDomain.
All properties are the road area in the set as reference picture of the pixel composition of road in reference picture.
In another embodiment of the present invention, as shown in fig. 6, the default road model in above-described embodiment can be by followingSeveral steps obtain.
Step S1051: choosing the pixel of multiple reflection non-rice habitats features in the non-rice habitats predeterminable area of reference picture,And it regard each pixel of selection as a non-rice habitats sample, obtain the non-rice habitats sample comprising multiple non-rice habitats samplesCollection.
In one particular embodiment of the present invention, it is instructed according to virtual controlling in the upper left corner and/or the right side of reference pictureChoose non-rice habitats predeterminable area in upper angle.For example, as shown in fig. 7, after the installation site and posture of vehicle-mounted vidicon are fixed, due toThe field range of video camera is limited, and the upper part in reference picture may be sky, the ground of distant place or object above ground levelDeng.Therefore, the pixel non-rice habitats sample on image in part, the upper left corner and upper right comer region can be chosen in a reference image,As shown in fig. 7, with the region that dotted line is filled be the non-rice habitats predeterminable area selected, wherein w in figurem、wl、wrIt is respectively selectedThe height of reference picture upper area, the reference picture upper left corner and reference picture upper right comer region width, ll、lrRespectivelyThe length in the reference picture upper left corner and reference picture upper right comer region.
Also, when vehicle turns to the left, since the movement speed of right side wheels is greater than left side wheel, l is enabledlSubtractSmall, lrIncrease；When vehicle turns to the right, since the movement speed of right side wheels is greater than left side wheel, l is enabledlIncrease, lrReduce, to meet the actual conditions of vehicle driving.
Using above-mentioned selected region as the non-rice habitats predeterminable area of reference picture, non-rice habitats predeterminable area is superimposed upon ginsengExamine on image, specific stacked system and when choosing road sample by the first predeterminable area and the second predeterminable area and reference pictureThe mode of superposition is similar, and details are not described herein again, chooses reflection non-rice habitats on a reference based on non-rice habitats predeterminable areaThe pixel of feature is as non-rice habitats sample.
Step S1052: multiple pixels undetermined are chosen in a reference image according to preset strategy；
For example, can to reference picture interval sampling, i.e., every several pixels acquire a pixel, also, with everyCentered on the pixel of acquisition, the pixel characteristic of the pixel periphery preset quantity pixel is obtained, according to peripheral image vegetarian refreshmentsPixel characteristic determine the pixel characteristic of the pixel, and using the pixel as its own and periphery present count can be representedMeasure the pixel undetermined of a pixel.Choose multiple pixels undetermined, each picture undetermined in a reference image in the manner described aboveVegetarian refreshments can represent the pixel characteristic of preset quantity pixel around it.It, can after the attribute for determining certain pixel undeterminedThink that the attribute for all pixels point that the pixel undetermined represents is consistent with the attribute of the pixel.
Step S1053: non-according to each of each of road sample set road sample and non-rice habitats sample setRoad sample determines that the attribute of each pixel reflection undetermined, attribute are road or non-rice habitats respectively.
In one particular embodiment of the present invention, as shown in figure 8, completing step S1053 by following steps:
Step S531: the pixel characteristic of each of road sample set road sample is obtained, in non-rice habitats sample setThe pixel characteristic of each non-rice habitats sample, and, the pixel characteristic of each pixel undetermined.
Step S532: forming a similar set for the similar pixel of pixel characteristic, similar set have it is multiple, any twoThe pixel characteristic of pixel is dissimilar in a similar set, includes: at least one pixel undetermined in each similar set, withAnd at least one pixel in road sample set and non-rice habitats sample set.Analyze each of road sample set roadEach of the pixel characteristic of sample, non-rice habitats the sample set pixel characteristic of non-rice habitats sample and each pixel undeterminedThe pixel characteristic of point, pixel characteristic includes at least the parameters such as brightness, the coloration of pixel, by the similar pixel group of pixel characteristicPixel at an independent similar set, also, pixel characteristic dissmilarity belongs to different similar sets.For example, oneIt include similar 71 pixels of pixel characteristic in similar set, wherein including 50 pixels undetermined, 20 road samples and 1A non-rice habitats sample.
Step S533: judge whether the quantity of road sample in each similar set is greater than the number of non-rice habitats sample respectivelyAmount.
The quantity of road sample and the quantity of non-rice habitats sample in the similar set that above-mentioned steps S532 is obtained are counted, pointDo not judge whether the quantity of road sample in each similar set is greater than the quantity of non-rice habitats sample.
Step S534: if the quantity of road sample be greater than non-rice habitats sample quantity, determine in the similar set toThe attribute for determining pixel is road.
Step S535: it if the quantity of road sample is less than the quantity of non-rice habitats sample, determines undetermined in similar setThe attribute of pixel is non-rice habitats.In this way, it determines the attribute of each pixel undetermined, that is, determines that each is undeterminedPixel is reflection roadway characteristic or reflection non-rice habitats feature.
It wherein, is not the picture undetermined in similar set if the quantity of road sample is equal to the quantity of non-rice habitats sampleVegetarian refreshments determines attribute, gives up the pixel undetermined in the similar set.
Step S1054: according to the non-road of each of each of road sample set road sample, non-rice habitats sample setRoad sample and each pixel undetermined establish default road model.
Pixel is had determined that according in each of non-rice habitats sample set non-rice habitats sample in step S1051, step S1053Each of road sample set road sample in the pixel and above-described embodiment undetermined of feature establishes default road mouldType, for example, default road model can be also possible to high using mixing to use support vector machines for the classifier algorithm of representativeThe disaggregated model that this model establishes road sample.
In another embodiment of the present invention, it is contemplated that with the operation of unmanned vehicle, road like that unmanned vehicle is runCondition may change, for example, unmanned vehicle runs to dirt road by asphalt road, at this point, the road sample obtained by former reference pictureThe default road model of this and the foundation of non-rice habitats sample may be not suitable for identifying road, therefore, this reality in current reference imageIt applies example and provides a kind of scheme of default road model of real-time update, as shown in figure 9, the program mainly includes following stepIt is rapid:
Step S201: choosing the pixel of multiple reflection non-rice habitats features in the non-rice habitats predeterminable area of reference picture,And it regard each pixel of selection as a non-rice habitats sample, obtain the non-rice habitats sample comprising multiple non-rice habitats samplesCollection.
Step S202: multiple pixels undetermined are chosen in a reference image according to preset strategy.
Step S203: the category of each pixel reflection undetermined is determined respectively according to road sample set and non-rice habitats sample setProperty, attribute is road or non-rice habitats.
Wherein, above-mentioned steps S201, step S202 and step S203 respectively with step S5031, step in above-described embodimentS5032 is similar with step S5033, and details are not described herein again.
Step S204: differentiate the attribute of each pixel undetermined respectively using default road model.
Judge the attribute of each pixel undetermined respectively using default road model, that is, judges each pixel undeterminedAttribute be road or non-rice habitats, obtain default road model to the differentiation result of each pixel undetermined.
Step S205: judge the differentiations result of default road model and fixed pixel undetermined attribute whether oneIt causes.
Step S206: being directed to each pixel undetermined, if the differentiation result of above-mentioned default road model and above-mentioned realityThe attribute for applying the pixel undetermined determined in step S203 in example is consistent, determines that the differentiation result of default road model is correct.ExampleSuch as, the attribute of some pixel undetermined is road, and default road model differentiates that the attribute of the pixel undetermined is also road, is saidThe differentiation result of bright default road model is consistent with the actual attribute of pixel undetermined, it may be determined that is directed to the pixel undetermined, in advanceIf the differentiation of road model is the result is that correctly.
Step S207: being directed to each pixel undetermined, if the differentiation result of above-mentioned default road model and above-mentioned realityThe attribute for applying in example the pixel undetermined determined in step S203 is inconsistent, determines the differentiation result of default road model not justReally.For example, the attribute of certain pixel undetermined is road, and the attribute of the pixel undetermined is determined as non-road by default road modelTherefore the differentiation of road model is preset the result is that incorrect in road.
Step S208: the corresponding pixel quantity undetermined of judicious differentiation result, and the ratio of pixel quantity undeterminedWhether value is less than preset threshold.
Step S209: if the quantity of the corresponding pixel undetermined of result is correctly differentiated, with all pixel numbers undeterminedThe ratio of amount is less than preset threshold, corrects default road model using the corresponding pixel undetermined of incorrect differentiation result.ExampleSuch as, if pixel quantity undetermined is 1000, and after being differentiated using default road model to each pixel undetermined,Determining has the differentiation of 600 pixels the result is that correctly, differentiating the correct pixel quantity 600 of result and all pixels undeterminedPoint quantity 1000 ratio be 0.6, if set preset threshold be 0.99, obtain preset road model differentiation result it is incorrect400 pixels undetermined, and default road model is corrected using this 400 pixels undetermined, for example, using replacement basic pixelThe modes such as the attribute of point or amendment original basis pixel correct default road model.
Step S210: if the quantity of the corresponding pixel undetermined of result is correctly differentiated, with all pixel numbers undeterminedThe ratio of amount is not less than preset threshold, then does not correct default road model.
The basic principle of default road model is successively to analyze each pixel in reference picture, and attribute is trueFixed pixel obtains the attribute of each pixel in reference picture as basic pixel, that is, obtaining with attribute is roadPixel there is the pixel of similar pixel feature, and determine that the attribute of pixel with similar pixel feature is similarlyRoad.
Therefore, if in default road model the attribute of certain basic pixel point be it is wrong, will lead to default road mouldType cannot be appropriately determined the attribute of pixel undetermined similar with the pixel characteristic of the basic pixel point, that is, default road mould occursThe differentiation of type is the result is that incorrect situation.
In the present embodiment, default road is corrected using the incorrect pixel undetermined of the differentiation result of default road modelRoad model, and only amendment causes pixel undetermined to differentiate the incorrect basic pixel point of result, including replacement basic pixel pointOr the attribute etc. of amendment original basis pixel, it can avoid being modified to not needing modified basic pixel point, structure be effectively reducedBuild the computing resource and calculate the time that default road model occupies.For example, in a specific embodiment, acquisition causes to belong to certainProperty be road pixel undetermined attribute misjudgement be non-rice habitats basic pixel point, constructed based on the basic pixel point singleGauss model is divided into the distribution that certain original model of representative is beyond expression to express the mistake, on this basis, by existing passIt is combined together in the mixed Gauss model of road and several new single Gauss models, re-using E-M algorithm constructs newMixed Gauss model is to update default road model.
Figure 10 is a kind of schematic diagram of road Identification device provided in an embodiment of the present invention, is taken the photograph applied to Vehicular body front installationAs the vehicle of head, which is characterized in that device includes that acquiring unit 1, trajectory unit 2, superpositing unit 3, sample unit 4 and identification are singleMember 5:
Acquiring unit 1 is connect with camera and the control equipment for the virtual controlling instruction for sending user's input respectively, is used forObtain vehicle movement in front of reference picture, and receive user input virtual controlling instruction, wherein virtual controlling instruct toIt less include: wheel pivot angle, throttle amount and braking amount etc.；
In one particular embodiment of the present invention, control equipment can be human-computer interaction device, and human-computer interaction device canTo be steering wheel, gas pedal and the brake pedal of emulation, other controlling equipments being also possible to based on mouse-keyboard, behaviourWork person obtains virtual controlling instruction by human-computer interaction device.
Trajectory unit 2 is connect with acquiring unit 1, for the driving trace line according to virtual controlling branch prediction vehicle；
Superpositing unit 3 is connect with trajectory unit 2 and acquiring unit 1 respectively, for being added to driving trace line with reference to figureAs upper；
Sample unit 4 is connect with superpositing unit 3 and acquiring unit 1 respectively, is used for according to driving trace line in reference pictureThe middle pixel for choosing multiple reflection roadway characteristics, and it regard each pixel of selection as a road sample, it is wrappedRoad set of stereotypes containing multiple road samples；
Recognition unit 5 is connect with sample unit 4, for utilizing all road pattern recognitions in road sample set with reference to figureRoad area as in.
In another embodiment of the present invention, the trajectory unit 2 in above-described embodiment includes vehicle front wheel track unit2, vehicle front wheel track unit 2 is for obtaining wheel trajectories line；
Sample unit 4 includes the first predeterminable area selection unit, and the first predeterminable area selection unit is used in two wheelsThe first predeterminable area is chosen in the interval region of trajectory line；
Sample unit 4 is including further including the second predeterminable area selection unit, and the second predeterminable area selection unit is for basisTwo wheel trajectories line selections take two the second predeterminable areas, and each track of vehicle line corresponds to second predeterminable area, andEach second predeterminable area is using corresponding wheel trajectories line as center line；
Sample unit 4 further includes connecting respectively with the first predeterminable area selection unit and the second predeterminable area selection unitRoad sample selection unit, road sample selection unit according to the first predeterminable area and/or the second predeterminable area for referring toRoad sample is chosen in image.
In another embodiment of the present invention, the recognition unit 5 in above-described embodiment includes model unit, attribute listMember, road pixel point acquiring unit and road Identification unit；
The model unit is used to establish default road model using all road samples in the road sample set；
Template(-let) is used to differentiate using default road model that the attribute of each pixel in reference picture, attribute to beRoad or non-rice habitats；
Road pixel point acquiring unit is connect with template(-let), for obtaining ginseng according to the differentiation result for presetting road modelExamine the pixel of all reflection roadway characteristics in image；
Road Identification unit is connect with road pixel point acquiring unit, for special using reflection roads all in reference pictureRoad area in the pixel identification reference picture of sign.
In another embodiment of the present invention, the sample unit 4 in above-described embodiment includes non-rice habitats sample unit, non-Road sample unit is used to choose the pixel of multiple reflection non-rice habitats features in the non-rice habitats predeterminable area of reference picture, andIt regard each pixel of selection as non-rice habitats sample, obtains the non-rice habitats sample set comprising multiple non-rice habitats samples；
Recognition unit 5 includes unit undetermined, determination unit, judgement unit, differentiates result determination unit and amending unit；
Unit undetermined for choosing multiple pixels undetermined according to preset strategy in a reference image；
Determination unit is connect with non-rice habitats sample unit and unit undetermined respectively, for according to road sample set and non-rice habitatsSample set determines that the attribute of each pixel undetermined, attribute are road or non-rice habitats respectively；
Judgement unit is connect with template(-let), for differentiating that each in reference picture waits for respectively using default road modelDetermine the attribute of pixel；
Differentiate that result unit is connect with determination unit and judgement unit respectively, for being directed to each pixel undetermined,When the differentiation result of default road model is consistent with the attribute of pixel undetermined, the differentiation result of default road model is being determined justReally；
Amending unit is connect with result determination unit is differentiated, for correctly differentiating the corresponding pixel number undetermined of resultAmount utilizes the corresponding pixel undetermined of incorrect differentiation result when being less than preset threshold with the ratio of pixel quantity undeterminedCorrect default road model.
In another embodiment of the present invention, the sample unit 4 in above-described embodiment includes non-rice habitats sample unit, non-The pixel that road sample unit is used to choose multiple reflection non-rice habitats features in the non-rice habitats predeterminable area of reference picture is madeFor non-rice habitats sample；
Recognition unit 5 includes unit undetermined, determination unit and road model unit；
Unit undetermined for choosing multiple pixels undetermined according to preset strategy in a reference image；
Determination unit is connect with non-rice habitats sample unit and unit undetermined respectively, for according to road sample set and non-rice habitatsSample set determines that the attribute of each pixel undetermined, attribute are road or non-rice habitats respectively；
Road model unit is connect with non-rice habitats sample unit and determination unit, for according to road sample, non-rice habitats sampleThis and pixel undetermined establish default road model.
In another embodiment of the present invention, the non-rice habitats sample unit in above-described embodiment includes the choosing of non-rice habitats regionTake unit；
Non-rice habitats region selection unit is used to be instructed according to virtual controlling on the top, the upper left corner and/or the right side of reference pictureNon-rice habitats predeterminable area is chosen at upper angle, chooses picture of multiple pixels as reflection non-rice habitats feature in non-rice habitats predeterminable areaVegetarian refreshments.
In another embodiment of the present invention, the determination unit in above-described embodiment includes pixel characteristic unit, setComponent units, judging unit and characteristics determining unit；
Pixel characteristic unit is connect with sample unit 4 and unit undetermined respectively, each in road sample set for obtainingEach of the pixel characteristic of a road sample, the non-rice habitats sample set pixel characteristic of non-rice habitats sample and each wait forDetermine the pixel characteristic of pixel；
Set component units connects with pixel characteristic unit, are used for similar by pixel characteristic similar pixel composition oneSet, similar set have multiple, and the pixel characteristic of pixel is dissimilar in any two similar set, in each similar setIt include: at least one pixel undetermined, and, at least one pixel in road sample set and non-rice habitats sample set；
Judging unit is connect with set component units, for judging the quantity of road sample in each similar set respectivelyWhether the quantity of non-rice habitats sample is greater than；
Characteristics determining unit is connect with judging unit, and the quantity of non-rice habitats sample is greater than for the quantity in road sampleWhen, determine that the attribute of the pixel undetermined in set is road；It is not more than the quantity of non-rice habitats sample in the quantity of road sampleWhen, determine that the attribute of the pixel undetermined in set is non-rice habitats.
In another embodiment of the present invention, the superpositing unit 3 in above-described embodiment includes camera parameter unit, matchingRelation unit and image superimposition unit；
Camera parameter unit is used to obtain the setting parameter of the video camera of shooting reference picture；
Matching relationship unit is connect with camera parameter unit, for the setting gain of parameter subject using video cameraThe picture position of object and the matching relationship between the spatial position in real space in a reference image；
Image superimposition unit 3 is connect with matching relationship unit, for driving trace line to be superimposed upon ginseng according to matching relationshipIt examines on image.
It should be noted that, in this document, the relational terms of such as " first " and " second " or the like are used merely to oneA entity or operation with another entity or operate distinguish, without necessarily requiring or implying these entities or operation itBetween there are any actual relationship or orders.Moreover, the terms "include", "comprise" or its any other variant are intended toCover non-exclusive inclusion, so that the process, method, article or equipment for including a series of elements not only includes thoseElement, but also including other elements that are not explicitly listed, or further include for this process, method, article or settingStandby intrinsic element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded thatThere is also other identical elements in the process, method, article or apparatus that includes the element.
The above is only a specific embodiment of the invention, is made skilled artisans appreciate that or realizing this hairIt is bright.Various modifications to these embodiments will be apparent to one skilled in the art, as defined hereinGeneral Principle can be realized in other embodiments without departing from the spirit or scope of the present invention.Therefore, of the inventionIt is not intended to be limited to the embodiments shown herein, and is to fit to and the principles and novel features disclosed herein phase oneThe widest scope of cause.