The content of the invention
It is an object of the invention to provide identification different crowd carry out advertisement pushing method, storage device and mobile terminal,To solve at least one of drawbacks described above existing for prior art, compared to the robot for fixing or having track on the market at presentFor, there is the functions such as accurate dispensing, mobility, man-machine interaction, amusement, inquiry.
For the above-mentioned purpose, the technical solution adopted by the present invention is as follows：
A kind of method for identifying different crowd and carrying out advertisement pushing, it is characterised in that methods described includes：
A, face characteristic storehouse is established；
B, indoor map is built；
C, recognition of face；
D, personage's coordinate information is obtained；
E, coordinate information is transformed into indoor map relevant position；
F, motion planning and robot control；
G, corresponding advertisement plays.
Preferably, the step A gathers the feature samples of face and bone by using 3D cameras, through opencv'sEigenFace methods handle to obtain face characteristic storehouse.
Preferably, the step B scans whole room area by using laser radar remote-controlled robot, with reference to ros systemsThe slam algorithms of system, establish a complete indoor map.
Preferably, during the step C identifications, by setting investigative range, when someone is appeared in investigative range, start3D cameras obtain the image information and depth information of face, and by being contrasted with feature database, determine belonging to this personClassification.
Preferably, human depth's information that the step D is obtained by 3D cameras, can be inferred that what is currently identifiedPeople is in the direction of robot and the information of distance and position.
Preferably, the step E passes through the coordinate system that is converted into matching with map by the distance of people and directional information.
Preferably, the step F be specially by robot by Navigational Movements arrive from people front 1 meter at a distance from.
Preferably, the step G plays corresponding advertisement specifically by the people's classification information recognized before.
A kind of storage device, it is characterised in that the storage device is stored with computer program, the computer program energyEnough it is performed to realize the method for identification different crowd progress advertisement pushing as described above.
A kind of mobile terminal, it is characterised in that including：Processor, the memory being connected with the processor communication, it is describedMemory storage has computer program, and the computer program realizes identification different crowd described above when being executed by processorThe method for carrying out advertisement pushing；
The processor is used to call the computer program in the memory, is entered with the identification different crowd for performing described aboveThe method of row advertisement pushing.
Compared with prior art, the present invention at least has the advantages that：The present invention includes A, face characteristic storehouse is established；B, indoor map is built；C, recognition of face；D, personage's coordinate information is obtained；E, coordinate information is transformed into indoor map corresponding positionsPut；F, motion planning and robot control；G, corresponding advertisement plays, using above-mentioned steps, compared to fixed or rail on the market at presentFor the robot in road, there is the functions such as accurate dispensing, mobility, man-machine interaction, amusement, inquiry.
Figure of description
Fig. 1 is that present pre-ferred embodiments identify that different crowd carries out the flow chart of the method for advertisement pushing.
Embodiment
The present invention will be further described with reference to the accompanying drawings and examples.
As shown in figure 1, the identification different crowd of the present embodiment carries out method, storage device and the movement of advertisement pushing eventuallyEnd, it is characterised in that methods described includes：
A, face characteristic storehouse is established；
B, indoor map is built；
C, recognition of face；
D, personage's coordinate information is obtained；
E, coordinate information is transformed into indoor map relevant position；
F, motion planning and robot control；
G, corresponding advertisement plays.
In specific this method, human face identification and the foundation of feature database are specifically introduced first：
This method hardware components are carried out recognition of face, are worked as machine using eyes of the 3D cameras of light in ratio difficult to understand as robotWhen people detects that people is appeared in default effective scope（Such as 5 meters）, then face and bone are caught and gathered；Software is knownOther technology uses the opencv recognitions of face storehouse of famous open source software, and the inside includes feature recognition algorithms EigenFace,EigenFace is a kind of big data matrix recognition methods, implements fairly simple, and concrete principle and implementation are：
1）We gather the picture sample of the different peoples such as youth, child, old man, men and women, are numbered according to different types, willEveryone face image of training set all elongates a row, and they are grouped together into a big matrix A.Assuming that each faceImage is MxM sizes, then the dimension for pulling into each face sample after a row is exactly d=MxM sizes.Assuming that there is N number of face figurePicture, then sample matrix A dimension is exactly dxN.
2）All N number of faces are added up in corresponding dimension, then asks individual average, has just obtained one " averagelyFace ".
3）N number of image is all subtracted into that average face image, obtains the data matrix Φ of error image.
4）Calculate covariance matrix C=Φ Φ T.Eigenvalues Decomposition is carried out to it again.It can be obtained by desired characteristic vector（Eigenface）.
5）The image of training set image and test set is all projected in these characteristic vectors, then to each of test setImage finds arest neighbors or k neighbours in training set, is finally classified.So we have just established a faceFeature database.
The specific specific implementation step for introducing lower map structuring and movement navigation again is as follows：
Hardware components：Master control uses the high performance 64 ARM chips of ARM3399, and CPU frequency is up to 2.3GHz, 4G internal memories, 16GROM, operational capability are powerful；Software section：Map structuring part uses the Gmapping algorithms inside robot ROS systems, itsInput as 2D laser radars, odometer, IMU etc., by particle filter, one increment type map of formation, and real-time recorderThe coordinate position that people is currently located, when Robot Scanning completely individual room area, with also just establishing a complete indoor 2DFigure.
The depth data of people is recognized according to 3D cameras in robot, the position of current robot is being combined, according to TFTransformational relation can extrapolate position coordinates of the people in map, after the coordinate of people and the coordinate of robot determine, be led according to ROSBit path plans that motor moves to the position at 1 meter in front of people.
Path planning algorithm, using famous A* algorithms, concrete principle is：Initial point is reached the cost g (n) of the nodeCombine with the cost h (n) from the node to destination node and node is evaluated：
f(n)=g(n)+h(n)
Wherein：G (n) is represented from start node to node n path cost
H (n) represents the estimation band value of the lowest-cost paths from node n to destination node
F (n) is exactly the lowest costs estimate on solutions cost to destination node by node n
When being used for the path planning of mobile robot using A* methods, robot is first according to known environmental map and cooked upOne paths, moved then along this track, when robot sensor detects environmental information and original environmental informationWhen inconsistent, robot plans that the path from current location to target point is so circulated until robot reaches againTarget point either finds that but if the unreachable robots of target point move in dynamic environment or circumstances not known, machineDevice people is likely to fairly frequently run into current detection environmental information and the unmatched situation of priori environment information, and this just needsCarry out path plan again again planning algorithm be still the global search from current location to target point process, computingLarger is measured in planning period again, robot either selects to be stopped for new generation path or the road according to mistakeFootpath continues to move therefore, and quickly planning algorithm is that very important .A* methods represent map, grid using grid againGranularity is smaller, and the expression of barrier is also more accurate, but the scope of algorithm search can exponentially increase using changing simultaneouslyThe local paths planning method for entering Artificial Potential Field optimizes to A* methods, can effectively increase the grid granularity of A* methods,Reaching reduces the purpose of A* method operands.
Under last specific introduction
Advertisement pushing part：
The people information recognized is classified, such as old man, child, youth, man, the packet of female's various information, according to differencePacket, will be played out in the advertisement pushing of association to display screen.
Preferably, the step A gathers the feature samples of face and bone by using 3D cameras, through opencv'sEigenFace methods handle to obtain face characteristic storehouse.
Preferably, the step B scans whole room area by using laser radar remote-controlled robot, with reference to ros systemsThe slam algorithms of system, establish a complete indoor map.
Preferably, during the step C identifications, by setting investigative range, when someone is appeared in investigative range, start3D cameras obtain the image information and depth information of face, and by being contrasted with feature database, determine belonging to this personClassification.
Preferably, human depth's information that the step D is obtained by 3D cameras, can be inferred that what is currently identifiedPeople is in the direction of robot and the information of distance and position.
Preferably, the step E passes through the coordinate system that is converted into matching with map by the distance of people and directional information.
Preferably, the step F be specially by robot by Navigational Movements arrive from people front 1 meter at a distance from.
Preferably, the step G plays corresponding advertisement specifically by the people's classification information recognized before.
Specifically, that for example detect is minor child, then according to its age size, finds corresponding object for appreciationHave the advertisement of class, because child is relatively low for the repellence of toy, be bound to be attracted, then this advertisement will play phaseThe effect being a good；Meanwhile such as detect be 40 or so a middle-aged person, then people's cause of this age bracket is comparativelyIt is stable, then may to compare for health and value, the advertisement of some insurance classes or the advertisement of health care productses can be pushed, its is wideThe effect of announcement will be relatively preferable.
A kind of storage device, it is characterised in that the storage device is stored with computer program, the computer program energyEnough it is performed to realize the method for identification different crowd progress advertisement pushing as described above.
A kind of mobile terminal, it is characterised in that including：Processor, the memory being connected with the processor communication, it is describedMemory storage has computer program, and the computer program realizes identification different crowd described above when being executed by processorThe method for carrying out advertisement pushing；
The processor is used to call the computer program in the memory, is entered with the identification different crowd for performing described aboveThe method of row advertisement pushing.
In summary, the present invention includes A, face characteristic storehouse is established；B, indoor map is built；C, recognition of face；D, obtainPersonage's coordinate information；E, coordinate information is transformed into indoor map relevant position；F, motion planning and robot control；G, corresponding advertisementPlay, using above-mentioned steps, compared to fixing or having for the robot of track on the market at present, there is accurate dispensing, it is removableThe dynamic function such as property, man-machine interaction, amusement, inquiry.
The present invention is described in detail above by specific embodiment, these detailed description are only limited to helpSkilled artisan understands that present disclosure, can not be interpreted as limiting the scope of the invention.The guarantor of the present inventionShield scope is defined by the content of claims.