CN103810475B - A kind of object recognition methods and device - Google Patents

Abstract

The embodiment of the invention discloses a kind of object recognition methods and device.The method includes：At least two images and image taking information that acquisition is obtained from least two angle shots to object；Object identification is carried out at least two images respectively using setting recognizer, the first image and the second image are determined from least two images according to recognition result；Determine the fisrt feature point on object region in the first image, calculate attributive character of the fisrt feature o'clock in the first image；Second feature point in second image corresponding to fisrt feature point position is determined according to image taking information, attributive character of the second feature o'clock in the second image is calculated；According to the similarity between the attributive character of fisrt feature point and second feature point, the accuracy of object recognition result in the first image and/or the second image is obtained.The embodiment of the present invention realize to there is geometric deformation multiple image in object identification, improve object recognition result accuracy.

Description

Target object identification method and device

Technical Field

The embodiment of the invention relates to the technical field of image recognition, in particular to a target object recognition method and device.

Background

At present, in the field of image recognition technology, the recognition of a target object becomes a hotspot, and especially, the recognition of a traffic sign in a collected image becomes a necessary task for a vehicle-mounted system. The traffic sign is used as an indicator symbol and is important for the correct driving and traffic safety of the driver.

In the prior art, one identification technique is object identification based on a single acquired image. However, due to complex illumination changes in natural environments, different observation angles increase great difficulty in identifying the target object, and it is very difficult to achieve high identification rate under different observation conditions in practical use.

The other identification technology is to perform linear fusion on a plurality of images in the acquired video sequence to obtain a fused image and perform target object identification on the fused image. However, on one hand, the linear fusion can result in the averaging of the recognition results of the target object, and the recognition rate of the target object is reduced to a certain extent; on the other hand, the method only has a better recognition result for the target object with smaller geometric deformation in the adjacent frame images of the video sequence.

Disclosure of Invention

The embodiment of the invention provides a target object identification method and device, which are used for identifying a target object with larger geometric deformation in a plurality of images and improving the accuracy of a target object identification result.

In a first aspect, an embodiment of the present invention provides a target object identification method, where the method includes:

acquiring at least two images obtained by shooting a target object from at least two angles and image shooting information thereof, wherein the image shooting information comprises a shooting angle, a shooting distance and shooting equipment parameters;

respectively identifying the target objects of the at least two images by adopting a set identification algorithm, and determining a first image and a second image from the at least two images according to an identification result;

determining a first feature point on an area where a target object is located in the first image, and calculating attribute features of the first feature point in the first image;

determining a second feature point corresponding to the position of the first feature point in the second image according to the image shooting information, and calculating the attribute feature of the second feature point in the second image;

and obtaining the accuracy of the target object identification result in the first image and/or the second image according to the similarity between the attribute features of the first feature points and the attribute features of the second feature points.

In a second aspect, an embodiment of the present invention further provides an object identification apparatus, where the apparatus includes:

acquiring at least two images obtained by shooting a target object from at least two angles and image shooting information thereof, wherein the image shooting information comprises a shooting angle, a shooting distance and shooting equipment parameters;

respectively identifying the target objects of the at least two images by adopting a set identification algorithm, and determining a first image and a second image from the at least two images according to an identification result;

determining a first feature point on an area where a target object is located in the first image, and calculating attribute features of the first feature point in the first image;

determining a second feature point corresponding to the position of the first feature point in the second image according to the image shooting information, and calculating the attribute feature of the second feature point in the second image;

and obtaining the accuracy of the target object identification result in the first image and/or the second image according to the similarity between the attribute features of the first feature points and the attribute features of the second feature points.

According to the technical scheme provided by the invention, on one hand, based on the shooting angle relation between two images, the target object recognition result of one image is utilized to correct the target object recognition result of the other image, and the accuracy of the target object recognition result in the first image and/or the second image is obtained according to the corrected target object recognition result, so that the overall target object recognition rate is greatly improved; on the other hand, the mapping relation between the positions of the characteristic points of the two images is found by utilizing the image shooting information, so that the identification of the target object in the image with large geometric deformation can be realized from the imaging model of the image.

Drawings

Fig. 1 is a schematic flowchart of a target object identification method according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of an embodiment of the present invention, wherein the object is photographed from two angles;

fig. 3 is a schematic diagram illustrating a display of three images of a target object taken from different angles according to an embodiment of the present invention;

fig. 4 is a schematic flowchart of a target object identification method according to a second embodiment of the present invention;

fig. 5 is a schematic structural diagram of a target recognition device according to a third embodiment of the present invention;

fig. 6 is a schematic diagram of an imaging model of a camera according to an embodiment of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting of the invention. It should be further noted that, for the convenience of description, only some of the structures related to the present invention are shown in the drawings, not all of the structures.

Example one

Fig. 1 is a schematic flowchart of a target object identification method according to an embodiment of the present invention. The embodiment can be applied to the situation that after a certain target object in a real scene is subjected to image acquisition from different shooting angles, the target object in a plurality of acquired images is identified. For example, a navigator mounted on a vehicle captures traffic signs from different angles during traveling, and recognizes a plurality of images. The method may be performed by an object recognition apparatus, which is implemented by software and/or hardware. Referring to fig. 1, the target object identification method specifically includes the following operations:

110. at least two images obtained by shooting a target object from at least two angles and image shooting information thereof are obtained.

In this embodiment, the target object may be a traffic sign, a license plate, a face, or the like; the image capturing information includes a capturing angle, a capturing distance, and a capturing apparatus parameter. The photographing apparatus parameters are related to the internal structure of the photographing apparatus. The image capture information may be read directly from the capture device locally for parameters or determined in real time from the captured image. For example, the position and angle at which each image is taken can be acquired by a GPS positioning device.

120. And respectively identifying the target objects of the at least two images by adopting a set identification algorithm, and determining a first image and a second image from the at least two images according to the identification result.

After acquiring the plurality of images, the target recognition apparatus may first perform target recognition on the acquired single images respectively by using a preset recognition algorithm, and then perform a category identification operation on the images according to a recognition result to determine a first image and a second image of the at least two images.

The first image and the second image are images to be compared later, and in order to take the difference of multiple angles into consideration, the images with large difference or large difference of angles are preferably selected according to the recognition result to be compared. The images to be compared are not limited to two images, and may be two or more images, which are compared separately. Further, the target object recognition device determines the first image and the second image from the at least two images according to the recognition result, which may specifically be: an image capable of acquiring the recognition result is set as a first image, and an image incapable of acquiring the recognition result is set as a second image.

In a specific implementation manner of this embodiment, the target recognizing apparatus, which performs target recognition on the acquired image by using a preset recognition algorithm, may include: extracting features of a target object within at least one image using one or more filters; aggregating and combining the extracted features into a score of possible presence of the target object within the image; fusing scores corresponding to the plurality of images to obtain fusion scores; if the fusion score indicates that the object is likely to appear within a region of the image, the model is aligned to the region to determine whether the region in the image represents the object. The specific details of the identification algorithm can be found in the traffic sign detection algorithm in the patent with publication number "CN 101023436".

It should be understood by those skilled in the art that the set recognition algorithm may also be other algorithms for recognizing the target object, and this embodiment does not limit this. Thus, the first image and the second image of the at least two images may also be determined in other ways. For example, when the recognition result of each image is represented by the magnitude of the probability that the target object exists, an image corresponding to a probability equal to or higher than a set threshold value is regarded as a first image, and an image corresponding to a probability lower than the set threshold value is regarded as a second image.

130. And determining a first characteristic point on the area where the target object is located in the first image, and calculating the attribute characteristic of the first characteristic point in the first image.

The process of determining the first feature point may specifically be: and calculating the attribute characteristics of at least one point on the region where the target object is located in the first image, and taking the point with the attribute characteristics meeting the set conditions as a first characteristic point.

In this embodiment, the area where the target object is located is determined by the result of the target object recognition performed on the first image, and the area may be a target object area composed of points corresponding to the target object recognized in the first image, or may be a shape area of the target object area, such as a minimum circumscribed rectangle or a circle.

The determination of the setting condition is associated with the feature of the target object itself. For example, when the target object is a traffic sign, since the edge characteristics of the traffic sign are obvious and texture rarely exists, the first feature point may be an edge point, the attribute feature may include a gray gradient vector, and the point satisfying the setting condition may be a point in the first image where a modulus of the gray gradient vector is equal to or greater than a set value; when the target object is a license plate, since color information of the license plate is significant, the first feature point may be a point having set color information, the attribute feature may include an average value of the color information (gray values on R, G, B three components) of all points in a neighborhood window centered on the point and a variance thereof, and the point satisfying the set condition may be a point in the first image where the average value falls within a set first threshold range and the variance thereof falls within a set second threshold range.

140. And determining a second feature point corresponding to the position of the first feature point in the second image according to the image shooting information, and calculating the attribute feature of the second feature point in the second image.

Fig. 2 is a schematic diagram of photographing an object from two angles according to an embodiment of the present invention. As shown in fig. 2, the photographing apparatus photographs the object 210 at a first physical location 220 and a second physical location 230, respectively. Fig. 3 is a schematic display diagram of three images of a target object taken from different angles according to an embodiment of the present invention. Referring to fig. 3, the three images are: the image capturing apparatus includes a first captured image 310 captured from a first capturing angle with respect to a target object 300 in a real scene, a second captured image 320 captured from a second capturing angle with respect to the target object 300 in the real scene, and a third captured image 330 captured from a third capturing angle with respect to the target object 300 in the real scene. As can be seen, the target object in the first captured image 310 and the third captured image 330 is geometrically deformed compared to the target object 300 in the real scene.

After acquiring at least two images obtained by shooting a target object from at least two angles and image shooting information thereof, the target object recognition device can obtain a mapping relation of an imaging area of the same target object between the images according to the shooting angle, the shooting distance and the shooting equipment parameters of each image, namely, a corresponding position of the same target object in other images can be deduced from the acquired image.

Let two images be L1 and L2 with imaging matrices of M1 and M2, then the same point [ X YZ ] in the world coordinate system is referenced]' (e.g., a point on an object in a real scene) coordinates in image L1Comprises the following steps:

for the same point [ X Y Z ] in the world coordinate system]' coordinates in image L2Comprises the following steps:

so for the same point [ X Y Z ] in the world coordinate system]', coordinates in the image L1With coordinates in image L2The mapping relationship between the two is as follows:

wherein, M1 and M2 are both imaging matrices, and the imaging matrices can be divided into products of internal reference matrices and external reference matrices, that is: m1 ═ Ma11 ═ Ma 12; m2 Ma21 Ma 22.

If the image L1 and the image L2 are by the same photographing apparatus (e.g., camera), the internal reference matrix Ma11 and the internal reference matrix Ma21 are equal, and at this time,

accordingly, the process of determining a second feature point in the second image corresponding to the position of the first feature point may include:

obtaining a first internal reference matrix of the first shooting device according to shooting device parameters in the image shooting information corresponding to the first image, and obtaining a first external reference matrix of the first shooting device according to a shooting angle and a shooting distance in the image shooting information corresponding to the first image;

obtaining a second internal parameter matrix of the second shooting device according to shooting device parameters in the image shooting information corresponding to the second image, and obtaining a second external parameter matrix of the shooting device according to a shooting angle and a shooting distance in the image shooting information corresponding to the second image;

determining a second feature point in the second image corresponding to the position of the first feature point according to the following formula:

wherein,coordinates of the second characteristic point in the second image are obtained;coordinates of the first characteristic point in the first image are obtained; m2_outIs a second appearance parameter matrix; m2_inIs a second internal reference matrix; m1_outIs a first external parameter matrix; m1_inIs a first internal reference matrix.

In the present embodiment, there are various techniques for determining the internal reference matrix and the external reference matrix of the photographing apparatus. The coordinate mapping relationship from the three-dimensional world coordinate system to the two-dimensional image plane can be obtained by using the parameter matrixes. A typical camera imaging model is shown in fig. 6. Referring to FIG. 6, the coordinate O is shown_pObject coordinate system X as origin_pY_pZ_pAn object point Q is a point on a target object (such as a traffic sign) that corresponds to a coordinate point in the world coordinate system. Xyz at the projection center S refers to the camera coordinate system, which is also three-dimensional. X where the image point q is located_pixel,y_pixelThe coordinate system is a two-dimensional image coordinate system. Wherein the external parameter matrix of the camera is used for representing X_pY_pZ_pThe mapping relationship to xyz is such that,the reference matrix is used to represent xyz to x_pixely_pixelThe mapping relationship of (2).

The external parameter matrix is continuously changed in the shooting process and can be accurately calculated in real time according to the shooting angle of the camera and the position of the camera. The internal reference matrix is fixed for each camera and only needs to be calculated once, for example, a planar calibration method of Zhangyingyou can be adopted (ICCV 1999).

In this embodiment, calculating the attribute feature of the second feature point in the second image is similar to calculating the attribute feature of the first feature point in the first image, and is not described herein again. The attribute features may be any feature that reflects the attributes of the pixels of the image, such as gray scale, color, etc., for subsequent comparison.

150. And obtaining the accuracy of the target object identification result in the first image and/or the second image according to the similarity between the attribute features of the first feature points and the attribute features of the second feature points.

In this embodiment, the target object recognition apparatus may calculate, for all first feature points in the first image, similarity between the attribute feature of each first feature point and the attribute feature of the second feature point at the corresponding position in the second image one by one, then integrate all the obtained similarities, and determine, according to the integrated result, a correct rate of the target object recognition result in the first image and/or the second image obtained when the operation 120 is performed, so as to determine whether to perform further recognition of the target object detail information.

When all the obtained similarities are integrated, all the obtained similarities can be directly weighted or averaged. When the weighted value or the average value is larger, the probability that the target object exists in the second image is higher, the accuracy of the target object recognition result in the first image obtained when the operation 120 is performed is higher, and the accuracy of the target object recognition result in the second image is lower.

It should be noted that the execution sequence of each operation provided in the above technical solution is only a specific example, and the technical solution provided by the present invention is not limited thereto. The operation of "calculating the attribute feature of the first feature point in the first image" may also be performed after the operation of "determining the second feature point corresponding to the first feature point position in the second image according to the image capturing information", and the sequential execution order between the two operations is not limited in this embodiment.

According to the technical scheme provided by the embodiment, on one hand, the target object recognition result of one image is corrected by using the target object recognition result of the other image based on the shooting angle relationship between the two images, and the accuracy of the target object recognition result in the first image and/or the second image is obtained according to the corrected target object recognition result, so that the overall target object recognition rate is greatly improved; on the other hand, the mapping relation between the positions of the characteristic points of the two images is found by utilizing the image shooting information, so that the identification of the target object in the image with large geometric deformation can be realized from the imaging model of the image.

Example two

Fig. 4 is a schematic flowchart of a target object identification method according to a second embodiment of the present invention. In this embodiment, based on the above embodiments, the case where the first feature point is an edge point on the area where the object is located in the first image and the attribute feature is a gray gradient vector is further explained. Referring to fig. 4, the target object identification method specifically includes the following operations:

410. acquiring at least two images obtained by shooting a target object from at least two angles and image shooting information of the images;

420. respectively identifying the target objects of the at least two images by adopting a set identification algorithm, and determining a first image and a second image from the at least two images according to an identification result;

430. determining a first edge point on an area where a target object in the first image is located, and calculating a gray gradient vector of the first edge point in the first image;

440. determining a second edge point corresponding to the first edge point position in the second image according to the image shooting information, and calculating a gray gradient vector of the second edge point in the second image;

450. and obtaining the accuracy of the target object identification result in the first image and/or the second image according to the similarity between the gray gradient vector of the first edge point and the gray gradient vector of the second edge point.

In this embodiment, determining the first edge point on the area where the target object is located in the first image may specifically be: calculating a gray gradient vector of at least one point in the region where the target object is located in the first image, wherein the gray gradient vector is obtained by synthesizing gradient components of the gray of the point in different directions in the first image; the first edge point is a point at which the modulus of the gradation gradient vector or the modulus of the gradient component in a certain direction reaches a set threshold value.

When the first image is represented by an RGB color space, the gray scale of a point may be the gray scale of R, G, B corresponding to the point; when each point in the first image is represented by a YUV color space, the gray scale of the point may be the gray scale on the Y component corresponding to the point; when each point in the first image is represented in IHS color space, the gray scale of the point may be the gray scale on the I-component corresponding to the point. There are various ways to calculate the similarity between the gray gradient vectors, for example, the following way can be adopted, that is, obtaining the accuracy of the target object recognition result in the first image and/or the second image according to the similarity between the gray gradient vector of the first edge point and the gray gradient vector of the second edge point, which may include:

calculating a similarity between the gray gradient vectors according to the first edge point and the gray gradient vectors according to the second edge point according to the following formula:

wherein S is similarity; p1 is a first edge point in the first image; p is a first edge point set in the first image;modulo of the gray gradient vector of the first edge point p 1; angle (p1) is the direction of the gray gradient vector of the first edge point p 1; angle (p2) is the direction of the gray gradient vector of the second edge point p2 corresponding to the first edge point p 1; diff (angle (p1), angle (p2)) is a metric function describing the degree of difference between angle (p1) and angle (p 2).

The metric function diff (angle (p1), angle (p2)) may be a function that varies non-linearly with the angle difference of the direction of the gray gradient vector of the first edge point and the direction of the gray gradient vector of the second edge point, the larger the angle difference, the larger the slope of the function value. For example, the angle difference is 3 degrees, and the function value is 5; the angle difference is 6 degrees, and the function value is 12; the angular difference was 9 degrees and the function value was 20.

In this embodiment, when comparing two images with different shooting angles and different target object recognition results, only the similarity between the gray gradient vectors of the edge points in the images is calculated, but not the similarity between the gray gradient vectors of all the points in the region where the target object is located, so that the calculation speed can be increased on the premise of not affecting the accuracy of the target object recognition result as much as possible.

The typical application scenario of the above technical solution is identification of a traffic sign, and when the target object is the traffic sign, acquiring at least two images and image capturing information thereof captured by the target object from at least two angles may include:

acquiring at least two images obtained by shooting a traffic sign in a preset time period of vehicle driving by a camera configured on a vehicle and position data of a current vehicle generated by a navigator configured on the vehicle when the images are shot;

and obtaining the shooting angle and the shooting distance in the image shooting information according to the acquired position data of the vehicle, and taking the camera parameters configured on the vehicle as the shooting equipment parameters in the image shooting information.

The traffic sign usually carries out image acquisition outdoors, and both the change of a shooting angle can exist when a vehicle runs and the influence of the change of an outdoor illumination angle on the traffic sign image exists, so that the image deformation is large and the identification is often difficult. By adopting the technical scheme of the embodiment of the invention, the identification result can be verified based on the comparison between the images. Particularly, the technical scheme does not integrate and identify the parameters of the multiple images, and does not neutralize the image with good shooting effect and the image with poor shooting effect, so that the target object can be identified by using the high-quality image, the multiple images can be used for verification, and the identification accuracy is improved.

EXAMPLE III

Fig. 5 is a schematic structural diagram of a target recognition device according to a third embodiment of the present invention. Referring to fig. 5, the object recognition apparatus includes an image acquisition unit 510, an image classification unit 520, a first feature extraction unit 530, a second feature extraction unit 540, and a recognition result generation unit 550. Wherein:

an image obtaining unit 510, configured to obtain at least two images of a target object captured from at least two angles and image capturing information thereof, where the image capturing information includes a capturing angle, a capturing distance, and a capturing device parameter;

an image classification unit 520, configured to perform target object recognition on the at least two images respectively by using a set recognition algorithm, and determine a first image and a second image from the at least two images according to a recognition result;

a first feature extraction unit 530, configured to determine a first feature point on an area where a target object in the first image is located, and calculate an attribute feature of the first feature point in the first image;

a second feature extraction unit 540, configured to determine, according to the image capturing information, a second feature point in the second image corresponding to the first feature point position, and calculate an attribute feature of the second feature point in the second image;

the recognition result generating unit 550 is configured to obtain a correct rate of the target object recognition result in the first image and/or the second image according to a similarity between the attribute features of the first feature point and the attribute features of the second feature point.

Further, the image classification unit 520 is specifically configured to:

and respectively identifying the target objects of the at least two images by adopting a set identification algorithm, wherein an image capable of acquiring an identification result is taken as a first image, and an image incapable of acquiring the identification result is taken as a second image.

Further, the first feature extraction unit 530 is specifically configured to:

and calculating the attribute characteristics of at least one point on the region where the target object is located in the first image, and taking the point with the attribute characteristics meeting set conditions as a first characteristic point.

Further, the second feature extraction unit 540 is specifically configured to:

obtaining a first internal reference matrix of first shooting equipment according to shooting equipment parameters in image shooting information corresponding to the first image, and obtaining a first external reference matrix of the first shooting equipment according to a shooting angle and a shooting distance in the image shooting information corresponding to the first image;

obtaining a second internal parameter matrix of second shooting equipment according to shooting equipment parameters in the image shooting information corresponding to the second image, and obtaining a second external parameter matrix of the shooting equipment according to a shooting angle and a shooting distance in the image shooting information corresponding to the second image;

determining a second feature point in the second image corresponding to the first feature point position according to the following formula:

wherein,coordinates of the second feature point in the second image;coordinates of the first feature point in the first image are obtained; m2_outIs the second appearance parameter matrix; m2_inThe second reference matrix is obtained; m1_outIs the first external parameter matrix; m1_inIs the first internal reference matrix;

and calculating the attribute feature of the second feature point in the second image.

Further, the first feature point is an edge point on an area where the target object is located in the first image;

the first feature extraction unit 530 is specifically configured to: calculating a gray gradient vector of the first feature point in the first image; taking the calculated gray gradient vector as the attribute feature of the first feature point;

the second feature extraction unit 540 is specifically configured to: calculating a gray gradient vector of the second feature point in the second image; and taking the calculated gray gradient vector as the attribute feature of the second feature point.

Further, the attribute features include a gray gradient vector;

the recognition result generating unit 550 is specifically configured to:

calculating a similarity between the gray gradient vector of the first feature point and the gray gradient vector of the second feature point according to the following formula:

wherein S is similarity; p1 is a first feature point in the first image; p is a first feature point set in the first image;is the modulus of the gray scale gradient vector of the first feature point p 1; angle (p1) is the direction of the gray gradient vector of the first feature point p 1; angle (p2) is the direction of the gray gradient vector of the second feature point p2 corresponding to the first feature point p 1; diff (angle (p1), angle (p2)) is a metric function describing the degree of difference between angle (p1) and angle (p 2);

and determining the accuracy of the target object recognition result in the first image and/or the second image according to the similarity.

Further, the target object is a traffic sign;

the image obtaining unit 510 is specifically configured to:

acquiring at least two images obtained by shooting a traffic sign in a preset time period of vehicle driving by a camera configured on a vehicle and position data of a current vehicle generated by a navigator configured on the vehicle when the images are shot;

and obtaining the shooting angle and the shooting distance in the image shooting information according to the acquired position data of the vehicle, and taking the camera parameters configured on the vehicle as the shooting equipment parameters in the image shooting information.

The product can execute the method provided by any embodiment of the invention, and has corresponding functional modules and beneficial effects of the execution method.

It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims (12)

1. A target object recognition method, comprising:

acquiring at least two images obtained by shooting a target object from at least two angles and image shooting information thereof, wherein the image shooting information comprises a shooting angle, a shooting distance and shooting equipment parameters;

respectively identifying the target objects of the at least two images by adopting a set identification algorithm, and determining a first image and a second image from the at least two images according to an identification result;

determining a first feature point on an area where a target object is located in the first image, and calculating attribute features of the first feature point in the first image;

determining a second feature point corresponding to the position of the first feature point in the second image according to the image shooting information, and calculating the attribute feature of the second feature point in the second image;

according to the similarity between the attribute features of the first feature points and the attribute features of the second feature points, obtaining the accuracy of the target object identification result in the first image and/or the second image;

the method for recognizing the target object in the at least two images by adopting a set recognition algorithm and determining the first image and the second image from the at least two images according to the recognition result comprises the following steps:

and respectively identifying the target objects of the at least two images by adopting a set identification algorithm, wherein an image capable of acquiring an identification result is taken as a first image, and an image incapable of acquiring the identification result is taken as a second image.

2. The method for identifying the target object according to claim 1, wherein determining a first feature point on an area where the target object is located in the first image, and calculating the attribute feature of the first feature point in the first image comprises:

and calculating the attribute characteristics of at least one point on the region where the target object is located in the first image, and taking the point with the attribute characteristics meeting set conditions as a first characteristic point.

3. The method for identifying the target object according to claim 1, wherein determining a second feature point in the second image corresponding to the first feature point position according to the image capturing information, and calculating an attribute feature of the second feature point in the second image comprises:

obtaining a first internal reference matrix of first shooting equipment according to shooting equipment parameters in image shooting information corresponding to the first image, and obtaining a first external reference matrix of the first shooting equipment according to a shooting angle and a shooting distance in the image shooting information corresponding to the first image;

obtaining a second internal parameter matrix of second shooting equipment according to shooting equipment parameters in the image shooting information corresponding to the second image, and obtaining a second external parameter matrix of the shooting equipment according to a shooting angle and a shooting distance in the image shooting information corresponding to the second image;

determining a second feature point in the second image corresponding to the first feature point position according to the following formula:

$\left[\begin{array}{c}x2\\ y2\\ 1\end{array}\right]=(M{2}_{out}*M{2}_{in})*{(M{1}_{out}*M{1}_{in})}^{-1}*\left[\begin{array}{c}x1\\ y1\\ 1\end{array}\right]$

wherein,coordinates of the second feature point in the second image;coordinates of the first feature point in the first image are obtained; m2_outIs the second appearance parameter matrix; m2_inThe second reference matrix is obtained; m1_outIs the first external parameter matrix; m1_inIs the first internal reference matrix;

and calculating the attribute feature of the second feature point in the second image.

4. The object identification method according to claim 1, wherein the first feature point is an edge point on an area where an object is located in the first image;

calculating attribute features of the first feature point in the first image, including: calculating a gray gradient vector of the first feature point in the first image; taking the calculated gray gradient vector as the attribute feature of the first feature point;

calculating attribute features of the second feature points in the second image, including: calculating a gray gradient vector of the second feature point in the second image; and taking the calculated gray gradient vector as the attribute feature of the second feature point.

5. The object recognition method according to claim 4, wherein the attribute feature includes a gray gradient vector;

obtaining the accuracy of the target object recognition result in the first image and/or the second image according to the similarity between the attribute features of the first feature points and the attribute features of the second feature points, including:

calculating a similarity between the gray gradient vector of the first feature point and the gray gradient vector of the second feature point according to the following formula:

$S={\mathrm{\Σ}}_{p1\∈P}\left|\▿p1\right|*diff\left(angle\right(p1),angle(p2\left)\right)$

wherein S is similarity; p1 is a first feature point in the first image; p is a first feature point set in the first image;is the modulus of the gray scale gradient vector of the first feature point p 1; angle (p1) is the direction of the gray gradient vector of the first feature point p 1; angle (p2) is the direction of the gray gradient vector of the second feature point p2 corresponding to the first feature point p 1; diff (angle (p1), angle (p2)) is a degree describing the degree of difference between angle (p1) and angle (p2)A quantity function;

and determining the accuracy of the target object recognition result in the first image and/or the second image according to the similarity.

6. The object identification method according to any one of claims 1 to 5, wherein the object is a traffic sign;

the method for acquiring at least two images of a target object shot from at least two angles and image shooting information of the images comprises the following steps:

acquiring at least two images obtained by shooting a traffic sign in a preset time period of vehicle driving by a camera configured on a vehicle and position data of a current vehicle generated by a navigator configured on the vehicle when the images are shot;

and obtaining the shooting angle and the shooting distance in the image shooting information according to the acquired position data of the vehicle, and taking the camera parameters configured on the vehicle as the shooting equipment parameters in the image shooting information.

7. An object recognition device, comprising:

the device comprises an image acquisition unit, a processing unit and a processing unit, wherein the image acquisition unit is used for acquiring at least two images obtained by shooting a target object from at least two angles and image shooting information of the images, and the image shooting information comprises a shooting angle, a shooting distance and shooting equipment parameters;

the image classification unit is used for respectively identifying the target objects of the at least two images by adopting a set identification algorithm and determining a first image and a second image from the at least two images according to an identification result;

the first feature extraction unit is used for determining a first feature point on an area where a target object is located in the first image and calculating the attribute feature of the first feature point in the first image;

a second feature extraction unit, configured to determine, according to the image capturing information, a second feature point in the second image corresponding to the first feature point position, and calculate an attribute feature of the second feature point in the second image;

the recognition result generating unit is used for obtaining the accuracy of the recognition result of the target object in the first image and/or the second image according to the similarity between the attribute features of the first feature points and the attribute features of the second feature points;

the image classification unit is specifically configured to:

and respectively identifying the target objects of the at least two images by adopting a set identification algorithm, wherein an image capable of acquiring an identification result is taken as a first image, and an image incapable of acquiring the identification result is taken as a second image.

8. The object recognition device according to claim 7, wherein the first feature extraction unit is specifically configured to extract the first feature

And calculating the attribute characteristics of at least one point on the region where the target object is located in the first image, and taking the point with the attribute characteristics meeting set conditions as a first characteristic point.

9. The object recognition device according to claim 7, wherein the second feature extraction unit is specifically configured to:

obtaining a first internal reference matrix of first shooting equipment according to shooting equipment parameters in image shooting information corresponding to the first image, and obtaining a first external reference matrix of the first shooting equipment according to a shooting angle and a shooting distance in the image shooting information corresponding to the first image;

obtaining a second internal parameter matrix of second shooting equipment according to shooting equipment parameters in the image shooting information corresponding to the second image, and obtaining a second external parameter matrix of the shooting equipment according to a shooting angle and a shooting distance in the image shooting information corresponding to the second image;

determining a second feature point in the second image corresponding to the first feature point position according to the following formula:

$\left[\begin{array}{c}x2\\ y2\\ 1\end{array}\right]=(M{2}_{out}*M{2}_{in})*{(M{1}_{out}*M{1}_{in})}^{-1}*\left[\begin{array}{c}x1\\ y1\\ 1\end{array}\right]$

wherein,is the second characteristic pointCoordinates in the second image;coordinates of the first feature point in the first image are obtained; m2_outIs the second appearance parameter matrix; m2_inThe second reference matrix is obtained; m1_outIs the first external parameter matrix; m1_inIs the first internal reference matrix;

and calculating the attribute feature of the second feature point in the second image.

10. The object recognition device according to claim 7, wherein the first feature point is an edge point on an area where an object is located in the first image;

the first feature extraction unit is specifically configured to: calculating a gray gradient vector of the first feature point in the first image; taking the calculated gray gradient vector as the attribute feature of the first feature point;

the second feature extraction unit is specifically configured to: calculating a gray gradient vector of the second feature point in the second image; and taking the calculated gray gradient vector as the attribute feature of the second feature point.

11. The object identifying apparatus of claim 10, wherein the attribute feature includes a gray gradient vector;

the identification result generation unit is specifically configured to:

calculating a similarity between the gray gradient vector of the first feature point and the gray gradient vector of the second feature point according to the following formula:

$S={\mathrm{\Σ}}_{p1\∈P}\left|\▿p1\right|*diff\left(angle\right(p1),angle(p2\left)\right)$

wherein S is similarity; p1 is a first feature point in the first image; p is a first feature point set in the first image;is the modulus of the gray scale gradient vector of the first feature point p 1; angle (p1) is the direction of the gray gradient vector of the first feature point p 1; angle (p2) is the direction of the gray gradient vector of the second feature point p2 corresponding to the first feature point p 1; diff (angle (p1), angle (p2)) is a metric function describing the degree of difference between angle (p1) and angle (p 2);

and determining the accuracy of the target object recognition result in the first image and/or the second image according to the similarity.

12. The object identifying device according to any one of claims 7 to 11, wherein the object is a traffic sign;

the image acquisition unit is specifically configured to:

acquiring at least two images obtained by shooting a traffic sign in a preset time period of vehicle driving by a camera configured on a vehicle and position data of a current vehicle generated by a navigator configured on the vehicle when the images are shot;

and obtaining the shooting angle and the shooting distance in the image shooting information according to the acquired position data of the vehicle, and taking the camera parameters configured on the vehicle as the shooting equipment parameters in the image shooting information.