Method, device, equipment, medium and product for selecting region of interest based on vehicle navigation

Abstract

The present disclosure provides a region of interest selection method and device based on vehicle navigation, equipment, medium and product, relating to the technical field of artificial intelligence, especially in the field of autonomous driving. The specific implementation scheme is: determining the standard recognition result corresponding to the road video collected by the vehicle-mounted camera; using a region selection strategy to select a candidate region of interest for the road video to obtain at least one candidate region of interest; performing lane line recognition on the candidate region of interest to obtain a to-be-compared recognition result, and determining the selection result of the corresponding candidate region of interest according to the to-be-compared recognition result and the standard recognition result; determining the target recognition result that meets the result selection condition according to the selection result corresponding to each of the at least one candidate region of interest, and determining the candidate region of interest corresponding to the target recognition result as the target region of interest. The embodiment of the present disclosure realizes efficient and accurate selection of the target region of interest.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the field of automatic driving in the field of artificial intelligence, and in particular to a method and device for selecting a region of interest based on vehicle navigation, equipment, medium and product. BACKGROUND

[0002] In the vehicle navigation scene, especially in the augmented reality (AR, full name: Augmented Reality) navigation scene, a camera needs to be used to collect a video in front of a road, and the collected video needs to be fused with an electronic map. In order to improve the accuracy of navigation, the image frames in the video can be cropped, that is, a region of interest is determined to crop the image frames using the region of interest, so that the cropped region accurately contains lane lines, so as to identify the lane lines, ensure that the output map video can contain accurate lane lines, complete real scene navigation, and improve the accuracy of navigation. Therefore, the selection accuracy of the region of interest is important for navigation. In actual application, different vehicle models can use different cameras, and different cameras need to be set with a region of interest to accurately select the region of interest. However, the selection of the region of interest is usually manually set by a person, and when a single selection is inaccurate, it needs to be reselected, and the selection precision and accuracy are not high. SUMMARY

[0003] The present disclosure provides a method and device for selecting a region of interest based on vehicle navigation.

[0004] According to a first aspect of the present disclosure, a method for selecting a region of interest based on vehicle navigation is provided, comprising:

[0005] determining a standard recognition result corresponding to a road video collected by a vehicle-mounted camera;

[0006] using a region selection strategy to select a candidate region of interest for the road video to obtain at least one candidate region of interest;

[0007] performing lane line recognition on the candidate region of interest to obtain a to-be-compared recognition result, and determining a selection result of the corresponding candidate region of interest according to the to-be-compared recognition result and the standard recognition result;

[0008] determining a target recognition result that meets a result selection condition according to the selection result corresponding to each of the at least one candidate region of interest, and determining a candidate region of interest corresponding to the target recognition result as a target region of interest.

[0009] According to a second aspect of the present disclosure, a device for selecting a region of interest based on vehicle navigation is provided, comprising:

[0010] The standard acquisition unit is used to determine the standard recognition result corresponding to the road video captured by the vehicle-mounted camera.

[0011] A region selection unit is used to select candidate regions of interest for the road video using a region selection strategy, so as to obtain at least one candidate region of interest.

[0012] The region analysis unit is used to identify lane lines in the candidate regions of interest to obtain comparison identification results, and to determine the selection result of the corresponding candidate regions of interest based on the comparison identification results and the standard identification results.

[0013] The target determination unit is used to determine the target recognition result that satisfies the result selection condition based on the selection results corresponding to at least one candidate region of interest, and to determine the candidate region of interest corresponding to the target recognition result as the target region of interest.

[0014] According to a third aspect of this disclosure, an electronic device is provided, comprising:

[0015] At least one processor; and

[0016] A memory communicatively connected to the at least one processor; wherein,

[0017] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in the first aspect.

[0018] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are configured to cause the computer to perform the method described in the first aspect.

[0019] According to a fifth aspect of this disclosure, a computer program product is provided, the computer program product comprising: a computer program stored in a readable storage medium, wherein at least one processor of an electronic device can read the computer program from the readable storage medium, and the at least one processor executes the computer program to cause the electronic device to perform the method described in the first aspect.

[0020] The technology disclosed herein solves the problem of low efficiency in selecting regions of interest. It automatically selects candidate regions of interest by using a region selection strategy and compares the candidate regions of interest with a standard. The comparison results can be used to select candidate regions of interest automatically, thereby achieving automatic selection of the target region of interest by using the selection results. The selection process adopts result selection constraints, which can improve selection efficiency.

[0021] It is to be understood that the details set forth herein do not limit the scope of the embodiments of the present disclosure but merely constitute illustrative of typical embodiments. Other embodiments can readily be devised in light of the disclosure herein without departing from the scope and spirit of the present disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0022] The accompanying drawings are included to provide a further understanding of the present application, and are incorporated in and constitute a part of this specification. In the drawings:

[0023] Figure 1 is a system architecture diagram for implementing the method for selecting a region of interest based on vehicle navigation according to a first embodiment of the present disclosure;

[0024] Figure 2 is a flowchart of the method for selecting a region of interest based on vehicle navigation according to a second embodiment of the present disclosure;

[0025] Figure 3 is a flowchart of the method for selecting a region of interest based on vehicle navigation according to a third embodiment of the present disclosure;

[0026] Figure 4 is a flowchart of the method for selecting a region of interest based on vehicle navigation according to a fourth embodiment of the present disclosure;

[0027] Figure 5 is a flowchart of the method for selecting a region of interest based on vehicle navigation according to a fifth embodiment of the present disclosure;

[0028] Figure 6 is a flowchart of the method for selecting a region of interest based on vehicle navigation according to a sixth embodiment of the present disclosure;

[0029] Figure 7 is a structural schematic diagram of the device for selecting a region of interest based on vehicle navigation according to a seventh embodiment of the present disclosure;

[0030] Figure 8 is a block diagram of an electronic device for implementing the method for selecting a region of interest based on vehicle navigation according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0031] Exemplary embodiments of the present disclosure are described herein with reference to the accompanying drawings, which are included to provide a further understanding of the present disclosure and are incorporated in and constitute a part of this specification. Embodiments of the present disclosure described herein should be considered in a descriptive sense only and not for purposes of limitation. Therefore, various modifications can be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Likewise, the present disclosure is not intended to be limited to the various aspects of the embodiments described herein, but rather, the scope of the present disclosure includes all alternatives, modifications, and equivalents falling within the scope of the claims.

[0032] The technical solutions of the present disclosure can be applied to a vehicle navigation scene based on augmented reality (AR, full name: Augmented Reality) technology, and the region of interest is selected by using a region selection strategy for identification detection, and the target region of interest is automatically selected by using the selection result. The selection efficiency can be improved by using result selection constraints during selection.

[0033] In related technologies, in a vehicle navigation scene based on augmented reality (AR, full name: Augmented Reality) technology, in order to accurately navigate the road, the video of the region of interest where the road is located can be intercepted, and then lane line recognition is performed on the multiple region images obtained by cropping, to obtain the recognition result of each region image, and then the lane line corresponding to the recognition result is displayed in combination with an electronic map. Generally, the region of interest can be manually set and selected by the vehicle camera installer. This selection method of the region of interest has low accuracy and low efficiency.

[0034] To solve the above technical problems, in the embodiments of the present disclosure, at least one candidate region of interest is selected according to a preset region selection strategy, and then each candidate region of interest is identified and tested to obtain a corresponding selection result, so that the selection result of the region of interest is used for automatic selection of the region.

[0035] In the embodiments of the present disclosure, after determining the road video collected by the vehicle camera, the standard recognition result corresponding to the road video can be determined. The region selection strategy is used to select the candidate region of interest for the road video to obtain at least one candidate region of interest. For each selected candidate region of interest, lane line recognition can be performed on the candidate region of interest to obtain a to-be-compared recognition result. The to-be-compared recognition result is compared with the standard recognition result to obtain a selection result corresponding to the candidate region of interest. The standard recognition result is used to determine the selection of the candidate region of interest, so that the target recognition result that meets the result selection condition is determined according to the selection result corresponding to each of the at least one candidate region of interest, and the candidate region of interest corresponding to the target recognition result is determined as the target region of interest. The candidate region of interest is accurately selected by using the selection result. By comparing the candidate region of interest, the candidate region of interest can be automatically selected by using the comparison result selection condition, so that the target region of interest is automatically selected by using the selection result, and the selection efficiency can be improved by using the result selection constraint during selection.

[0036] The present disclosure provides a region of interest selection method, device, equipment, medium and product based on vehicle navigation, which can be applied to the field of automatic driving in the field of artificial intelligence to achieve fast and accurate selection of the region of interest.

[0037] The technical solutions of the present disclosure will be described in detail below with reference to the drawings.

[0038] For ease of understanding, Figure 1 For the system architecture diagram provided by the first embodiment of the present disclosure for implementing the method for selecting the region of interest based on the vehicle navigation, the network architecture can include an electronic device 1 and a plurality of vehicles 2 connected to the electronic device 1 through a local area network or a wide area network. The vehicle 2 can be configured with a camera 21 to implement vehicle navigation based on augmented reality technology. The types and poses of the cameras in different vehicles 2 can be different, so as to solve the problem of automatic selection of the region of interest of each camera, the road video collected by each camera can be transmitted to the electronic device 1. The electronic device 1 can be, for example, a general server, a cloud server, etc., and can also be a computer, a notebook, a supercomputer device, etc. The specific type of the electronic device 1 is not limited in the present disclosure. Taking the cloud server shown in the figure as an example, when the cloud server 1 obtains the road video collected by the vehicle camera, the automatic selection of the target region of interest can be realized based on the technical solutions of the present disclosure. In addition, the cloud server 1 can also send the selected target region of interest to the vehicle 2, so that the vehicle 2 identifies the lane line according to the target region of interest, and fuses the identified lane line with the electronic map for user display. Then, the cloud server 1 can also update the electronic map according to the obtained target region of interest, and send the electronic map to the vehicle 2 for display.

[0039] As Figure 2 As shown in the flowchart of the method for selecting the region of interest based on the vehicle navigation provided by the second embodiment of the present disclosure, the method can be executed by a device for selecting the region of interest based on the vehicle navigation, which can be located in an electronic device. Therefore, the method for selecting the region of interest based on the vehicle navigation can include the following steps:

[0040] 201: Determine the standard recognition result corresponding to the road video collected by the vehicle camera.

[0041] The camera can be located on a vehicle driving on the road to collect the road video in front of the vehicle. The road video can be collected in real time or pre-collected. The vehicle in the present disclosure can refer to a vehicle with vehicle navigation function. For example, it can be an autonomous vehicle. The vehicle navigation involved in the present disclosure can be, for example, vehicle navigation based on augmented reality (full name: Augmented Reality, abbreviated as AR) technology.

[0042] In the field of vehicle navigation, the cameras of different models of vehicles are different. For example, the camera model, the shooting angle, the camera coefficient, and other camera parameters related to video acquisition are different. In order to achieve accurate navigation and display accurate navigation roads, different cameras can be set with corresponding regions of interest (full name: Region of Interest, abbreviated: RIO) in order to perform lane recognition on the image area within the region of interest of different cameras, display accurate lane lines and regional maps. Therefore, accurate selection of the region of interest can achieve accurate navigation of the electronic map.

[0043] The standard recognition result can be a relatively accurate recognition result corresponding to the road video. When using the standard recognition result for map navigation, a relatively accurate navigation effect can be obtained.

[0044] 202: A region selection strategy is used to select a candidate region of interest for the road video to obtain at least one candidate region of interest.

[0045] The region selection strategy can be pre-set. The candidate region of interest can be selected using the region selection strategy. The lane line recognition can be performed on each candidate region of interest to obtain the corresponding to-be-compared recognition result.

[0046] The at least one candidate region of interest is composed of all candidate regions of interest selected using the region selection strategy.

[0047] 203: The lane line recognition is performed on the candidate region of interest to obtain the to-be-compared recognition result, and the selection result of the corresponding candidate region of interest is determined according to the to-be-compared recognition result and the standard recognition result.

[0048] When any candidate region of interest is obtained, the lane line recognition can be performed on the candidate region of interest to obtain the corresponding to-be-compared result. The selection result of the candidate region of interest is determined according to the to-be-compared result and the standard recognition result.

[0049] The selection result can quantify the region selection effect of the candidate region of interest on the road video, and can achieve accurate selection of the final target region of interest.

[0050] 204: According to the selection result corresponding to each of the at least one candidate region of interest, a target recognition result satisfying a result selection condition is determined, and a candidate region of interest corresponding to the target recognition result is determined as a target region of interest.

[0051] The result selection condition can be pre-set. The target recognition result can be a selection result satisfying the result selection condition, which can be selected from the selection results corresponding to each of the at least one candidate region of interest.

[0052] In the embodiments of the present disclosure, after determining the road video collected by the vehicle-mounted camera, a region selection strategy can be used to select a candidate region of interest for the road video to obtain at least one candidate region of interest and a selection result corresponding to each candidate region of interest, the selection result can be lane line recognition based on the candidate region of interest for road recognition, so as to determine a target recognition result meeting a result selection condition according to the selection result corresponding to each candidate region of interest, and determine that the candidate region of interest corresponding to the target recognition result is a target region of interest. Through automatic selection of the region of interest and result detection, the result selection condition can be used to automatically select the candidate region of interest, so as to realize automatic selection of the target region of interest by using the selection result, and the selection result selection constraint can be used in the selection, thereby improving the selection efficiency.

[0053] As shown in Figure 3 FIG. 3 is a flowchart of a method for selecting a region of interest based on a vehicle-mounted navigation according to a third embodiment of the present disclosure. The method can include the following steps:

[0054] 301: Correcting a road video collected by a vehicle-mounted camera to obtain a corrected video.

[0055] Some steps in the embodiments of the present disclosure can refer to the description of other embodiments. For the sake of brevity, they will not be described here.

[0056] The corrected video can be obtained by correcting the road video collected by the vehicle-mounted camera.

[0057] The algorithm of the corrected video is relatively complex. In actual application, if the road video is directly corrected, because there are many vehicles, the video collected by each vehicle is corrected, which can cause a very high calculation amount of the pre-step of lane line recognition, that is, the video correction step. Therefore, in actual application, the video collected by each vehicle is not corrected to identify the lane line. However, in the embodiments of the present disclosure, the region of interest is set for different types of cameras, the video is corrected, the standard recognition result is obtained by video correction, so as to identify and compare the candidate region of interest, obtain the target region of interest corresponding to the camera, and obtain the accurate target region of interest. Taking the camera as the recognition basis, the video correction and selection of the target region of interest are performed for each type of camera, so that the selected target region of interest is used for each type of camera, the excessive mathematical calculation is avoided, and the calculation efficiency is improved.

[0058] 302: Lane line recognition is performed on the corrected video to obtain a standard recognition result.

[0059] Optionally, a lane line recognition algorithm can be used to perform lane line recognition on the corrected video to obtain the standard recognition result.

[0060] The lane line recognition algorithm can be pre-set and can be a lane line detection algorithm known in the related art. For example, it can be a semantic segmentation or instance segmentation algorithm based on visual features, such as SCNN (Spatial Convolutional Neural Networks for Traffic Lane Detection). The specific type of lane line recognition algorithm in the embodiments of the present disclosure is not limited too much.

[0061] 303: A region selection strategy is used to select a candidate region of interest for the road video to obtain at least one candidate region of interest.

[0062] 304: Lane line recognition is performed on the candidate region of interest to obtain a to-be-compared recognition result, and a selection result of the corresponding candidate region of interest is determined according to the to-be-compared recognition result and the standard recognition result.

[0063] 305: According to the selection result corresponding to each of the at least one candidate region of interest, a target recognition result that satisfies a result selection condition is determined, and a candidate region of interest corresponding to the target recognition result is determined as a target region of interest.

[0064] In the embodiments of the present disclosure, when obtaining the standard recognition result, the road video can be corrected to obtain a corrected video, and the lane line recognition is performed on the corrected video to obtain the standard recognition result. By correcting the road video, accurate recognition of the video can be performed to obtain the corresponding recognition result.

[0065] As a possible implementation, the lane line recognition is performed on the corrected video to obtain the standard recognition result, including:

[0066] Lane line recognition is performed on each of the at least one corrected image to obtain a first lane line parameter corresponding to each of the at least one corrected image.

[0067] It is determined that the first lane line parameter corresponding to each of the at least one corrected image constitutes the standard recognition result.

[0068] In actual application, the corrected video can include at least one corrected image, that is, at least one image can be combined into the corrected video according to a timestamp. Lane line identification can be performed on each corrected image to obtain the first lane line parameter corresponding to each of the at least one corrected image. Of course, in actual application, in order to reduce the identification calculation amount, image sampling can be performed on all corrected images in the corrected video, and lane line identification can be performed on the at least one sampled corrected image to obtain the first lane line parameter corresponding to each of the at least one corrected image. The sampling frequency of the corrected image can be set according to the actual accuracy requirement. The higher the sampling frequency, the higher the identification accuracy, and the lower the sampling frequency, the lower the identification accuracy.

[0069] In the embodiments of the present disclosure, for the at least one corrected image in the corrected video, lane line identification can be performed on each corrected image to obtain the first lane line parameter corresponding to each of the at least one corrected image, so as to accurately obtain the lane line parameter of each corrected image in the corrected video, and the first lane line parameter corresponding to each of the at least one corrected image constitutes a standard identification result, so as to accurately calculate the standard identification result and improve the calculation accuracy.

[0070] In a possible design, the corrected video can be obtained by performing correction processing on the road video, which can include:

[0071] The camera intrinsic parameter and the distortion coefficient of the vehicle-mounted camera are obtained.

[0072] The corrected video is obtained by performing correction processing on the road video by using the camera intrinsic parameter and the distortion coefficient.

[0073] Optionally, the camera intrinsic parameter and the distortion coefficient can be preset. The corrected video can be obtained by performing correction processing on the road video by using the camera intrinsic parameter and the distortion coefficient. The existing video correction algorithm can be used to perform correction processing on the road video by using the camera intrinsic parameter and the distortion coefficient, so as to obtain the corrected video.

[0074] In the embodiments of the present disclosure, when the road video is corrected, the camera intrinsic parameter and the distortion coefficient of the vehicle-mounted camera can be obtained, and the road video can be corrected. For the vehicle-mounted camera, the camera intrinsic parameter and the distortion coefficient corresponding to the camera are used to accurately correct the camera, so that the accuracy of the standard identification result corresponds to the camera parameter, and the identification accuracy is higher.

[0075] In some embodiments, the camera intrinsic parameter and the distortion correction coefficient of the vehicle-mounted camera can be obtained, which can include:

[0076] The camera intrinsic parameter and the distortion correction coefficient of the vehicle-mounted camera are obtained by performing parameter calibration processing on the vehicle-mounted camera based on a preset checkerboard video.

[0077] Optionally, the checkerboard video can be used to calibrate the camera intrinsic parameters and distortion correction coefficients of the vehicle-mounted camera, to obtain the corresponding camera intrinsic parameters and distortion correction coefficients. The distortion correction coefficients can include one or more of the following coefficients: distortion coefficients, lens field of vision (FOV) coefficients, and the like.

[0078] In the embodiments of the present disclosure, the checkerboard video can be used to calibrate the camera intrinsic parameters and distortion correction coefficients of the vehicle-mounted camera, to obtain the camera intrinsic parameters and distortion correction coefficients of the vehicle-mounted camera, to accurately obtain the video correction parameters, and to improve the accuracy of video correction.

[0079] As an optional implementation, the lane line recognition on the candidate region of interest can include:

[0080] Based on the candidate region of interest, at least one region image corresponding to the road video is obtained by image cropping on at least one image frame in the road video.

[0081] The lane line recognition is performed on the at least one region image, to obtain the second lane line parameters corresponding to the at least one region image, respectively.

[0082] The second lane line parameters corresponding to the at least one region image, respectively, constitute the to-be-compared recognition result.

[0083] In actual applications, the region video can include at least one region image, i.e., at least one image can be combined into a region video according to their respective time stamps. The lane line recognition can be performed on each region image, to obtain the second lane line parameters corresponding to the at least one region image, respectively. Of course, in actual applications, to reduce the recognition calculation amount, image sampling can be performed on all region images in the region video, and the lane line recognition can be performed on the at least one region image after sampling, to obtain the second lane line parameters corresponding to the at least one region image, respectively. The sampling frequency of the region image can be set according to the actual accuracy requirement. The higher the sampling frequency, the higher the recognition accuracy, and the lower the sampling frequency, the lower the recognition accuracy. When sampling, the sampling starting position and the sampling frequency of the region video and the correction video are the same, i.e., the image frames in the road video corresponding to the sampling results are the same.

[0084] In the embodiments of the present disclosure, each image frame in the road video can be cropped based on the candidate region of interest, at least one region image can be obtained, the region image is a local image corresponding to the region of interest, and the cropping of the region of interest is realized. Lane line identification is performed on the at least one region image respectively, and lane line parameters corresponding to the at least one region image are obtained, and accurate lane line parameter identification is realized. By video cropping of the image frames in the road video according to the candidate region of interest, lane line identification can be performed on the single frame of the local image of interest, and accurate identification results to be compared are obtained.

[0085] In some embodiments, according to the identification results to be compared and the standard identification results, the selection result of the corresponding candidate region of interest can include:

[0086] According to the lane line identification results to be compared and the standard lane line identification results, the identification accuracy of the candidate region of interest is determined.

[0087] According to the identification results to be compared, the first detection rate corresponding to the candidate region of interest is calculated.

[0088] The identification accuracy and the first detection rate constitute the selection result of the candidate region of interest.

[0089] The identification accuracy can be a numerical value representing the identification accuracy of the identification results to be compared corresponding to the candidate region of interest.

[0090] In the embodiments of the present disclosure, according to the lane line identification results to be compared and the standard lane line identification results, the identification accuracy of the candidate region of interest can be determined, and the identification accuracy can evaluate the identification effect of the candidate region of interest. Then, according to the identification results to be compared, the first detection rate corresponding to the candidate region of interest is calculated, which can confirm the detection effectiveness of the region video corresponding to the candidate region of interest. Thus, the selection result of the candidate region of interest constituted by the identification accuracy and the first detection rate is determined, the accurate calculation of the selection result of the candidate region of interest is realized, so as to be applied to the selection of the target identification result in the subsequent, and the selection accuracy of the target region of interest is improved.

[0091] In order to accurately calculate the detection rate, in one possible design, according to the identification results to be compared, the first detection rate corresponding to the candidate region of interest is calculated, including:

[0092] According to the second lane line parameters corresponding to the at least one region image, the first missed detection image in the at least one region image is determined.

[0093] The difference between the first image quantity of the at least one region image and the second image data of the first missed detection image is calculated, and a third image quantity is obtained.

[0094] The quotient of the third image quantity and the first image quantity is calculated to obtain the first detection rate corresponding to the candidate region of interest.

[0095] The region image can correspond to second lane line parameters, which can include lane line position, length, shape, angle with the horizontal line, and other data in the image. The lane line shape can be a straight line or a curve, for example.

[0096] The first detection rate can be the ratio of the number of region images in which normal lane line detection calculation is performed to the total number of region images.

[0097] In the embodiments of the present disclosure, the first missed detection image in the at least one region image is determined by using the second lane line parameters corresponding to the at least one region image, and the difference between the first image quantity of the at least one region image and the second image data of the first missed detection image is calculated to obtain the third image quantity. The first detection rate of the candidate region of interest can be obtained by calculating the quotient of the third image quantity and the first image quantity. The first detection rate can be obtained by determining the number of first missed detection images and the number of normally detected images to calculate the proportion of normally detected images to all region images. By accurately calculating the first detection rate, the to-be-compared recognition result of the candidate region of interest can be quantified numerically to facilitate quick target selection.

[0098] As an optional implementation, the standard recognition result includes: first lane line parameters corresponding to the at least one corrected image; and the at least one corrected image is an image obtained by performing video correction on a road image.

[0099] According to the to-be-compared lane line recognition result and the standard lane line recognition result, the recognition accuracy of the candidate region of interest is determined, including:

[0100] For at least one image frame in the road video, region images and corrected images corresponding to the at least one image frame are determined;

[0101] The second lane line parameters corresponding to the region image of any image frame and the first lane line parameters corresponding to the corrected image of the image frame are calculated to obtain lane line difference values corresponding to the at least one image frame.

[0102] According to the lane line difference values corresponding to the at least one image frame, lane line variance is calculated.

[0103] The lane line variance is normalized to obtain the recognition accuracy of the candidate region of interest.

[0104] Optionally, when calculating the recognition accuracy, the lane line parameter difference between the region image and the corrected image of any image frame can be calculated to obtain the corresponding lane line difference value.

[0105] Optionally, the lane line parameter can include at least one sub-parameter, at least one first sub-parameter in the first lane line parameter of each image frame and at least one second sub-parameter in the second lane line parameter can be respectively calculated by difference of each sub-parameter, for example, a position difference corresponding to a position sub-parameter can be calculated, a shape difference corresponding to a shape can be calculated, etc. By calculating the difference of at least one sub-parameter respectively, an array or one-dimensional vector composed of at least one sub-parameter difference obtained by calculation can be combined to obtain a lane line difference. When calculating the lane line variance according to the lane line difference corresponding to at least one image frame respectively, the square of the lane line difference corresponding to any image frame can be calculated to obtain the square of the lane line difference corresponding to at least one image frame respectively; then the sum of the squares of all lane line differences is added and square rooted to obtain the lane line variance.

[0106] Optionally, when calculating the square of the lane line difference, adding the square and square rooting to obtain the lane line variance, the variance calculation can be performed on the difference of any sub-parameter, that is, the square calculation of the sub-parameter difference corresponding to at least one image frame respectively is performed, the sum of the squares obtained by at least one image frame respectively corresponding to the sub-parameter is added, and the square of the sum is obtained to obtain the sub-variance corresponding to the sub-parameter. The array or one-dimensional vector obtained by combining the sub-variances corresponding to at least one sub-parameter respectively can be the lane line variance.

[0107] The obtained lane line variance can include at least one sub-variance corresponding to each sub-parameter, and different sub-parameters represent the lane line from different dimensions. In order to obtain the recognition accuracy of the lane line, at least one sub-variance of at least one sub-parameter can be normalized according to the weight corresponding to each sub-parameter in the lane line variance to obtain the recognition accuracy of the candidate region of interest. By converting the sub-variances of at least one sub-parameter in different dimensions into the same dimension, the corresponding recognition accuracy can be obtained.

[0108] In the embodiments of the present disclosure, for at least one image frame in the road video, the region image and the corrected image corresponding to at least one image frame respectively can be determined to obtain the first lane line parameter and the second lane line parameter corresponding to each image frame, and then the lane line variance can be calculated by using the first lane line parameter and the second lane line parameter corresponding to the image frame, that is, the lane line recognition result to be compared and the standard lane line recognition result are compared to realize the comparison of the candidate region of interest corresponding to the recognition result to be compared. The recognition accuracy of the candidate region of interest can be accurately calculated, and the selection result of the candidate region of interest can be accurately determined.

[0109] In some embodiments, the first missed image in the at least one region image is determined according to the second lane line parameter corresponding to the at least one region image, comprising:

[0110] selecting a lane line parameter with an empty parameter from the second lane line parameter corresponding to the at least one region image;

[0111] determining the region image corresponding to the empty lane line parameter as the first missed image.

[0112] Optionally, the second lane line parameter being empty can mean that the parameter value corresponding to the at least one sub-parameter is zero, and the lane line in the region image is not detected.

[0113] In the embodiments of the present disclosure, when determining the first missed image, a lane line parameter with an empty parameter can be selected from the second lane line parameter corresponding to the at least one region image, that is, the region image for which the second lane line parameter is not obtained is determined as the first missed image, so as to quantitatively select the first missed image and obtain an accurate recognition result.

[0114] In a possible design, the target recognition result satisfying the result selection condition is determined according to the selection result corresponding to the at least one candidate region of interest, comprising:

[0115] The target recognition result satisfying the first selection condition and the second selection condition is determined according to the recognition accuracy and the first detection rate corresponding to the at least one candidate region of interest.

[0116] Optionally, the target recognition result can be the selection result satisfying the first selection condition and the second selection condition. The first selection condition can include that the recognition accuracy is higher than a preset accuracy threshold. The second selection condition can include that the detection rate is higher than a detection rate threshold. The accuracy threshold and the detection rate threshold can be preset.

[0117] Optionally, the target recognition result can include one, and the target recognition result can be directly output. In some embodiments, the target recognition result can include multiple, and the recognition result with the highest recognition accuracy can be selected from the target recognition result as the final target recognition result for output. The recognition result with the highest first detection rate can also be selected from the target recognition result as the final target recognition result for output. Of course, the multiple target recognition results can also be respectively identified and scored according to the recognition accuracy and the first detection rate, and the recognition result with the highest score can be obtained as the final target recognition result for output.

[0118] In the embodiments of the present disclosure, when the target recognition result is selected, the target recognition result that meets the first selection condition in terms of recognition accuracy and the second selection condition in terms of the first detection rate can be selected from the at least one candidate region of interest. By using the constraints of the first selection condition corresponding to the recognition accuracy and the second selection condition corresponding to the first detection rate, the target recognition result can be accurately selected, and the obtained target recognition result meets the first selection condition and the second selection condition at the same time, so that the target recognition result is accurately selected, and the selection reliability is higher.

[0119] As shown in Figure 4 FIG. 4 is a flowchart of a method for selecting a region of interest based on a vehicle-mounted navigation according to a fourth embodiment of the present disclosure. The method can include the following steps.

[0120] 401: Correcting a road video collected by a vehicle-mounted camera to obtain a corrected video.

[0121] In the embodiments of the present disclosure, some steps are the same as some steps in the foregoing embodiments, and for the sake of brevity of description, they will not be described here.

[0122] 402: Lane line recognition is performed on the corrected video to obtain a standard recognition result. The standard recognition result includes: first lane line parameters corresponding to at least one corrected image; and the at least one corrected image is an image obtained by performing video correction on a road image.

[0123] 403: A region selection strategy is used to select a candidate region of interest for the road video to obtain at least one candidate region of interest.

[0124] 404: Lane line recognition is performed on the candidate region of interest to obtain a to-be-compared recognition result. The to-be-compared recognition result includes: second lane line parameters corresponding to at least one region image; and the at least one region image is obtained by image cropping on at least one image frame in the road image.

[0125] 405: The recognition accuracy of the candidate region of interest is determined according to the to-be-compared lane line recognition result and the standard lane line recognition result.

[0126] 406: The first detection rate corresponding to the candidate region of interest is calculated according to the to-be-compared recognition result.

[0127] 407: The recognition accuracy and the first detection rate constitute a selection result of the candidate region of interest.

[0128] 408: A target recognition result that meets the first selection condition in terms of recognition accuracy and the second selection condition in terms of the first detection rate is determined according to the recognition accuracy and the first detection rate corresponding to the at least one candidate region of interest.

[0129] 409: The candidate region of interest corresponding to the target recognition result is determined as the target region of interest.

[0130] In this embodiment, after correcting the road video captured by the vehicle-mounted camera, lane line recognition can be performed on the corrected video to obtain a standard recognition result. The standard recognition result may include first lane line parameters corresponding to at least one corrected image. After selecting candidate sensing regions for the road video using a region selection strategy, lane line recognition can be performed on each selected candidate sensing region to obtain a comparison recognition result. This comparison recognition result may include second lane line parameters corresponding to at least one region image. The recognition accuracy and first detection rate of the region of interest are accurately calculated using the first lane line parameters of at least one corrected image and the second lane line parameters of at least one region image to determine the selection result composed of the recognition accuracy and the first detection rate. The candidate regions of interest can be quantified using the recognition accuracy and the first detection rate, allowing for accurate selection of the target recognition result based on the recognition accuracy and the first detection rate corresponding to at least one region of interest. This results in higher accuracy and a more accurate target region of interest. This process does not require user intervention, effectively improving the efficiency of selecting the target region of interest.

[0131] like Figure 5 The diagram shows a flowchart of a method for selecting a region of interest based on vehicle navigation, according to a second embodiment of this disclosure. This method may include the following steps:

[0132] 501: Determine the standard recognition result corresponding to the road video captured by the vehicle-mounted camera.

[0133] Some steps in this embodiment are the same as some steps in the foregoing embodiments, and for the sake of brevity, they will not be repeated here.

[0134] 502: Utilize a region selection strategy to select candidate regions of interest for road videos.

[0135] 503: Perform lane line recognition on candidate regions of interest to obtain recognition results to be compared, and determine the selection result of the corresponding candidate regions of interest based on the recognition results to be compared and the standard recognition results.

[0136] 504: Determine whether the comparison and identification results of the candidate region of interest have not triggered the stop selection condition. If not, return to step 502 to continue execution. If yes, proceed to step 505.

[0137] When the candidate region of interest triggers the stop selection condition, the selection of the candidate region of interest can be stopped, and when the candidate region of interest does not trigger the stop selection condition, the selection of the candidate region of interest can be continued.

[0138] The stop selection condition can include that the to-be-compared recognition result does not reach a preset minimum recognition result.

[0139] 505: Stop selecting the candidate region of interest for the road video by using the region selection strategy, and obtain at least one candidate region of interest constituted by all candidate regions of interest selected for the road video.

[0140] 506: According to the selection result corresponding to each of the at least one candidate region of interest, determine a target recognition result that meets a result selection condition, and determine the candidate region of interest corresponding to the target recognition result as a target region of interest.

[0141] In the embodiments of the present disclosure, when determining the standard recognition result corresponding to the road video collected by the vehicle-mounted camera, the region selection strategy can be used to select the candidate region of interest for the road video. The lane line recognition is performed on the candidate region of interest to obtain the to-be-compared recognition result, and the selection result of the corresponding candidate region of interest is determined according to the to-be-compared recognition result and the standard recognition result. If it is determined whether the to-be-compared recognition result of the candidate region of interest triggers the stop selection condition, if it triggers, the selection is stopped, and the candidate region of interest at which the selection is stopped is obtained. If it does not trigger, the selection of the candidate region of interest can be continued. By using the to-be-compared recognition result of the candidate region of interest, the selectability of the selection strategy corresponding to the candidate region of interest can be judged, so as to set the stop selection condition for the region selection strategy, avoid the appearance of invalid candidate regions of interest, and improve the performance of the region selection. The selection accuracy of the candidate region of interest is ensured, and the region selection efficiency is improved.

[0142] As an embodiment, whether the to-be-compared recognition result of the candidate region of interest triggers the stop selection condition includes:

[0143] According to the to-be-compared recognition result of the candidate region of interest, whether the stop selection condition is triggered is judged in combination with the standard recognition result.

[0144] Whether the to-be-compared recognition result of the candidate region of interest triggers the stop selection condition can be judged by using the to-be-compared recognition result and the standard recognition result.

[0145] In the embodiments of the present disclosure, when judging whether the candidate region of interest triggers the stop selection condition, the to-be-compared recognition result of the candidate region of interest can be compared with the standard recognition result to confirm the difference between the to-be-compared recognition result and the standard recognition result, and whether the candidate region of interest triggers the stop selection condition can be judged according to the difference between the to-be-compared recognition result and the standard recognition result, so as to realize accurate selection of the candidate region of interest and improve the judgment accuracy.

[0146] In a possible design, according to the to-be-compared recognition result of the candidate region of interest, whether the stop selection condition is triggered can be judged in combination with the standard recognition result, and the method can include the following steps.

[0147] According to the to-be-compared recognition result, a first detection rate corresponding to the candidate region of interest is calculated.

[0148] According to the standard recognition result, a second detection rate corresponding to the candidate region of interest is calculated.

[0149] A first ratio of the first detection rate to the second detection rate is calculated.

[0150] If the first ratio is less than a stop ratio threshold, it is determined that the stop selection condition is triggered.

[0151] If the first ratio is greater than or equal to the stop ratio threshold, it is determined that the stop selection condition is not triggered.

[0152] The calculation manner of the first detection rate is the same as the calculation manner of the first detection rate corresponding to the candidate region of interest calculated by using the to-be-compared recognition result in the foregoing embodiments, and details are not repeated here.

[0153] The second detection rate can be a ratio of a number of correction images in which lane line detection is normally performed to a total number of correction images.

[0154] Optionally, according to the standard recognition result, the second detection rate corresponding to the candidate region of interest can be calculated, which can include the following steps. According to the first lane line parameter corresponding to each of the at least one correction image, a second missed detection image in the at least one correction image is determined. A difference between a fourth image number of the at least one correction image and a fifth image number of the second missed detection image is calculated to obtain a sixth image number. A quotient of the sixth image number and the fourth image number is calculated to obtain the second detection rate.

[0155] The determination of the second missed detection image includes the following steps. From the first lane line parameter corresponding to each of the at least one correction image, a correction image corresponding to a lane line parameter with an empty parameter is selected as the second missed detection image.

[0156] The ratio threshold can be obtained in advance, for example, can be set to 80%, that is, the detection can be stopped when the first detection rate is less than 80% of the second detection rate.

[0157] In this embodiment of the disclosure, the second detection rate can be calculated for the corresponding candidate region of interest using the identification results to be compared, and the third detection rate can be calculated for the candidate region of interest based on the standard identification results. By comparing the first detection rate of the candidate region of interest with the second detection rate of the standard identification results, the detection effect of the candidate region of interest can be accurately compared, so as to confirm whether the candidate region of interest has triggered the stop selection condition and obtain an accurate trigger judgment result.

[0158] In some embodiments, the region selection strategy includes at least one; using the region selection strategy to select candidate regions of interest for the road video includes:

[0159] Based on the selection order corresponding to at least one region selection strategy, determine the target selection strategy from at least one region selection strategy;

[0160] A target selection strategy is used to select candidate regions of interest for road videos.

[0161] Optionally, the selection order corresponding to at least one region selection strategy can be randomly determined.

[0162] In this embodiment of the disclosure, when selecting candidate regions of interest using at least one region selection strategy, a target selection strategy can be determined from the at least one region selection strategy according to the selection order of each region selection strategy, so as to select candidate regions of interest for the road video using the target selection strategy. By using the selection order to select the target selection strategy, any region selection strategy can be selected, which can avoid the omission or duplicate selection of region selection strategies, and effectively improve the efficiency and accuracy of the use of region selection strategies.

[0163] like Figure 6 The diagram shows a flowchart of a method for selecting a region of interest based on vehicle navigation, according to a sixth embodiment of this disclosure. This method may include the following steps:

[0164] 601: Determine the standard recognition result corresponding to the road video captured by the vehicle-mounted camera.

[0165] 602: Determine the target selection strategy from at least one region selection strategy according to the selection order corresponding to each of the at least one region selection strategy.

[0166] 603: Utilize a target selection strategy to select candidate regions of interest for road videos.

[0167] 604: Perform lane line recognition on candidate regions of interest to obtain recognition results to be compared, and determine the selection result of the corresponding candidate regions of interest based on the recognition results to be compared and the standard recognition results.

[0168] 605: judging whether the to-be-compared recognition result of the candidate region of interest triggers a stop selection condition; if not, returning to step 603 for continuous execution; if yes, executing step 606.

[0169] 606: judging whether the at least one region selection strategy is ended, if not, returning to step 602 for continuous execution; if yes, executing step 607.

[0170] Optionally, the at least one region selection strategy is ended includes: if it is determined that the currently used target selection strategy is the last one in the at least one region selection strategy, the at least one region selection strategy is ended; if it is determined that the currently used target selection strategy does not reach the last one in the at least one region selection strategy, the at least one region selection strategy is not ended.

[0171] 607: stopping selecting the candidate region of interest for the road video by using the target selection strategy, and obtaining at least one candidate region of interest composed of all the selected candidate regions of interest for the road video.

[0172] 608: determining a target recognition result satisfying a result selection condition according to the selection result corresponding to the at least one candidate region of interest respectively, and determining the candidate region of interest corresponding to the target recognition result as a target region of interest.

[0173] In the embodiments of the present disclosure, at least one region selection strategy is provided. When a standard recognition result of a road video is obtained, a target selection strategy can be determined from the at least one region selection strategy according to a corresponding selection order of the at least one region selection strategy, so as to select a region of interest for the road video by using the target selection strategy. By performing lane line recognition on the candidate region of interest, a to-be-compared recognition result is obtained. The selection result of the candidate region of interest is determined according to the to-be-compared recognition result and the standard recognition result. Then, it can be determined whether the to-be-compared recognition result of the candidate region of interest meets a stop selection condition. If the to-be-compared recognition result does not trigger the stop selection condition, the selection of the candidate region of interest continues to be performed by using the current target selection strategy. If the to-be-compared recognition result triggers the stop selection condition, it is determined whether the at least one region selection strategy is ended. If the at least one region selection strategy is not ended, a new target selection strategy can be selected, and the selection of the candidate region of interest continues. When the to-be-compared recognition result of the final candidate region of interest triggers the stop selection condition and the at least one region selection strategy is ended, at least one candidate region of interest composed of all selected candidate regions of interest is obtained, the automatic and accurate selection of the at least one candidate region of interest is realized, the target recognition result is determined according to the selection result of the at least one candidate region of interest, the candidate region of interest corresponding to the target recognition result is selected as the target region of interest, the automatic and accurate selection of the target region of interest is realized, and the selection efficiency and selection accuracy are improved.

[0174] As an example, the region selection strategy includes at least one of a moving selection strategy, an enlargement selection strategy, and a reduction selection strategy.

[0175] According to the target selection strategy, a candidate region of interest is selected for the road video, including:

[0176] A basic region of interest in an image frame of the road video is determined.

[0177] In the global region corresponding to the image frame of the road video, a new candidate region of interest is obtained by moving from the region position of the previous candidate region of interest according to a predetermined moving step and a moving direction in a moving order, taking the basic region of interest as the first candidate region of interest.

[0178] And / or, in the global region corresponding to the image frame of the road video, a new candidate region of interest is obtained by enlarging the previous candidate region of interest according to a predetermined enlargement data, taking the basic region of interest as the first candidate region of interest.

[0179] And / or, in the global region corresponding to the image frame of the road video, the above basic interest region is the first candidate interest region, and the previous candidate interest region is reduced according to the predetermined reduction data according to the position of the region where the previous candidate interest region is located, to obtain a new candidate interest region.

[0180] Optionally, the mobile selection strategy can be combined with the magnification selection strategy to select the candidate interest region. In one possible design, the mobile selection strategy can be used to select each candidate interest region, and the magnification selection strategy can be used to select the candidate interest region based on the mobile selection strategy. In another possible design, the magnification selection strategy can be used to magnify the candidate interest region, and then each selected magnified candidate interest region can be moved according to the mobile selection strategy to select a new candidate interest region. The mobile selection strategy and the magnification selection strategy can be combined to comprehensively select the candidate interest region.

[0181] Optionally, the mobile selection strategy can be combined with the magnification selection strategy to select the candidate interest region. In one possible design, the mobile selection strategy can be used to select each candidate interest region, and the magnification selection strategy can be used to select the candidate interest region based on the mobile selection strategy. In another possible design, the magnification selection strategy can be used to magnify the candidate interest region, and then each selected magnified candidate interest region can be moved according to the mobile selection strategy to select a new candidate interest region. The mobile selection strategy and the magnification selection strategy can be combined to comprehensively select the candidate interest region.

[0182] Optionally, the mobile selection strategy, the magnification selection strategy, and the reduction selection strategy can be used alone, and the mobile selection strategy, the magnification selection strategy, and the reduction selection strategy can all be used as a region selection strategy to participate in the strategy traversal of at least one region selection strategy.

[0183] Of course, when the mobile selection strategy is combined with the magnification selection strategy, and / or the mobile selection strategy is combined with the reduction selection strategy, the mobile selection strategy and the magnification selection strategy can be used as a whole region selection strategy, and the mobile selection strategy and the reduction selection strategy can be used as a whole region selection strategy, so as to traverse the strategy selection of at least one region selection strategy.

[0184] When at least one of the above mobile selection strategy, magnification selection strategy, and reduction selection strategy is used to select the candidate interest region, the trigger judgment of the stop selection condition and the end judgment of the selection strategy traversal can be performed on the to-be-compared recognition result of each selected candidate interest region, so as to accurately judge the selection strategy.

[0185] In this embodiment, the region selection strategy includes at least one of a moving selection strategy, a zooming selection strategy, and a zooming selection strategy. Using a moving selection strategy, a corresponding region of interest can be generated by moving the region of interest based on a base region of interest, thus achieving moving selection of the region of interest. Furthermore, zooming selection can be performed on the region of interest to select candidate regions of interest of different sizes. By setting at least one region selection strategy, automatic and effective selection of candidate regions of interest can be achieved. During the selection process, the comparison and identification results of the candidate regions of interest can be used to determine whether the candidate regions of interest trigger the stop selection condition, thereby implementing selection constraints for each selection strategy, achieving efficient and effective region of interest selection, avoiding the selection of invalid regions of interest, and improving the selection efficiency and accuracy of regions of interest.

[0186] like Figure 7 The diagram shown is a structural schematic of a region of interest selection device based on vehicle navigation according to the seventh embodiment of this disclosure. The region of interest selection device 700 based on vehicle navigation may include the following units:

[0187] Standard acquisition unit 701: used to determine the standard recognition result corresponding to the road video captured by the vehicle-mounted camera.

[0188] Region selection unit 702: Used to select candidate regions of interest for road video using a region selection strategy to obtain at least one candidate region of interest.

[0189] Region analysis unit 703: used to identify lane lines in candidate regions of interest to obtain identification results to be compared, and to determine the selection result of the corresponding candidate regions of interest based on the identification results to be compared and the standard identification results.

[0190] Target determination unit 704: is used to determine the target recognition result that satisfies the result selection condition based on the selection results corresponding to at least one candidate region of interest, and to determine the candidate region of interest corresponding to the target recognition result as the target region of interest.

[0191] In the embodiments of the present disclosure, after determining the road video collected by the vehicle-mounted camera, a region selection strategy can be used to select a candidate region of interest for the road video to obtain at least one candidate region of interest and a selection result corresponding to each candidate region of interest, the selection result can be lane line recognition based on the candidate region of interest, and the target recognition result meeting the result selection condition can be determined according to the selection result corresponding to each candidate region of interest, so that the candidate region of interest corresponding to the target recognition result is determined as the target region of interest. Through automatic selection of the region of interest and result detection, the result selection condition can be used to automatically select the candidate region of interest, so that the target region of interest is automatically selected by using the selection result, and the selection efficiency can be improved by using the result selection constraint during selection.

[0192] As an embodiment, the standard obtaining unit comprises:

[0193] The video correction module is configured to perform correction processing on the road video to obtain a corrected video.

[0194] The result obtaining module is configured to perform lane line recognition on the corrected video to obtain a standard recognition result.

[0195] In a possible design, the result obtaining module comprises:

[0196] The first calculation sub-module is configured to perform lane line recognition on at least one corrected image in the corrected video respectively to obtain a first lane line parameter corresponding to each corrected image.

[0197] The first determination sub-module is configured to determine that the first lane line parameter corresponding to each corrected image constitutes the standard recognition result.

[0198] In some embodiments, the video correction module comprises:

[0199] The parameter obtaining sub-module is configured to obtain camera intrinsic parameters and distortion coefficients of the vehicle-mounted camera.

[0200] The video correction sub-module is configured to perform correction processing on the road video by using the camera intrinsic parameters and the distortion coefficients to obtain a corrected video.

[0201] As an optional manner, the parameter obtaining sub-module is specifically configured to:

[0202] The camera intrinsic parameters and the distortion correction coefficients of the vehicle-mounted camera are obtained by performing parameter calibration processing on the vehicle-mounted camera based on the checkerboard video.

[0203] In a possible design, the region analysis unit comprises:

[0204] The image acquisition module is configured to perform image cropping on at least one image frame in the road video respectively based on the candidate region of interest to obtain at least one region image corresponding to the road video.

[0205] The image recognition module is configured to perform lane line recognition on the at least one region image respectively to obtain second lane line parameters corresponding to the at least one region image respectively.

[0206] The result acquisition module is configured to determine that the second lane line parameters corresponding to the at least one region image respectively constitute a to-be-compared recognition result.

[0207] In some embodiments, the region analysis unit comprises:

[0208] The first determination module is configured to determine, according to the to-be-compared lane line recognition result and the standard lane line recognition result, an identification accuracy of the candidate region of interest.

[0209] The first calculation module is configured to calculate, according to the to-be-compared recognition result, a first detection rate corresponding to the candidate region of interest.

[0210] The second determination module is configured to determine a selection result of the candidate region of interest constituted by the identification accuracy and the first detection rate.

[0211] As a possible implementation manner, the first calculation module comprises:

[0212] The missed detection determination submodule is configured to determine, according to the second lane line parameters corresponding to the at least one region image respectively, a first missed detection image in the at least one region image.

[0213] The second calculation submodule is configured to calculate a difference between a first image number of the at least one region image and second image data of the first missed detection image to obtain a third image number.

[0214] The third calculation submodule is configured to calculate a quotient of the third image number and the first image number to obtain the first detection rate corresponding to the candidate region of interest.

[0215] In a possible design, the standard recognition result comprises: first lane line parameters corresponding to the at least one corrected image respectively; and the at least one corrected image is an image obtained by performing video correction on the road image. The first calculation module is specifically configured to:

[0216] The third determination submodule is configured to determine, for the at least one image frame in the road video, the region image and the corrected image corresponding to the at least one image frame respectively.

[0217] The difference calculation sub-module is configured to calculate lane line difference values by using the second lane line parameters corresponding to the region images of any image frame and the first lane line parameters corresponding to the corrected images of the image frame, so as to obtain lane line difference values corresponding to the at least one image frame respectively.

[0218] The variance calculation sub-module is configured to calculate lane line variances according to the lane line difference values corresponding to the at least one image frame respectively.

[0219] The normalization calculation sub-module is configured to normalize the lane line variances to obtain the recognition accuracy of the candidate region of interest.

[0220] In some embodiments, the missed detection determination sub-module is specifically configured to:

[0221] select lane line parameters with no parameters from the second lane line parameters corresponding to the at least one region image respectively.

[0222] determine that the region image corresponding to the empty lane line parameter is the first missed detection image.

[0223] In a possible design, the target determination unit includes:

[0224] The target selection module is configured to determine a target recognition result that satisfies a first selection condition in terms of the recognition accuracy and a second selection condition in terms of the first detection rate, according to the recognition accuracy corresponding to the at least one candidate region of interest respectively and the first detection rate.

[0225] In some embodiments, the method further includes:

[0226] The first selection unit is configured to, if it is determined that the to-be-compared recognition result of the candidate region of interest does not trigger the stop selection condition, return to the step of selecting the candidate region of interest for the road video by using the region selection strategy for continuous execution.

[0227] The second selection unit is configured to, if it is determined that the to-be-compared recognition result of the candidate region of interest triggers the stop selection condition, stop selecting the candidate region of interest for the road video by using the region selection strategy, and obtain at least one candidate region of interest constituted by all the candidate regions of interest selected for the road video.

[0228] As an embodiment, whether the to-be-compared recognition result of the candidate region of interest triggers the stop selection condition includes:

[0229] The selection judgment unit is configured to judge whether the stop selection condition is triggered according to the to-be-compared recognition result of the candidate region of interest in combination with the standard recognition result.

[0230] In some embodiments, the selection judgment unit includes:

[0231] The second calculation module is configured to calculate a second detection rate corresponding to the candidate region of interest according to the to-be-compared recognition result.

[0232] The third calculation module is configured to calculate a third detection rate corresponding to the candidate region of interest according to the standard recognition result.

[0233] The fourth calculation module is configured to calculate a first ratio of the first detection rate and the second detection rate.

[0234] The first comparison module is configured to determine that the stop selection condition is triggered if the first ratio is less than a stop ratio threshold.

[0235] The second comparison module is configured to determine that the stop selection condition is not triggered if the first ratio is greater than or equal to the stop ratio threshold.

[0236] As an optional mode, the region selection strategy includes at least one region selection strategy, and the region selection unit includes:

[0237] The strategy selection module is configured to determine a target selection strategy from the at least one region selection strategy according to a selection order corresponding to each of the at least one region selection strategy.

[0238] The region selection module is configured to select the candidate region of interest for the road video by using the target selection strategy.

[0239] In a possible design, the first selection unit includes:

[0240] The first selection module is configured to return to the step of selecting the candidate region of interest for the road video by using the target selection strategy if it is determined that the to-be-compared recognition result of the candidate region of interest does not trigger the stop selection condition.

[0241] The second selection module is configured to return to the step of determining the target selection strategy from the at least one region selection strategy according to the selection order corresponding to each of the at least one region selection strategy if it is determined that the to-be-compared recognition result of the candidate region of interest triggers the stop selection condition when the at least one region selection strategy is not iterated to an end.

[0242] The result determination module is configured to stop the step of selecting the candidate region of interest for the road video by using the target selection strategy and obtain at least one candidate region of interest composed of all the candidate regions of interest selected for the road video when the at least one region selection strategy is iterated to the end.

[0243] In some embodiments, the region selection strategy includes at least one of a moving selection strategy, an enlarging selection strategy, and a reducing selection strategy.

[0244] The region selection module includes:

[0245] a base determining sub-module, configured to determine a base region of interest in an image frame of a road video;

[0246] a moving selecting sub-module, configured to, in a global region corresponding to the image frame of the road video, take the base region of interest as a first candidate region of interest, and move the position of a previous candidate region of interest according to a predetermined moving step and a moving direction, to obtain a new candidate region of interest;

[0247] and / or an enlarging selecting sub-module, configured to, in the global region corresponding to the image frame of the road video, take the base region of interest as the first candidate region of interest, and enlarge the previous candidate region of interest according to predetermined enlargement data, to obtain the new candidate region of interest;

[0248] and / or a reducing selecting sub-module, configured to, in the global region corresponding to the image frame of the road video, take the base region of interest as the first candidate region of interest, and reduce the previous candidate region of interest according to predetermined reduction data, to obtain the new candidate region of interest.

[0249] It should be noted that the vehicle-mounted camera or the road video in the embodiment is not a camera or a video for a specific user, and cannot reflect the personal information of a specific user. It should be noted that the road video in the embodiment comes from a public data set.

[0250] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in the technical solution comply with relevant laws and regulations and do not violate public order and good customs.

[0251] According to the embodiments of the present disclosure, the present disclosure further provides an electronic device, a readable storage medium and a computer program product.

[0252] According to the embodiments of the present disclosure, the present disclosure further provides a computer program product, which comprises a computer program stored in a readable storage medium, and at least one processor of an electronic device can read the computer program from the readable storage medium, and the at least one processor executes the computer program to make the electronic device execute the scheme provided in any of the above embodiments.

[0253] Figure 8A schematic block diagram of an example electronic device 800 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0254] like Figure 8 As shown, device 800 includes a computing unit 801, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 802 or a computer program loaded from storage unit 808 into random access memory (RAM) 803. RAM 803 may also store various programs and data required for the operation of device 800. The computing unit 801, ROM 802, and RAM 803 are interconnected via bus 804. Input / output (I / O) interface 805 is also connected to bus 804.

[0255] Multiple components in device 800 are connected to I / O interface 805, including: input unit 806, such as keyboard, mouse, etc.; output unit 807, such as various types of monitors, speakers, etc.; storage unit 808, such as disk, optical disk, etc.; and communication unit 809, such as network card, modem, wireless transceiver, etc. Communication unit 809 allows device 800 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0256] The computing unit 801 can be various general and / or special purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 performs various methods and processes described above, such as the vehicle navigation based region of interest selection method. For example, in some embodiments, the vehicle navigation based region of interest selection method can be implemented as a computer software program tangibly embodied in a machine readable medium, such as the storage unit 808. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded onto the RAM 803 and executed by the computing unit 801, one or more steps of the vehicle navigation based region of interest selection method described above can be performed. Alternatively, in other embodiments, the computing unit 801 can be configured to perform the vehicle navigation based region of interest selection method by any other suitable means, such as by means of firmware.

[0257] Various implementations of the systems and techniques described above can be realized in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a programmable logic device (PLD), a computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0258] Program code for carrying out methods of the present disclosure can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces a means for implementing the functions / acts specified in the flowcharts and / or block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine and partially on a remote machine or entirely on a remote machine or server.

[0259] In the context of the present disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0260] To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0261] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0262] The computer system can include clients and servers. This relationship can be. remote, where each server is stored on a remote computer from a client. The clients and the servers can be connected through a communication network. The relationship can be a client-server relationship over a network. Servers can be cloud servers, also known as cloud computing servers or cloud hosts, which are a host product in the cloud computing service system. The servers can be servers of a distributed system, or servers combined with a blockchain.

[0263] It should be understood that the various forms of flow shown above can be reordered, steps added or removed. For example, the steps described in the present disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions of the present disclosure can be achieved, which are not limited herein.

[0264] The above detailed description does not constitute a limitation on the protection scope of the present disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for selecting a region of interest based on vehicle navigation, comprising: Determine the standard recognition results corresponding to the road videos captured by the vehicle-mounted camera; Using a region selection strategy, candidate regions of interest are selected for the road video to obtain at least one candidate region of interest; Lane line recognition is performed on the candidate regions of interest to obtain the recognition results to be compared, and the selection result of the corresponding candidate regions of interest is determined based on the recognition results to be compared and the standard recognition results; Based on the selection results corresponding to at least one of the candidate regions of interest, a target recognition result that satisfies the result selection condition is determined, and the candidate region of interest corresponding to the target recognition result is determined as the target region of interest; The process of identifying lane lines in the candidate regions of interest to obtain the identification results to be compared includes: Based on the candidate region of interest, at least one image frame in the road video is cropped to obtain at least one region image corresponding to the road video. Lane line recognition is performed on at least one of the said regional images to obtain the second lane line parameters corresponding to each of the at least one said regional images; Determine the second lane line parameters corresponding to at least one of the said region images to form the comparison and recognition result; The step of determining the selection result of the corresponding candidate region of interest based on the identification result to be compared and the standard identification result includes: The recognition accuracy of the candidate region of interest is determined based on the lane line recognition results to be compared and the standard lane line recognition results. Select lane line parameters that are empty from the second lane line parameters corresponding to at least one region image; The region image corresponding to the empty lane line parameters is determined as the first missed detection image; The difference between the number of first images of at least one of the region images and the second image data of the first missed image is calculated to obtain the number of third images; Calculate the quotient of the number of third images and the number of first images to obtain the first detection rate corresponding to the candidate region of interest; The recognition accuracy and the first detection rate are used to determine the selection result of the candidate region of interest.

2. The method according to claim 1, wherein, The determination of the standard recognition result corresponding to the road video captured by the vehicle-mounted camera includes: The road video captured by the vehicle-mounted camera is corrected to obtain a corrected video; Lane line recognition is performed on the corrected video to obtain standard recognition results.

3. The method according to claim 2, wherein, The process of performing lane line recognition on the corrected video to obtain a standard recognition result includes: Lane line recognition is performed on at least one corrected image in the corrected video to obtain first lane line parameters corresponding to at least one corrected image. The standard recognition result is formed by determining at least one first lane line parameter corresponding to each of the corrected images.

4. The method according to claim 2 or 3, wherein, The process of correcting the road video to obtain a corrected video includes: Obtain the camera intrinsic parameters and distortion coefficients of the vehicle-mounted camera; The road video is corrected using the camera intrinsic parameters and the distortion coefficients to obtain the corrected video.

5. The method according to claim 4, wherein, The process of obtaining the camera intrinsic parameters and distortion correction coefficients of the vehicle-mounted camera includes: The vehicle-mounted camera is calibrated based on a pre-set checkerboard video to obtain its intrinsic camera parameters and distortion correction coefficients.

6. The method according to claim 1, wherein, The standard recognition result includes: at least one first lane line parameter corresponding to each of the corrected images; at least one of the corrected images is an image obtained by video correction of a road image; The step of determining the recognition accuracy of the candidate region of interest based on the lane line recognition results to be compared and the standard lane line recognition results includes: For at least one image frame in the road video, determine at least one region image and a correction image corresponding to each image frame; For the second lane line parameters corresponding to the region image of any image frame and the first lane line parameters corresponding to the corrected image of the image frame, calculate the lane line difference to obtain at least one lane line difference corresponding to each of the image frames. Calculate the lane line variance based on the lane line difference value corresponding to at least one of the image frames; The variance of the lane lines is normalized to obtain the recognition accuracy of the candidate region of interest.

7. The method according to claim 6, wherein, The step of determining the target recognition result that satisfies the result selection condition based on the selection results corresponding to at least one of the candidate regions of interest includes: Based on the recognition accuracy and the first detection rate corresponding to at least one of the candidate regions of interest, a target recognition result is determined where the recognition accuracy satisfies the first selection condition and the first detection rate satisfies the second selection condition.

8. The method according to any one of claims 1-3 and 5-7, wherein, After determining the selection result of the corresponding candidate region of interest based on the identification result to be compared and the standard identification result, the method further includes: If it is determined that the comparison and identification results of the candidate region of interest do not trigger the stop selection condition, then return to the step of using the region selection strategy to select candidate regions of interest for the road video and continue to execute; If the comparison and identification result of the candidate region of interest triggers the stop selection condition, then the region selection strategy is stopped, candidate regions of interest are selected for the road video, and at least one candidate region of interest is obtained by combining all the candidate regions of interest selected for the road video.

9. The method according to claim 8, wherein, Whether the comparison and identification results of the candidate region of interest trigger the stop selection condition includes: Based on the comparison and identification results of the candidate regions of interest, and in conjunction with the standard identification results, it is determined whether the stop selection condition is triggered.

10. The method according to claim 9, wherein, The step of determining whether to trigger a stop selection condition based on the comparison and identification results of the candidate region of interest, combined with the standard identification results, includes: Based on the identification results to be compared, calculate the first detection rate corresponding to the candidate region of interest; Based on the standard identification results, calculate the second detection rate corresponding to the candidate region of interest; Calculate the first ratio of the first detection rate to the second detection rate; If the first ratio is less than the stop ratio threshold, a stop selection condition is determined to be triggered; If the first ratio is greater than or equal to the stop ratio threshold, it is determined that the stop selection condition has not been triggered.

11. The method according to any one of claims 9-10, wherein, The region selection strategy includes at least one; the step of using the region selection strategy to select candidate regions of interest for the road video includes: Based on the selection order corresponding to at least one of the region selection strategies, a target selection strategy is determined from at least one of the region selection strategies; The target selection strategy is used to select candidate regions of interest for the road video.

12. The method according to claim 11, wherein, If the comparison and recognition result of the candidate region of interest does not trigger the stop selection condition, the process returns to the step of selecting candidate regions of interest for the road video using the region selection strategy, including: If it is determined that the comparison and identification result of the candidate region of interest does not trigger the stop selection condition, the process returns to the step of using the target selection strategy to select candidate regions of interest for the road video and continue execution. If the comparison and recognition result of the candidate region of interest triggers a stop selection condition, the region selection strategy is stopped from being used to select candidate regions of interest for the road video, and at least one candidate region of interest is obtained by combining all the candidate regions of interest selected for the road video, including: If the comparison and identification result of the candidate region of interest triggers the stop selection condition, the process returns to the step of determining the target selection strategy from at least one region selection strategy according to the selection order corresponding to at least one region selection strategy before the process ends. When at least one of the region selection strategies is completed, the target selection strategy is stopped, candidate regions of interest are selected for the road video, and at least one candidate region of interest is obtained by combining all the candidate regions of interest selected for the road video.

13. The method according to claim 12, wherein, The region selection strategy includes at least one of the following: a moving selection strategy, a zoom-in selection strategy, and a zoom-out selection strategy; The step of selecting candidate regions of interest for the road video using the target selection strategy includes: Determine the basic region of interest in the image frames of the road video; In the global region corresponding to the image frame of the road video, the basic region of interest is taken as the first candidate region of interest. The new candidate region of interest is obtained by moving according to the predetermined movement step size and movement direction in the region where the previous candidate region of interest is located. And / or, in the global region corresponding to the image frame of the road video, taking the above-mentioned basic region of interest as the first candidate region of interest, and at the location of the previous candidate region of interest, the previous candidate region of interest is magnified according to the predetermined magnification data to obtain a new candidate region of interest; And / or, in the global region corresponding to the image frame of the road video, taking the aforementioned basic region of interest as the first candidate region of interest, and scaling down the previous candidate region of interest according to predetermined scaling down data at the location of the previous candidate region of interest, a new candidate region of interest is obtained.

14. A region of interest selection device based on vehicle navigation, comprising: The standard acquisition unit is used to determine the standard recognition result corresponding to the road video captured by the vehicle-mounted camera. A region selection unit is used to select candidate regions of interest for the road video using a region selection strategy, so as to obtain at least one candidate region of interest. The region analysis unit is used to identify lane lines in the candidate regions of interest to obtain comparison identification results, and to determine the selection result of the corresponding candidate regions of interest based on the comparison identification results and the standard identification results. The target determination unit is configured to determine a target recognition result that satisfies the result selection condition based on the selection results corresponding to at least one candidate region of interest, and to determine the candidate region of interest corresponding to the target recognition result as the target region of interest; The regional analysis unit includes: The image acquisition module is used to crop at least one image frame in the road video based on the candidate region of interest to obtain at least one region image corresponding to the road video; The image recognition module is used to perform lane line recognition on at least one of the said regional images respectively, and obtain the second lane line parameters corresponding to at least one of the said regional images respectively; The result acquisition module is used to determine that the second lane line parameters corresponding to at least one of the region images constitute the comparison and recognition result; The regional analysis unit includes: The first determining module is used to determine the recognition accuracy of the candidate region of interest based on the lane line recognition results to be compared and the standard lane line recognition results; The first calculation module is used to select lane line parameters that are empty from the second lane line parameters corresponding to at least one region image respectively; The region image corresponding to the empty lane line parameters is determined as the first missed detection image; The difference between the number of first images of at least one of the region images and the second image data of the first missed image is calculated to obtain the number of third images; Calculate the quotient of the number of third images and the number of first images to obtain the first detection rate corresponding to the candidate region of interest; The second determining module is used to determine the selection result of the candidate region of interest composed of the recognition accuracy and the first detection rate.

15. The apparatus according to claim 14, wherein, The standard acquisition unit includes: The video correction module is used to correct the road video to obtain a corrected video. The result acquisition module is used to perform lane line recognition on the corrected video and obtain standard recognition results.

16. The apparatus according to claim 15, wherein, The result acquisition module includes: The first calculation submodule is used to perform lane line recognition on at least one corrected image in the corrected video to obtain first lane line parameters corresponding to at least one corrected image. The first determining submodule is used to determine at least one first lane line parameter corresponding to each of the corrected images to constitute the standard recognition result.

17. The apparatus according to claim 15 or 16, wherein, The video correction module includes: The parameter acquisition submodule is used to acquire the camera intrinsic parameters and distortion coefficients of the vehicle-mounted camera; The video correction submodule is used to perform correction processing on the road video using the camera intrinsic parameters and the distortion coefficients to obtain the corrected video.

18. The apparatus according to claim 17, wherein the parameter acquisition submodule is specifically used for: The vehicle-mounted camera is calibrated using checkerboard video to obtain its intrinsic camera parameters and distortion correction coefficients.

19. The apparatus according to claim 14, wherein, The standard recognition result includes: at least one first lane line parameter corresponding to each of the at least one corrected image; at least one of the corrected images is an image obtained by video correction of a road image; the first calculation module is specifically used for: The third determining submodule is used to determine, for at least one image frame in the road video, at least one corresponding region image and a correction image; The difference calculation submodule is used to calculate the lane line difference for the second lane line parameters corresponding to the region image of any image frame and the first lane line parameters corresponding to the corrected image of the image frame, so as to obtain at least one lane line difference corresponding to each of the image frames. The variance calculation submodule is used to calculate the lane line variance based on the lane line difference values ​​corresponding to at least one of the image frames. The normalization calculation submodule is used to normalize the lane line variance to obtain the recognition accuracy of the candidate region of interest.

20. The apparatus according to claim 19, wherein, The target determination unit includes: The target selection module is used to determine the target recognition result based on the recognition accuracy and the first detection rate corresponding to at least one of the candidate regions of interest, respectively, and to determine the target recognition result whose recognition accuracy satisfies the first selection condition and whose first detection rate satisfies the second selection condition.

21. The apparatus according to any one of claims 14-16, 18-20, wherein, Also includes: The first selection unit is configured to return to the step of selecting candidate regions of interest for the road video using the region selection strategy if the comparison and recognition result of the candidate region of interest is determined not to trigger the stop selection condition. The second selection unit is configured to stop using the region selection strategy to select candidate regions of interest for the road video if the comparison and recognition result of the candidate region of interest is determined to trigger the stop selection condition, and to obtain at least one candidate region of interest composed of all candidate regions of interest selected for the road video.

22. The apparatus according to claim 21, wherein, Whether the comparison and identification results of the candidate region of interest trigger the stop selection condition includes: The selection and judgment unit is used to determine whether to trigger the stop selection condition based on the comparison and recognition results of the candidate region of interest and the standard recognition results.

23. The apparatus according to claim 22, wherein, The selection and determination unit includes: The second calculation module is used to calculate the second detection rate corresponding to the candidate region of interest based on the identification result to be compared; The third calculation module is used to calculate the third detection rate corresponding to the candidate region of interest based on the standard recognition result. The fourth calculation module is used to calculate a first ratio between the first detection rate and the second detection rate; The first comparison module is used to determine the triggering of the stop selection condition if the first ratio is less than the stop ratio threshold. The second comparison module is used to determine that the stop selection condition has not been triggered if the first ratio is greater than or equal to the stop ratio threshold.

24. The apparatus according to any one of claims 22-23, wherein, The region selection strategy includes at least one; the region selection unit includes: The strategy selection module is used to determine a target selection strategy from at least one of the region selection strategies according to the selection order corresponding to each of the at least one region selection strategy. The region selection module is used to select candidate regions of interest for the road video using the target selection strategy.

25. The apparatus according to claim 24, wherein, The first selection unit includes: The first selection module is used to return to the step of selecting candidate regions of interest for the road video and continuing execution if the comparison and recognition result of the candidate region of interest does not trigger the stop selection condition. The second selection module is used to, if the comparison and identification result of the candidate region of interest triggers the stop selection condition, return to the step of determining the target selection strategy from the at least one region selection strategy according to the selection order corresponding to the at least one region selection strategy before the at least one region selection strategy has been traversed and continue to execute. The result determination module is used to stop using the target selection strategy when at least one of the region selection strategies has been traversed, to select candidate regions of interest for the road video, and to obtain at least one candidate region of interest composed of all the candidate regions of interest selected for the road video.

26. The apparatus according to claim 25, wherein, The region selection strategy includes at least one of the following: a moving selection strategy, a zoom-in selection strategy, and a zoom-out selection strategy; The region selection module includes: The basic determination submodule is used to determine the basic region of interest in the image frames of the road video; The motion selection submodule is used to select a new candidate region of interest from the global region corresponding to the image frame of the road video, taking the basic region of interest as the first candidate region of interest, and moving in a predetermined motion step size and motion direction order according to the location of the previous candidate region of interest. And / or, the magnification selection submodule is used to, in the global region corresponding to the image frame of the road video, take the above-mentioned basic region of interest as the first candidate region of interest, and magnify the previous candidate region of interest according to the predetermined magnification data at the location of the previous candidate region of interest to obtain a new candidate region of interest; And / or, the zoom selection submodule is used to zoom out the previous candidate region of interest according to predetermined zoom data in the global region corresponding to the image frame of the road video, taking the above-mentioned basic region of interest as the first candidate region of interest, and to obtain a new candidate region of interest.

27. An electronic device comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-13.

28. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-13.

29. A computer program product comprising a computer program that, when executed by a processor, implements the steps of the method according to any one of claims 1-13.

Images