Obstacle detection method and device, electronic equipment and storage medium

Abstract

The application relates to the technical field of image recognition, and provides a detection method and device for obstacles, electronic equipment and a storage medium, the method comprising the following steps: acquiring 2D image information in the driving direction of a current vehicle; determining feature points in the 2D image information; the feature points are pixel points intersecting with a ground region in each frame of picture in the image information; determining a first detection contour in the 2D image information according to the feature points; the first detection contour is a contour representing an obstacle; determining a second detection contour according to the first detection contour, and displaying the second detection contour on the 2D image information; the second detection contour is a contour obtained by screening and modifying the first detection contour; by determining the pixel points intersecting with the ground region in each frame of picture in the image information, the contour of the obstacle in the image information is displayed based on the pixel points, so that the obstacle in front of the driving vehicle can be accurately positioned.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, and in particular to an obstacle detection method, system, electronic device, and storage medium. Background Technology

[0002] General obstacles mainly refer to target types in autonomous driving scenarios that are difficult to classify uniformly, such as rocks, fallen trees, discarded tires, and any abnormal protrusions on the ground. Existing methods for detecting general obstacles in autonomous driving scenarios mainly include: combining ultrasonic waves, millimeter waves, and LiDAR to detect general obstacles; segmenting drivable areas in images; and extracting general obstacles using traditional computer vision (CV) methods.

[0003] However, ultrasonic detection has a short range and is only suitable for low-speed scenarios (such as parking), and it has blind spots; radar detection mainly has a high reflection response to metallic objects and is prone to false detections; laser scanning is difficult to detect small objects; image segmentation-based methods generally have poor stability and generalization, and are prone to missed detections and false detections; traditional CV methods have complex processing procedures and generally poor results. Summary of the Invention

[0004] To address the problems existing in the prior art, the present invention provides an obstacle detection method, system, electronic device, and storage medium.

[0005] In a first aspect, the present invention provides a method for detecting obstacles, comprising:

[0006] Obtain 2D image information of the current vehicle in the direction of travel;

[0007] The feature points in the 2D image information are determined; the feature points are the pixels in each frame of the image information that intersect with the ground area.

[0008] A first detection contour is determined in the 2D image information based on the feature points; the first detection contour is a contour representing an obstacle.

[0009] A second detection contour is determined based on the first detection contour, and the second detection contour is displayed on the 2D image information; the second detection contour is a contour that has been filtered and modified from the first detection contour.

[0010] In one embodiment, determining a first detection contour in the 2D image information based on feature points includes:

[0011] All identified feature points are smoothly connected to determine the first detection contour in the 2D image information; wherein, two feature points whose distance between adjacent feature points is greater than a preset distance are not smoothly connected.

[0012] In one embodiment, determining the second detection contour based on the first detection contour includes:

[0013] The first detected contour with two endpoints is compared with each standard contour in the pre-stored set of standard contours to determine the standard contour with the greatest similarity.

[0014] Determine the scaling ratio between a first detected contour with two endpoints and a standard contour with the highest similarity, and determine the incomplete contour corresponding to the first detected contour with two endpoints based on the scaling ratio;

[0015] The incomplete contour and the first detection contour with two endpoints are integrated into a second detection contour.

[0016] In one embodiment, determining the second detection contour based on the first detection contour includes:

[0017] The first detection profile, which does not have two endpoints, is directly identified as the second detection profile.

[0018] In one embodiment, determining the feature points in the 2D image information includes:

[0019] The pixel features of each frame in the 2D image information are input into the recognition model to obtain the feature points output by the recognition model.

[0020] The recognition model is a model trained by machine learning, which takes the pixel features of the intersection of obstacles and ground areas in the image sample as input, and is used to judge the pixel points where obstacles and ground areas intersect in the image.

[0021] In one embodiment, the recognition model includes a recognition layer and an association layer. The recognition layer is used to identify ground regions based on the pixel features of the image; the association layer is used to determine feature points based on the pixel features of the image and the edge pixel features of the ground regions. Accordingly, the pixel features of each frame in the 2D image information are input into the recognition model to obtain feature points output by the recognition model, including:

[0022] The pixel features of the image are input into the recognition layer, which then compares the pixel features of the image with the pre-stored pixel features of the ground to determine the ground area in the image.

[0023] The edge pixel features of the ground area and the pixel features of the image are input into the association layer. The association layer determines the intersecting edges based on the pixel features of the image and the edge pixel features of the ground area, and selects pixels as feature points based on the intersecting edges.

[0024] In one embodiment, the feature point is the pixel in each frame of the image information that intersects with the ground area and is closest to the bottom edge of the image.

[0025] In a second aspect, the present invention provides an obstacle detection device, comprising:

[0026] The acquisition module is used to acquire 2D image information of the current vehicle in the driving direction;

[0027] The recognition module is used to determine the feature points in the 2D image information; the feature points are the pixels in each frame of the image information that intersect with the ground area;

[0028] The processing module is used to determine a first detection contour in the 2D image information based on the feature points; the first detection contour is a contour representing an obstacle.

[0029] The detection module is used to determine a second detection contour based on the first detection contour and display the second detection contour on the 2D image information; the second detection contour is a contour that has been filtered and modified from the first detection contour.

[0030] Thirdly, the present invention provides an electronic device, including a memory and a memory storing a computer program, wherein the processor executes the program to implement the steps of the obstacle detection method described in the first aspect.

[0031] Fourthly, the present invention provides a processor-readable storage medium storing a computer program for causing the processor to perform the steps of the obstacle detection method described in the first aspect.

[0032] The obstacle detection method, system, electronic device, and storage medium provided by this invention determine the pixels that intersect with the ground area in each frame of the image information, and realize the outline display of the obstacle in the image information based on the pixels, so as to accurately locate the obstacle in front of the driving vehicle. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0034] Figure 1 This is a flowchart illustrating the obstacle detection method provided by the present invention;

[0035] Figure 2 This is a schematic diagram of the obstacle detection device provided by the present invention;

[0036] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention; Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0038] The following is combined Figures 1-3 The present invention describes an obstacle detection method, system, electronic device, and storage medium.

[0039] Figure 1 A flowchart illustrating an obstacle detection method according to the present invention is shown below. Figure 1 The method includes:

[0040] 11. Obtain 2D image information of the current vehicle in the direction of travel;

[0041] 12. Determine the feature points in the 2D image information; the first feature point is the pixel in each frame of the image information that intersects with the ground area;

[0042] 13. Determine the first detection contour in the 2D image information based on feature points; the first detection contour is the contour representing the obstacle;

[0043] 14. Determine the second detection contour based on the first detection contour, and display the second detection contour on the 2D image information; the second detection contour is the contour that is filtered and modified from the first detection contour.

[0044] Regarding steps 11-14, it should be noted that in this invention, the current vehicle is an intelligent vehicle equipped with autonomous driving capabilities. This vehicle is equipped with a forward-facing 2D camera device for acquiring 2D image information along the driving direction. The 2D camera device on the current vehicle can collect the required image information in real time.

[0045] After the image information is acquired, each frame of the image information needs to be identified in real time to identify any obstacles that may exist in the image and mark them on the image (e.g., with outline lines).

[0046] In this invention, autonomous vehicles primarily operate on designated main roads. Obstacles such as vehicles, rocks, trees, pedestrians, and bicycles appearing on these roads will have intersecting edges with the ground in the image. Therefore, it is necessary to identify the pixels where obstacles intersect with the ground in the image; these pixels serve as feature points in the detection process of this invention. Since each obstacle and ground interaction presents a different scenario, the pixel changes at the intersection edges exhibit regularity across different scenarios. Therefore, the intersection edges can be identified based on these pixel changes, and pixels on these intersection edges can then be selected as feature points.

[0047] In this invention, after identifying feature points in the image, these feature points are concatenated to form contour lines, which serve as preliminary detection contours representing potential obstacles obtained from image detection. For ease of subsequent description of the scheme, these detection contours are referred to here as the first detection contours.

[0048] Regarding the detection contours that may represent obstacles, it should be noted that some obstacles in the image do not intersect with the ground area entirely.

[0049] For example, a pedestrian on the road has an edge that intersects with the ground in the picture, while their upper body has an edge that intersects with the buildings or the sky in the picture.

[0050] For example, a large tree blown down by the wind lies across the road. Some of its lush branches intersect with the ground in the picture, while others intersect with buildings or the sky in the picture.

[0051] For example, the tires and lower body of the vehicle in front intersect with the ground area in the image, while the upper body intersects with the sky in the image.

[0052] Therefore, the first detected contour obtained based on feature points contains both complete closed-loop contours and incomplete open-loop contours.

[0053] In this invention, the first detection profile is then screened and refined to obtain a more complete detection profile. To facilitate differentiation from the detection profiles mentioned above, these refined detection profiles are referred to as the second detection profile.

[0054] For example, incomplete open-loop contours are repaired to obtain complete open-loop contours. The edges and corners of complete closed-loop contours are then processed to obtain more standardized complete open-loop contours. These processed contours are classified as the second detection contours mentioned above.

[0055] The second detected contour is then displayed on the corresponding screen. Since multiple frames constitute real-time 2D image information, the obtained second detected contour is dynamically displayed on the 2D image information.

[0056] The obstacle detection method provided by this invention determines the pixels that intersect with the ground area in each frame of the image information, and then uses these pixels to reconstruct and display the outline of the obstacle in the image information, thereby achieving the purpose of accurately locating obstacles in front of a moving vehicle.

[0057] The further explanation of the above method mainly focuses on the process of determining the first detection contour in 2D image information based on feature points, as detailed below:

[0058] All identified feature points are smoothly connected to determine the first detection contour in the 2D image information; however, two feature points whose distance between adjacent feature points is greater than a preset distance are not smoothly connected.

[0059] It should be noted that in this invention, there are situations where multiple obstacles appearing on a single screen are spaced apart. For example, different vehicles may have gaps between them on the road. In this case, the intersection edge of each obstacle with the ground area in the screen exists independently.

[0060] There are also cases where multiple obstacles appear in a single frame and overlap. For example, when a vehicle cuts in front of another vehicle on a road, multiple nearby vehicles may appear in the frame, their edges intersecting with the ground area.

[0061] In this invention, obstacles are marked using contour lines. All identified feature points are then smoothly connected to obtain individual detection contours. However, since many obstacles mentioned above exist independently at their intersection edges with the ground area in the image, a preset distance needs to be determined. When the distance between adjacent feature points is greater than the preset distance, it is equivalent to these two feature points belonging to two obstacles with gaps. In this case, these two feature points do not need to be smoothly connected. Through the aforementioned distance constraint, possible obstacles on the image can be distinguished.

[0062] A further method of the present invention, by smoothly connecting all feature points and limiting the spacing between feature points, can divide the image into multiple detection contours, thereby enabling simple differentiation of areas in the image that may be obstacles.

[0063] The further explanation of the above method mainly focuses on the process of determining the second detection contour based on the first detection contour, as follows:

[0064] The first detected contour with two endpoints is compared with each standard contour in the pre-stored set of standard contours to determine the standard contour with the greatest similarity.

[0065] Determine the scaling ratio between the first detected contour with two endpoints and the standard contour with the highest similarity, and determine the incomplete contour corresponding to the first detected contour with two endpoints based on the scaling ratio.

[0066] The incomplete contour and the first detection contour with two endpoints are integrated into a second detection contour.

[0067] It should be noted that, in this invention, the first detection profile with two endpoints is a non-complete open-loop profile, that is, the profile has an opening. The reason for the opening is that the obstacle corresponding to the profile does not have an intersecting edge with the ground area (such as having an intersecting edge with the sky).

[0068] In this invention, a set of standard contours is established, which stores a large number of standard contours. These standard contours are overall contours of different obstacles and the ground at different angles and sizes, in an intersecting state.

[0069] For example, changes in the distance between the vehicle in front of the current vehicle and the current vehicle will result in different intersection contours of the vehicle in front and the ground area captured by the current vehicle. Therefore, the standard contour set stores a reasonable number of contour data suitable for the intersection of the vehicle in front and the ground.

[0070] For example, the distance between pedestrians and vehicles, as well as their walking posture, will cause the intersection contours of pedestrians and the ground area captured by the vehicle to vary. Therefore, the standard contour set stores a reasonable number of contour data suitable for the intersection of pedestrians and the ground.

[0071] For other types of obstacles, the principle is the same as in the examples above; reasonable contour data will also be stored in the standard contour set.

[0072] In this invention, it is necessary to repair incomplete open-loop contours, that is, to supplement the open-loop contours into closed-loop and complete contours. Therefore, the detection contour with two endpoints is compared with each standard contour in the pre-stored standard contour set to determine the standard contour with the highest similarity as the contour of the same obstacle as the detection contour with two endpoints.

[0073] In this invention, since the standard contour may differ in size from the detection contour with two endpoints, the corresponding partial contour size of the standard contour with the greatest similarity to the detection contour with two endpoints is converted based on the partial contour size of the detection contour with two endpoints to obtain the scaling ratio between the two. Then, the incomplete contour corresponding to the first detection contour with two endpoints is determined according to the scaling ratio, that is, the contour of another part of the detection contour with two endpoints is restored according to the scaling ratio.

[0074] Finally, the incomplete contour and the first detection contour with two endpoints are integrated into a complete contour, the second detection contour. To easily distinguish it from the detection contour before integration, this integrated detection contour is called the second detection contour.

[0075] In addition, for the first detection contour that does not have two endpoints, it is actually a complete closed-loop contour, and these contours are directly determined as the second detection contour.

[0076] A further method of the present invention compares the detected contour with a standard contour to achieve complete correction of the incomplete detected contour, thereby obtaining a more complete contour indication of the obstacle.

[0077] The further explanation of the above method mainly focuses on the process of determining feature points in 2D image information, as detailed below:

[0078] The pixel features of each frame in the 2D image information are input into the recognition model to obtain the feature points output by the recognition model.

[0079] The recognition model is trained by machine learning based on the pixel features of the intersection of obstacles and ground areas in the image sample as input, and is used to determine the pixel points where obstacles and ground areas intersect in the image.

[0080] It should be noted that in this invention, the obtained feature points are pixels at the intersection of the obstacle and the ground area. Therefore, it is necessary to acquire the pixel features of each frame in the 2D image information, and then input these pixel features into the recognition model trained by the neural network model. The recognition model then processes the pixel features and outputs the feature points in the image.

[0081] In this invention, the recognition model is a model trained by machine learning, which takes the pixel features of the intersection of obstacles and ground areas in the image sample as input, and is used to determine the pixel points where obstacles and ground areas intersect in the image.

[0082] It should be noted that different types of obstacles can be placed on the ground, and then images can be captured using a camera. The captured images will show relatively complete intersection edges between the obstacles and the ground. The pixel features of these intersection edges can be obtained. That is, the pixel feature points of the obstacles near the intersection edges, as well as the pixel feature points of the ground area, can be obtained. Furthermore, since the shape of the obstacles may not be regular, directly obtaining the complete intersection edges of irregular obstacles with the ground may result in irregular contours after subsequent contour matching based on feature points, leading to inaccurate or difficult matching. Therefore, the pixel features of these intersection edges can be obtained by determining the pixels of the obstacles near the intersection edges that are closest to the bottom edge of the image, as well as the pixels of the ground area, based on each column of pixels in the image. For example, if the target obstacle has 20 pixels intersecting with the ground in a certain column of pixels, only the pixel closest to the bottom edge of the image needs to be selected.

[0083] These pixel feature points are input into the initial neural network model for training to obtain the recognition model.

[0084] In this invention, the recognition model can be updated and upgraded. The pixel features of each frame in the 2D image information are input into the recognition model, which can determine whether the pixel features of the entire image are intersecting edge features. If the determination is successful, the corresponding feature points can be output.

[0085] The present invention further explains the recognition model, which includes a recognition layer and an association layer. The recognition layer is used to identify the ground area based on the pixel features of the image; the association layer is used to determine feature points based on the pixel features of the image and the edge pixel features of the ground area.

[0086] Accordingly, the process of inputting the pixel features of each frame in the 2D image information into the recognition model to obtain the feature points output by the recognition model is specifically implemented as follows:

[0087] The pixel features of the image are input into the recognition layer, which then compares the pixel features of the image with the pre-stored pixel features of the ground to determine the ground area in the image.

[0088] The edge pixel features of the ground area and the pixel features of the image are input into the association layer. The association layer determines the intersecting edges based on the pixel features of the image and the edge pixel features of the ground area, and selects the pixels as feature points based on the intersecting edges.

[0089] It should be noted that, to facilitate the recognition of pixel features, the ground area in the image is first identified. Since the ground can be asphalt, gravel, or dirt roads, pixel features of different types of ground roads are pre-stored. The recognition layer in the model compares the pixel features of the image with the pre-stored pixel features of the ground to obtain the ground area in the image.

[0090] After obtaining the ground region in the image, we can obtain the edge pixel features of the ground region, that is, the pixel features at the edge positions of the ground region. These edge features are the pixels closest to the intersection edge with the obstacle. Therefore, through the model's association layer, we can extract pixel features that show significant differences between the image's pixel features and the ground region's edge pixel features. Then, based on the extracted pixel features and the ground region's edge pixel features, we can obtain the intersecting edges, and select pixels as feature points based on these intersecting edges. For example, we can select the pixel closest to the edge line as the feature point.

[0091] A further method of the present invention identifies feature points where obstacles intersect with the ground area based on pixel features using a trained model, which is beneficial for quickly identifying the probability of obstacles in image information.

[0092] The obstacle detection device provided by the present invention is described below. The obstacle detection device described below and the obstacle detection method described above can be referred to in correspondence.

[0093] Figure 2 A schematic diagram of the structure of an obstacle detection device provided by the present invention is shown below. Figure 2 The device includes an acquisition module 21, an identification module 22, a processing module 23, and a detection module 24, wherein:

[0094] The acquisition module 21 is used to acquire 2D image information of the current vehicle in the driving direction;

[0095] The recognition module 22 is used to determine feature points in 2D image information; the feature points are the pixels in each frame of the image information that intersect with the ground area;

[0096] Processing module 23 is used to determine a first detection contour in the 2D image information based on feature points; the first detection contour is a contour representing an obstacle;

[0097] The detection module 24 is used to determine the second detection contour based on the first detection contour and display the second detection contour on the 2D image information; the second detection contour is a contour that is filtered and modified from the first detection contour.

[0098] In a further description of the above device, the processing module is specifically used for:

[0099] All identified feature points are smoothly connected to determine the first detection contour in the 2D image information; wherein, two feature points whose distance between adjacent feature points is greater than a preset distance are not smoothly connected.

[0100] In a further description of the above device, the processing module, in the process of determining the second detection contour based on the first detection contour, is specifically used for:

[0101] The first detected contour with two endpoints is compared with each standard contour in the pre-stored set of standard contours to determine the standard contour with the greatest similarity.

[0102] Determine the scaling ratio between a first detected contour with two endpoints and a standard contour with the highest similarity, and determine the incomplete contour corresponding to the first detected contour with two endpoints based on the scaling ratio;

[0103] The incomplete contour and the first detection contour with two endpoints are integrated into a second detection contour.

[0104] In a further description of the above device, the processing module, in the process of determining the second detection contour based on the first detection contour, is specifically used for:

[0105] The first detection profile, which does not have two endpoints, is directly identified as the second detection profile.

[0106] In a further description of the above device, the identification module is specifically used for:

[0107] The pixel features of each frame in the 2D image information are input into the recognition model to obtain the feature points output by the recognition model.

[0108] The recognition model is a model trained by machine learning, which takes the pixel features of the intersection of obstacles and ground areas in the image sample as input, and is used to judge the pixel points where obstacles and ground areas intersect in the image.

[0109] In a further description of the above device, the recognition model includes a recognition layer and an association layer. The recognition layer is used to identify the ground region based on the pixel features of the image; the association layer is used to determine feature points based on the pixel features of the image and the edge pixel features of the ground region. Accordingly, in the process of inputting the pixel features of each frame of the 2D image information into the recognition model to obtain the feature points output by the recognition model, the recognition module is specifically used for:

[0110] The pixel features of the image are input into the recognition layer, which then compares the pixel features of the image with the pre-stored pixel features of the ground to determine the ground area in the image.

[0111] The edge pixel features of the ground area and the pixel features of the image are input into the association layer. The association layer determines the intersecting edges based on the pixel features of the image and the edge pixel features of the ground area, and selects pixels as feature points based on the intersecting edges.

[0112] Since the device described in this embodiment of the invention is based on the same principle as the method described in the above embodiments, more detailed explanations will not be repeated here.

[0113] It should be noted that, in the embodiments of the present invention, the relevant functional modules can be implemented by a hardware processor.

[0114] The obstacle detection device provided by the present invention determines the pixels that intersect with the ground area in each frame of the image information, and realizes the outline display of the obstacle in the image information based on the pixels, so as to accurately locate the obstacle in front of the driving vehicle.

[0115] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include a processor 31, a communication interface 32, a memory 33, and a communication bus 34. The processor 31, communication interface 32, and memory 33 communicate with each other via the communication bus 34. The processor 31 can call a computer program in the memory 33 to execute the steps of an obstacle detection method, such as: acquiring 2D image information of the current vehicle in its driving direction; determining feature points in the 2D image information; the first feature points are pixels in each frame of the image information that intersect with the ground area; determining a first detection contour in the 2D image information based on the feature points; the first detection contour being a contour representing an obstacle; determining a second detection contour based on the first detection contour; displaying the second detection contour on the 2D image information; and the second detection contour being a contour that has been filtered and refined from the first detection contour.

[0116] Furthermore, the logical instructions in the aforementioned memory 33 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0117] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein when the program instructions are executed by a computer, the computer is able to execute the obstacle detection method provided by the above methods, the method comprising: acquiring 2D image information of the current vehicle in the driving direction; determining feature points in the 2D image information; the first feature points being pixels in each frame of the image information that intersect with the ground area; determining a first detection contour in the 2D image information based on the feature points; the first detection contour being a contour representing an obstacle; determining a second detection contour based on the first detection contour, and displaying the second detection contour on the 2D image information; the second detection contour being a contour that has been filtered and modified from the first detection contour.

[0118] On the other hand, embodiments of this application also provide a processor-readable storage medium storing a computer program. The computer program is used to cause the processor to execute the obstacle detection method provided in the above embodiments, for example including: acquiring 2D image information of the current vehicle in the driving direction; determining feature points in the 2D image information; the first feature point being a pixel in each frame of the image information that intersects with the ground area; determining a first detection contour in the 2D image information based on the feature points; the first detection contour being a contour representing an obstacle; determining a second detection contour based on the first detection contour; displaying the second detection contour on the 2D image information; the second detection contour being a contour that has been filtered and modified from the first detection contour.

[0119] The processor-readable storage medium can be any available medium or data storage device that the processor can access, including but not limited to magnetic memory (e.g., floppy disk, hard disk, magnetic tape, magneto-optical disk (MO)), optical memory (e.g., CD, DVD, BD, HVD), and semiconductor memory (e.g., ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)).

[0120] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0121] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0122] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method of detecting an obstacle, characterized by, The method comprises the following steps: acquiring 2D image information in the driving direction of the current vehicle; determining feature points in the 2D image information; the feature points are pixel points intersecting with the ground region in each frame of the image information and are dynamically identified by a machine learning model; determining a first detection contour in the 2D image information according to the feature points; the first detection contour is a contour representing an obstacle; determining a second detection contour according to the first detection contour and displaying the second detection contour on the 2D image information; the second detection contour is a contour obtained by screening and modifying the first detection contour; the step of determining the first detection contour in the 2D image information according to the feature points comprises the following steps: smoothly connecting all the determined feature points to determine the first detection contour in the 2D image information; wherein two feature points with a distance greater than a preset distance are not smoothly connected; the step of determining the second detection contour according to the first detection contour comprises the following steps: comparing the first detection contour with each standard contour in a pre-stored standard contour set to determine a standard contour with the maximum similarity; determining a scaling ratio between the first detection contour with two end points and the standard contour with the maximum similarity, and determining a defective contour corresponding to the first detection contour with two end points according to the scaling ratio; integrating the defective contour and the first detection contour with two end points to obtain the second detection contour.

2. The method of obstacle detection according to claim 1, wherein, the step of determining the second detection contour according to the first detection contour comprises the following step: directly determining the first detection contour without two end points as the second detection contour.

3. The method of obstacle detection according to claim 1, wherein, the step of determining the feature points in the 2D image information comprises the following step: inputting the pixel features of each frame of the 2D image information into the identification model to obtain the feature points output by the identification model; wherein the identification model is obtained by machine learning training by taking the pixel features of the intersection between the obstacle and the ground region in the picture sample as the input, and is a model for judging the pixel points of the intersection between the obstacle and the ground region in the picture.

4. The method of obstacle detection according to claim 3, characterized in that, The identification model comprises an identification layer and an association layer; the identification layer is used to identify the ground region according to the pixel features of the picture; and the association layer is used to determine the feature points according to the pixel features of the picture and the edge pixel features of the ground region; correspondingly, the step of inputting the pixel features of each frame of the 2D image information into the identification model to obtain the feature points output by the identification model comprises the following steps: inputting the pixel features of the picture into the identification layer, comparing the pixel features of the picture with the pre-stored pixel features of the ground region by the identification layer, and determining the ground region in the picture; inputting the edge pixel features of the ground region and the pixel features of the picture into the association layer, determining the intersection edge according to the pixel features of the picture and the edge pixel features of the ground region by the association layer, and selecting the pixel points as the feature points based on the intersection edge.

5. The method of obstacle detection according to any one of claims 1 to 4, characterized in that, The feature points are the pixel points intersecting with the ground region and closest to the bottom edge of each frame of the image information.

6. An obstacle detection device characterized by comprising: The method comprises the following steps: an acquisition module is configured to acquire 2D image information in the driving direction of the current vehicle; The recognition module is configured to determine feature points in the 2D image information; the feature points are pixel points intersecting with a ground region in each frame of the image information and are dynamically determined by a machine learning model; The processing module is configured to determine a first detection contour in the 2D image information according to the feature points; the first detection contour is a contour representing an obstacle; The detection module is configured to determine a second detection contour according to the first detection contour and display the second detection contour on the 2D image information; the second detection contour is a contour obtained by screening and modifying the first detection contour; The processing module is specifically configured to: smoothly connect all the determined feature points to determine the first detection contour in the 2D image information; wherein two feature points with a distance greater than a preset distance are not smoothly connected; In the process of determining the second detection contour according to the first detection contour, the processing module is specifically configured to: compare the first detection contour with two end points with each standard contour in a pre-stored standard contour set to determine a standard contour with the maximum similarity; determine a scaling ratio between the first detection contour with two end points and the standard contour with the maximum similarity, and determine a defective contour corresponding to the first detection contour with two end points according to the scaling ratio; integrate the defective contour and the first detection contour with two end points to obtain the second detection contour.

7. An electronic device comprising a processor and a memory having a computer program stored therein, characterized in that The processor executes the computer program to implement the steps of the obstacle detection method in any one of claims 1 to 5.

8. A processor-readable storage medium, comprising: The processor readable storage medium stores a computer program, and the computer program is used to make the processor execute the steps of the obstacle detection method in any one of claims 1 to 5.

Images