Method and device for detecting boundary points of an object

By obtaining a panoramic image in an intelligent car and performing radial scanning from the center point of its feature map to the edge, it determines that the feature point with the largest feature value is the target boundary point, which solves the accuracy of the boundary point of the vehicle's travelable area in the prior art. and the problem of low efficiency, fast and low-cost boundary point detection is achieved.

CN111382696BActive Publication Date: 2024-11-12BEIJING BAIDU NETCOM SCI & TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010153383.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-03-06
Publication Date
2024-11-12
Estimated Expiration
2040-03-06

AI Technical Summary

Technical Problem

The technology used in existing smart cars to detect the vehicle's travelable areas is difficult to accurately identify target boundary points, resulting in slow detection speed, high hardware computing capabilities and high cost.

Method used

By acquiring a panoramic image, input the object detection neural network to extract the feature map of the feature point, and perform radial scanning from the center point of the feature map to the edge, determine the feature point with the largest feature value as the target boundary point, combining downsampling and upsampling techniques , improve detection efficiency and accuracy.

Benefits of technology

It effectively reduces the computing time of detecting target boundary points, reduces the requirements of hardware computing capabilities, reduces hardware costs, and improves detection speed and accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN111382696B_ABST
    Figure CN111382696B_ABST
Patent Text Reader

Abstract

The disclosed embodiments disclose a method and device for detecting boundary points of a target, and relate to the field of autonomous driving technology, especially autonomous parking technology. The method includes: obtaining a panoramic image to be processed; inputting the panoramic image to be processed into a target detection neural network to obtain a feature map output by the target detection neural network containing feature points of the target; performing radial scanning from the center point of the feature map to the edge, and determining the feature point with the largest feature value during each radial scan as the feature point of the target boundary. This method can effectively reduce the calculation time of detecting the feature points of the target boundary, speed up the detection speed, reduce the requirements for the computing power of the hardware, and reduce the cost of the hardware.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, specifically to the field of image recognition technology, and more particularly to a method and device for detecting boundary points of a target. Background Art

[0002] In the existing smart car scenarios, in order to detect the vehicle's drivable area, four fisheye cameras are installed on the vehicle, including a forward fisheye camera, a side fisheye camera on each side, and a rearward fisheye camera.

[0003] When the vehicle is moving, the four fisheye cameras transmit the acquired video data to the field programmable gate array (FPGA). Then, in the FPGA, four offline trained models (one corresponding to each of the four cameras) are used to infer the image data of the four cameras respectively, and finally four results are returned to the downstream for decision-making. Summary of the invention

[0004] The embodiments of the present disclosure provide a method and an apparatus for detecting boundary points of a target.

[0005] In a first aspect, an embodiment of the present disclosure provides a method for detecting boundary points of a target, comprising: acquiring a panoramic image to be processed; inputting the panoramic image to be processed into a target detection neural network to obtain a feature map output by the target detection neural network containing feature points of the target; performing radial scanning from the center point of the feature map to the edge, and determining the feature point with the largest feature value in each radial scan as the feature point of the boundary of the target.

[0006] In some embodiments, obtaining the panoramic image to be processed includes: obtaining multiple fisheye images captured at the same time; stitching the multiple fisheye images captured at the same time to obtain a stitched panoramic image; downsampling the stitched panoramic image to obtain the panoramic image to be processed.

[0007] In some embodiments, the method further includes: up-sampling feature points of a boundary of the target to obtain the boundary of the target.

[0008] In some embodiments, radial scanning is performed from the center point of the feature map to the edge, and the feature point with the largest eigenvalue in each radial scan is determined as the feature point of the boundary of the target, including: dividing the feature map at equal angles along the ring direction around the center point of the feature map to obtain multiple areas; and radial scanning is performed on each area at the same time, and the feature point with the largest eigenvalue in each radial scan is determined as the feature point of the boundary of the target.

[0009] In some embodiments, the target includes a drivable area.

[0010] In a second aspect, an embodiment of the present disclosure provides a device for detecting boundary points of a target, comprising: an image acquisition unit, configured to acquire a panoramic image to be processed; a feature map output unit, configured to input the panoramic image to be processed into a target detection neural network, and obtain a feature map output by the target detection neural network containing feature points of the target; a radial scanning unit, configured to perform radial scanning from the center point of the feature map to the edge, and determine the feature point with the largest feature value in each radial scan as the feature point of the boundary of the target.

[0011] In some embodiments, the image acquisition unit includes: an image acquisition subunit, configured to acquire multiple fisheye images captured at the same time; an image stitching subunit, configured to stitch the multiple fisheye images captured at the same time to obtain a stitched panoramic image; and a downsampling subunit, configured to downsample the stitched panoramic image to obtain a panoramic image to be processed.

[0012] In some embodiments, the device further includes: an upsampling unit configured to upsample feature points of a boundary of the object to obtain the boundary of the object.

[0013] In some embodiments, the radial scanning unit includes: a region division subunit, configured to divide the feature map at equal angles along a ring direction around the center point of the feature map to obtain multiple regions; a region scanning subunit, configured to simultaneously perform radial scanning on each region, and determine the feature point with the largest feature value in each radial scan as the feature point of the target boundary.

[0014] In some embodiments, the targets in the feature map output unit and the radial scanning unit include a drivable area.

[0015] In a third aspect, an embodiment of the present disclosure provides an electronic device / terminal / server, comprising: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by one or more processors, the one or more processors implement any of the methods for detecting boundary points of a target as described above.

[0016] In a fourth aspect, an embodiment of the present disclosure provides a computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements any of the above methods for detecting boundary points of a target.

[0017] The method and device for detecting the boundary points of a target provided by the embodiment of the present disclosure first obtain a panoramic image to be processed; then, input the panoramic image to be processed into a target detection neural network to obtain a feature map containing the feature points of the target output by the target detection neural network; then, radial scanning is performed from the center point of the feature map to the edge, and the feature point with the largest feature value during each radial scanning is determined as the feature point of the boundary of the target. In this process, the existing target detection neural network can be used to extract the feature map containing the feature points of the target from the panoramic image to be processed to improve the efficiency of generating the feature map containing the feature points of the target. Then, for the feature map containing the feature points of the target, since the row or column of the feature map includes more than one boundary point of the target, the boundary point of the target cannot be accurately identified when the row or column scanning is used, so radial scanning can be performed from the center point of the feature map to the edge, and the feature point with the largest feature value during each radial scanning is determined as the feature point of the boundary of the target, thereby effectively reducing the calculation time of detecting the feature points of the boundary of the target, accelerating the detection speed, reducing the requirements for the computing power of the hardware, and reducing the cost of the hardware. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Other features, objects and advantages of the present disclosure will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0019] Figure 1 is an exemplary system architecture diagram in which the present disclosure may be applied;

[0020] Figure 2 is a flowchart of an embodiment of a method for detecting boundary points of a target according to an embodiment of the present disclosure;

[0021] Figure 3 This is an exemplary application scenario of the method for detecting boundary points of a target according to an embodiment of the present disclosure;

[0022] Figure 4 is a flowchart of another embodiment of a method for detecting boundary points of a target according to an embodiment of the present disclosure;

[0023] Figure 5 is an exemplary structural diagram of an embodiment of a device for detecting boundary points of a target disclosed in the present invention;

[0024] Figure 6 It is a schematic diagram of the structure of a computer system suitable for implementing a server of an embodiment of the present disclosure. DETAILED DESCRIPTION

[0025] The present disclosure is further described in detail below in conjunction with the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings.

[0026] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present disclosure may be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0027] Figure 1 An exemplary system architecture 100 is shown to which an embodiment of a method for detecting boundary points of an object or an apparatus for detecting boundary points of an object of the present disclosure can be applied.

[0028] like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0029] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various applications can be installed on terminal devices 101, 102, 103, such as image processing applications, driving reminder applications, driving control applications, etc.

[0030] Terminal devices 101, 102, 103 can be hardware or software. When terminal devices 101, 102, 103 are hardware, they can be various electronic devices that support browser applications, including but not limited to car brains, smart terminals, tablet computers, laptop portable computers, desktop computers, etc. When terminal devices 101, 102, 103 are software, they can be installed in the electronic devices listed above. It can be implemented as multiple software or software modules for providing distributed services, or it can be implemented as a single software or software module. No specific limitation is made here.

[0031] The server 105 may be a server that provides various services, such as a background server that provides support for the terminal devices 101, 102, and 103. The background server may analyze and process the received request and other data, and feed back the processing results to the terminal device.

[0032] It should be noted that the server can be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or it can be implemented as a single server. When the server is software, it can be implemented as multiple software or software modules for providing distributed services, or it can be implemented as a single software or software module. No specific limitation is made here.

[0033] In practice, the method for detecting boundary points of a target provided by the embodiments of the present disclosure may be executed by the terminal devices 101, 102, 103 and / or the servers 105, 106, and the apparatus for detecting boundary points of a target may also be disposed in the terminal devices 101, 102, 103 and / or the servers 105, 106.

[0034] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0035] Continue to refer Figure 2 , Figure 2 A process 200 of an embodiment of a method for detecting boundary points of a target according to the present disclosure is shown. The method for detecting boundary points of a target comprises the following steps:

[0036] Step 201: Obtain a panoramic image to be processed.

[0037] In this embodiment, the execution subject (eg Figure 1 The terminal or server shown in the figure can obtain the panoramic image to be processed from a local or remote album or database, or can obtain the panoramic image to be processed via a local or remote image processing service.

[0038] The panoramic image to be processed here can be stitched together by using multiple fisheye images collected at the same time, and the stitched panoramic image can be used as the panoramic image to be processed; after the stitched panoramic image is obtained by using multiple fisheye images collected at the same time, the stitched panoramic image can be further processed (such as data cleaning, data enhancement, etc.) to obtain the panoramic image to be processed.

[0039] In some optional implementations of this embodiment, obtaining the panoramic image to be processed may include: obtaining multiple fisheye images captured at the same time; stitching the multiple fisheye images captured at the same time to obtain a stitched panoramic image; downsampling the stitched panoramic image to obtain the panoramic image to be processed.

[0040] In this implementation, after a stitched panoramic image is obtained based on multiple fisheye images captured at the same time, the stitched panoramic image is downsampled to obtain a panoramic image to be processed, which can reduce the amount of data that needs to be processed when performing subsequent data processing on the panoramic image to be processed.

[0041] Downsampling means downsampling an image I of size M×N by s (s>1) times, that is, obtaining a resolution image of size (M / s)×(N / s), where s is the common divisor of M and N. If the panoramic image to be downsampled is an image in matrix form, the image in the s×s window of the original image is converted into a pixel, and the value of this pixel is the mean of all pixels in the window. Downsampling can make the image size meet the needs of data processing.

[0042] Step 202: input the panoramic image to be processed into a target detection neural network to obtain a feature map output by the target detection neural network containing feature points of the target.

[0043] In this embodiment, the target detection neural network is a trained machine learning network that takes the panoramic image to be processed as input and outputs a feature map containing the feature points of the target. It can be a neural network used in existing technologies or future technologies to detect and mark the feature points of the target in the feature map, and this application does not limit this.

[0044] For example, the target detection neural network can be implemented based on R-CNN, spatial pyramid pooling network SPP, FAST-RCNN, FASTER-RCNN, etc.; when higher real-time performance is required, the target detection neural network can be implemented based on the target detection algorithm YOLO series, YOLO+SSD, support vector machine SVM or deep convolutional neural network DeepLab series, etc.

[0045] The target detection neural network can be obtained by taking the panoramic image samples as input, taking the feature map containing the feature points of the target corresponding to the panoramic image samples as the expected output, and training the initial network of the target detection neural network.

[0046] The target included in the feature map is a target surrounding the center point of the panorama identified from the panorama according to the needs of the application scenario. For example, a specific entity in the panorama. In a specific example, the target included in the feature map can be: in the scenario of autonomous driving, a drivable area identified from the panorama to be processed.

[0047] The feature map output by the target detection neural network contains the feature points of the target and outputs the maximum response value at the boundary point of the target, that is, the boundary point of the target has the largest feature value.

[0048] Step 203, radial scanning is performed from the center point of the feature map to the edge, and the feature point with the largest feature value in each radial scanning is determined as the feature point of the boundary of the target.

[0049] In this embodiment, for the feature map containing the feature points of the target obtained in step 202, the execution subject performs radial scanning from the center point of the feature map to the edge direction of the feature map. The radial scanning starts from the center point of the feature map, and scans radially to the edge of the feature map, and then returns to the center point of the feature map, and another radial line is generated adjacent to the previous radial line. The newly generated radial line is radially scanned to the edge of the feature map, and then returns to the center point of the feature map, so that a radial line appears after the previous radial line, as if each radial line is rotating around the center point of the feature map.

[0050] When performing radial scanning, the execution subject may determine a feature point with the largest feature value during each radial scanning, and determine the feature point with the largest feature value as a feature point of the boundary of the target.

[0051] In some optional implementations of this embodiment, the above method may further include: up-sampling feature points of the boundary of the target to obtain the boundary of the target.

[0052] In this implementation, corresponding to the above-mentioned downsampling of the stitched panoramic image to obtain the panoramic image to be processed, after obtaining the feature map containing the feature points of the boundary of the target, the feature points of the boundary of the target can be upsampled to obtain the boundary of the restored target. The upsampling here can be an upsampling method in the existing technology or future technology, such as bilinear interpolation, transposed convolution, upsampling (unsampling) and uppooling (unpooling), etc., which will not be repeated here.

[0053] In this implementation, by upsampling the feature points of the boundary of the target, the clarity of the boundary of the target can be improved.

[0054] The method for detecting the boundary points of a target in the above-mentioned embodiment of the present disclosure may adopt an existing target detection neural network to extract a feature map containing the feature points of the target from the panoramic image to be processed, so as to improve the efficiency of generating the feature map containing the feature points of the target. Afterwards, for the feature map containing the feature points of the target, since the row or column of the feature map includes more than one boundary points of the target, the boundary points of the target cannot be accurately identified when scanning the row or column. Therefore, a radial scan may be performed from the center point of the feature map to the edge, and the feature point with the largest feature value in each radial scan is determined as the feature point of the target boundary, thereby effectively reducing the calculation time of detecting the feature points of the target boundary, accelerating the detection speed, reducing the requirements on the computing power of the hardware, and reducing the cost of the hardware.

[0055] The following combination Figure 3 , describing an exemplary application scenario of the method for detecting boundary points of an object disclosed in the present invention.

[0056] like Figure 3 As shown, Figure 3 An exemplary application scenario of the method for detecting boundary points of an object according to the present disclosure is shown.

[0057] like Figure 3 As shown, the method 300 for detecting boundary points of a target is run in an electronic device 310 and may include:

[0058] First, a panoramic image to be processed is obtained 301;

[0059] Secondly, the panoramic image 301 to be processed is input into the target detection neural network 302, and a feature map 303 containing feature points of the target outputted by the target detection neural network is obtained;

[0060] Finally, radial scanning is performed from the center point 304 of the feature map to the edge 305 of the feature map, and the feature point 306 with the largest feature value in each radial scanning is determined as the feature point 307 of the boundary of the target.

[0061] It should be understood that the above Figure 3 The application scenario of the method for detecting boundary points of an object shown in the figure is only an exemplary description of the method for detecting boundary points of an object and does not represent a limitation of the method. Figure 3 The various steps shown in the above can be further implemented in more detailed ways. Figure 3 On this basis, further steps are added to generate images.

[0062] Further references Figure 4 , Figure 4 A schematic flow chart of another embodiment of a method for detecting boundary points of a target according to the present disclosure is shown.

[0063] like Figure 4 As shown, the method 400 for detecting boundary points of a target in this embodiment may include the following steps:

[0064] Step 401: Obtain a panoramic image to be processed.

[0065] In this embodiment, in this embodiment, the execution subject of the method for detecting the boundary points of the target (for example Figure 1 The terminal or server shown in the figure can obtain the panoramic image to be processed from a local or remote album or database, or can obtain the panoramic image to be processed via a local or remote image processing service.

[0066] The panoramic image to be processed here can be stitched together by using multiple fisheye images collected at the same time, and the stitched panoramic image can be used as the panoramic image to be processed; after the stitched panoramic image is obtained by using multiple fisheye images collected at the same time, the stitched panoramic image can be further processed (such as data cleaning, data enhancement, etc.) to obtain the panoramic image to be processed.

[0067] Step 402: input the panoramic image to be processed into a target detection neural network to obtain a feature map output by the target detection neural network containing feature points of the target.

[0068] In this embodiment, in this embodiment, the target detection neural network is a trained machine learning network that takes the panoramic image to be processed as input and takes a feature map containing the feature points of the target as output. It can be a neural network used in existing technology or future technology to detect and mark the feature points of the target in the feature map, and this application is not limited to this.

[0069] For example, the target detection neural network can be implemented based on R-CNN, spatial pyramid pooling network SPP, FAST-RCNN, FASTER-RCNN, etc.; when higher real-time performance is required, the target detection neural network can be implemented based on the target detection algorithm YOLO series, YOLO+SSD, support vector machine SVM or deep convolutional neural network DeepLab series, etc.

[0070] The feature map output by the target detection neural network contains the feature points of the target and outputs the maximum response value at the boundary point of the target, that is, the boundary point of the target has the largest feature value.

[0071] Those skilled in the art will appreciate that the above steps 401 to 402 are similar to Figure 2 The steps 201 to 202 in the embodiment shown in FIG. Figure 2 The operations and features described in steps 201 to 202 of the embodiment shown in FIG. 1 are also applicable to steps 401 to 402 , and will not be described in detail here.

[0072] Step 403, dividing the feature map into equal angles along a ring direction around the center point of the feature map to obtain multiple regions.

[0073] In this embodiment, the execution entity may divide 360 ​​degrees into multiple angles along a circumferential direction around the center point of the feature graph, thereby dividing the feature graph into multiple regions at equal angles.

[0074] Step 404 , radial scanning is performed on each area simultaneously, and the feature point with the largest feature value during each radial scanning is determined as the feature point of the boundary of the target.

[0075] In this embodiment, the execution subject performs radial scanning from the center point of the feature map to the edge direction of the feature map for each of the multiple divided regions. The radial scanning starts from the center point of the feature map and scans radially to the edge of the feature map, then returns to the center point of the feature map, and generates another radial line adjacent to the previous radial line. The newly generated radial line scans radially to the edge of the feature map, and then returns to the center point of the feature map, so that a radial line appears after the previous radial line, as if each radial line is rotating around the center point of the feature map.

[0076] When performing radial scanning, the execution subject may determine a feature point with the largest feature value during each radial scanning, and determine the feature point with the largest feature value as a feature point of the boundary of the target.

[0077] The method for detecting the boundary points of a target in the above embodiment of the present disclosure is Figure 2 Compared with the embodiment shown, the feature map can be divided into equal angles along the ring direction around the center point of the feature map to obtain multiple areas, and each of the multiple areas can be radially scanned from the center point of the feature map to the edge at the same time, and the feature point with the largest eigenvalue in each radial scan is determined as the feature point of the boundary of the target, thereby further effectively reducing the calculation time of detecting the feature points of the boundary of the target, accelerating the detection speed, reducing the requirements on the computing power of the hardware, and reducing the cost of the hardware.

[0078] As an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of a device for detecting boundary points of a target. Figure 2-Figure 4 Corresponding to the method embodiment shown in Figure 1 terminal or server as shown.

[0079] like Figure 5 As shown, the device 500 for detecting boundary points of a target in this embodiment may include: an image acquisition unit 501, configured to acquire a panoramic image to be processed; a feature map output unit 502, configured to input the panoramic image to be processed into a target detection neural network, and obtain a feature map output by the target detection neural network containing feature points of the target; a radial scanning unit 503, configured to perform radial scanning from the center point of the feature map to the edge, and determine the feature point with the largest feature value in each radial scanning as the feature point of the boundary of the target.

[0080] In some optional implementations of the present embodiment, the image acquisition unit 501 includes (not shown in the figure): an image acquisition subunit, configured to acquire multiple fisheye images captured at the same time; an image stitching subunit, configured to stitch the multiple fisheye images captured at the same time to obtain a stitched panoramic image; and a downsampling subunit, configured to downsample the stitched panoramic image to obtain a panoramic image to be processed.

[0081] In some optional implementations of this embodiment, the apparatus further includes: an up-sampling unit 504 configured to up-sample feature points of a boundary of the target to obtain the boundary of the target.

[0082] In some optional implementations of the present embodiment, the radial scanning unit 503 includes (not shown in the figure): a region division subunit, configured to divide the feature map at equal angles along a ring direction around the center point of the feature map to obtain multiple regions; a region scanning subunit, configured to radially scan each region at the same time, and determine the feature point with the largest feature value in each radial scan as the feature point of the target boundary.

[0083] In some optional implementations of this embodiment, the targets in the feature map output unit 502 and the radial scanning unit 503 include a drivable area.

[0084] It should be understood that the various units described in the apparatus 500 are similar to those described in the reference Figure 2-Figure 4 Therefore, the operations and features described above for the method are also applicable to the device 500 and the various units contained therein, and will not be repeated here.

[0085] Reference below Figure 6 , which shows an electronic device (eg, Figure 1 The terminal device in the embodiment of the present disclosure may include but is not limited to a notebook computer, a desktop computer, etc. Figure 6 The terminal device / server shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0086] like Figure 6As shown, the electronic device 600 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device 600 are also stored. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0087] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 6 The electronic device 600 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 6 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0088] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through a communication device 609, or installed from a storage device 608, or installed from a ROM 602. When the computer program is executed by the processing device 601, the above functions defined in the method of the embodiment of the present disclosure are executed. It should be noted that the computer-readable medium described in the embodiment of the present disclosure can be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection with one or more conductors, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In an embodiment of the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, an apparatus, or a device. In an embodiment of the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in combination with an instruction execution system, an apparatus, or a device. The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to: wire, optical cable, RF (radio frequency), etc., or any suitable combination of the foregoing.

[0089] The computer-readable medium may be included in the electronic device; or it may exist independently without being installed in the electronic device. The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device: obtains a panoramic image to be processed; inputs the panoramic image to be processed into a target detection neural network to obtain a feature map output by the target detection neural network containing feature points of the target; radially scans from the center point of the feature map to the edge, and determines the feature point with the largest feature value during each radial scan as the feature point of the target boundary.

[0090] Computer program code for performing the operations of embodiments of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on a user's computer, partially on a user's computer, as a separate software package, partially on a user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0091] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0092] The units involved in the embodiments described in the present disclosure may be implemented by software or by hardware. The units described may also be provided in a processor, for example, may be described as: a processor including an image acquisition unit, a feature map output unit, and a radial scanning unit. The names of these units do not, in some cases, constitute limitations on the units themselves, for example, the image acquisition unit may also be described as a "unit for acquiring a panoramic image to be processed".

[0093] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present disclosure is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above invention concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other.

Claims

1. A method for detecting boundary points of an object, used in an autonomous driving scenario, include: Acquiring a panoramic image to be processed, including: stitching a plurality of fisheye images acquired at the same time, and using the stitched panoramic image as the panoramic image to be processed; Inputting the panoramic image to be processed into a target detection neural network to obtain a feature map output by the target detection neural network containing feature points of the target, wherein the target is a drivable area; The feature map is divided into multiple regions at equal angles along a ring direction around the center point of the feature map. At the same time, each region is radially scanned from the center point of the feature map to the edge and then back to the center point of the feature map to generate a radial line, and then another radial line is generated adjacent to the previous radial line. The scanning is repeated in this way, and the feature point with the largest eigenvalue in each radial scan is determined as the feature point of the target boundary.

2. The method according to claim 1, in, The obtaining of the panoramic image to be processed comprises: Get multiple fisheye images collected at the same time; Stitching the multiple fisheye images collected at the same time to obtain a stitched panoramic image; The stitched panoramic image is downsampled to obtain a panoramic image to be processed.

3. The method according to claim 2, in, The method further comprises: Up-sampling is performed on feature points of the boundary of the object to obtain the boundary of the object.

4. A device for detecting boundary points of a target, used in an autonomous driving scenario, include: The image acquisition unit is configured to acquire the panoramic image to be processed, including: stitching a plurality of fisheye images acquired at the same time, and using the stitched panoramic image as the panoramic image to be processed; A feature map output unit is configured to input the panoramic image to be processed into a target detection neural network to obtain a feature map output by the target detection neural network containing feature points of the target, wherein the target is a drivable area; The radial scanning unit is configured to divide the feature map at equal angles along a ring direction around the center point of the feature map to obtain multiple areas; at the same time, each area is radially scanned from the center point of the feature map to the edge and then back to the center point of the feature map to generate a radial line, and then another radial line is generated adjacent to the previous radial line, and the scanning is repeated in this way, and the feature point with the largest feature value in each radial scan is determined as the feature point of the boundary of the target.

5. The device according to claim 4, in, The image acquisition unit comprises: An image acquisition subunit is configured to acquire a plurality of fisheye images collected at the same time; An image stitching subunit is configured to stitch the multiple fisheye images acquired at the same time to obtain a stitched panoramic image; The down-sampling sub-unit is configured to down-sample the stitched panoramic image to obtain a panoramic image to be processed.

6. The device according to claim 5, in, The device also includes: The up-sampling unit is configured to up-sample the feature points of the boundary of the object to obtain the boundary of the object.

7. An electronic device / terminal / server, include: one or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 3.

8. A computer readable medium having a computer program stored thereon, wherein when the program is executed by a processor, the method according to any one of claims 1 to 3 is implemented.