Target detection method and device and storage medium

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a target detection method and device and a storage medium. Comprising the steps of obtaining a current BEV feature corresponding to a target area and at least one historical BEV feature; the current BEV feature is extracted and generated by at least one original image, and the original image is an image corresponding to the target area; fusing the at least one historical BEV feature into the current BEV feature to obtain a fused BEV feature; and performing target detection on the fused BEV features to obtain a detection result. The embodiment of the invention is used for solving the problem of low accuracy of a detection result.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a target detection method, device and storage medium. Background Art

[0002] Image object detection technology has a wide range of applications in the field of computer vision. For example, in self-driving cars, it can be used to detect vehicles, pedestrians, traffic signs, etc. on the road, thereby achieving environmental perception and safe driving; in industrial production, it can be used to detect product defects, incorrect assembly, and other production problems to ensure product quality.

[0003] The current method of image target detection usually involves first collecting the original image and extracting features from it. After that, the extracted features are input into a pre-trained detection model to determine the detection target in the image. The features extracted by this target detection method are usually relatively simple, and the detection results are less accurate. Summary of the invention

[0004] In order to solve the above technical problems, the present application provides a target detection method, device and storage medium, which can improve the accuracy of detection results.

[0005] In a first aspect, the present application provides a target detection method, comprising: obtaining a current BEV feature corresponding to a target area, and at least one historical BEV feature; the current BEV feature is generated by extracting at least one original image, where the original image is an image corresponding to the target area; integrating at least one historical BEV feature into the current BEV feature to obtain a fused BEV feature; performing target detection on the fused BEV feature to obtain a detection result.

[0006] In a second aspect, the present application provides a target detection device, comprising:

[0007] An acquisition module is used to acquire a current BEV feature corresponding to a target area and at least one historical BEV feature; the current BEV feature is generated by extracting at least one original image, and the original image is an image corresponding to the target area; a processing module is used to integrate at least one historical BEV feature into the current BEV feature to obtain a fused BEV feature; a detection module is used to perform target detection on the fused BEV feature to obtain a detection result.

[0008] In a third aspect, the present application provides an electronic device, comprising: a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the target detection method of the first aspect is implemented.

[0009] In a fourth aspect, the present application provides a computer-readable storage medium, comprising: a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the target detection method of the first aspect is implemented.

[0010] In a fifth aspect, the present application provides a computer program product, comprising: when the computer program product runs on a computer, the computer implements the target detection method as in the first aspect.

[0011] The technical solution provided by the present application has the following advantages compared with the prior art: first, the current BEV feature corresponding to the target area and at least one historical BEV feature are obtained, and at least one historical BEV feature is integrated into the current BEV feature to obtain a fused BEV feature. After that, the fused BEV feature is subjected to target detection to obtain a detection result. In this way, the historical BEV feature can be integrated into the current BEV feature to obtain a fused BEV feature, and then the fused BEV feature is detected to obtain the detection result, that is, instead of using a single current BEV feature to detect the target, the historical BEV feature is integrated into the current BEV feature before detecting the target, which increases the BEV features that can be detected, thereby improving the accuracy of the detection result. In addition, the current BEV feature is generated by extracting at least one original image, and the original image is the image corresponding to the target area, that is, instead of using the features corresponding to a single picture to detect the target, the features of multiple original images corresponding to the target area are extracted to detect the target, which further improves the accuracy of the detection result. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0013] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0014] Figure 1 One of the scenario schematic diagrams of the target detection method is provided for the embodiment of the present application;

[0015] Figure 2 A second schematic diagram of a scene for providing a target detection method in an embodiment of the present application;

[0016] Figure 3 One of the flowcharts of the target detection method provided in the embodiment of the present application;

[0017] Figure 4The second flowchart of the target detection method provided in the embodiment of the present application;

[0018] Figure 5 The third flowchart of the target detection method provided in the embodiment of the present application;

[0019] Figure 6 A fourth flowchart of the target detection method provided in an embodiment of the present application;

[0020] Figure 7 A schematic diagram of the structure of a target detection device provided in an embodiment of the present application;

[0021] Figure 8 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to more clearly understand the above-mentioned purposes, features and advantages of the present application, the scheme of the present application will be further described below. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0023] In the following description, many specific details are set forth to facilitate a full understanding of the present application, but the present application may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only part of the embodiments of the present application, rather than all of the embodiments.

[0024] Figure 1 A schematic diagram of an application scenario of the target detection method provided in an embodiment of the present application includes a target area and a vehicle 10. The vehicle 10 is located in the target area, and an image acquisition device capable of acquiring images of the target area is installed on the periphery of the vehicle 10, for acquiring at least one original image corresponding to the target area in real time. The number of image acquisition devices is preset. For example, an image acquisition device is provided at each of A, B, C, and D around the vehicle 10, and four original images corresponding to the target area can be acquired in real time at each moment. Afterwards, the original image is sent to the target detection device, and the target detection method is executed by the target detection device.

[0025] Among them, the target detection device provided in the embodiment of the present application can be located inside the vehicle 10 to assist in driving the vehicle 10, or it can be located outside the vehicle 10 to provide support for other software or hardware that needs to make decisions based on the detection results. In addition, the target detection device provided in the embodiment of the present application can be hardware or software. When the target detection device is hardware, it can be various electronic devices with a function of running target detection, including but not limited to vehicle-mounted equipment, smart vehicles, mobile phones, computers, etc. When the target detection device is software, it can be installed in the electronic devices listed above. It can be implemented as multiple software or software modules, or it can be implemented as a single software or software module, which is not specifically limited here.

[0026] and Figure 1 Different from the illustrated scenario, the target detection method provided by the present application can also be applied to other scenarios, in which case the image acquisition device can be installed on other equipment or facilities besides the vehicle. For example, Figure 2 Another application scenario schematic diagram of the target detection method provided in the embodiment of the present application includes a target area, an image acquisition device 11, and a target detection device 12. Among them, the image acquisition device 11 is set on a device or facility at the edge of the target area, and the image acquisition device 11 is set toward the inside of the target area, and is used to obtain at least one original image corresponding to the target area in real time. The number of image acquisition devices 11 is preset. For example, the number of image acquisition devices 11 can be set according to the size of the target area, or it can be a value set by relevant personnel according to actual conditions. For another example, the number of image acquisition devices 11 is 5.

[0027] Among them, the target detection device provided in the embodiment of the present application can be inside the target area, or it can be located outside the target area, and the present application does not limit this. In addition, the target detection device provided in the embodiment of the present application can be hardware or software. When the target detection device is hardware, it can be various electronic devices with a function of running target detection, including but not limited to vehicle-mounted equipment, smart vehicles, mobile phones, computers, etc. When the target detection device is software, it can be installed in the electronic devices listed above. It can be implemented as multiple software or software modules, or it can be implemented as a single software or software module, which is not specifically limited here.

[0028] For example, when the video processing method provided in the embodiment of the present application is executed by a target detection device, the process is as follows: the target detection device obtains the current BEV feature corresponding to the target area and at least one historical BEV feature, and integrates at least one historical BEV feature into the current BEV feature to obtain a fused BEV feature. Afterwards, the target detection device performs target detection on the fused BEV feature to obtain a detection result. In this way, the historical BEV feature can be integrated into the current BEV feature before detecting the target, which increases the BEV features that can be detected, thereby improving the accuracy of the detection result.

[0029] Figure 3 A schematic diagram of a target detection method provided in an embodiment of the present application is shown in FIG. Figure 3 As shown, the target detection method may include the following steps.

[0030] S21. Obtain current BEV characteristics corresponding to the target area and at least one historical BEV characteristic.

[0031] The current bird's eye view (BEV) feature is generated by extracting at least one original image, where the original image is an image corresponding to the target area. The target area is a specific area, for example, Figure 1 or Figure 2 The target area in the target area. The historical BEV features can be obtained from the historical time period. The time length of the historical time period is preset, and the next moment after the end moment of the historical time period is the current time. For example, when the time length of the historical time period is 5s and the current time is a hour, b minute, and c second, the historical time period is a hour, b minute, and c-5 seconds to a hour, b minute, and c-1 second.

[0032] First, the current BEV features corresponding to the target area are obtained.

[0033] Specifically, the method for obtaining the current BEV features corresponding to the target area can be to use a specific image acquisition device to acquire the BEV view of the target area in real time, and extract the current BEV features through the BEV view. The specific image acquisition device is an image acquisition device capable of acquiring the BEV view, for example, it can be a camera installed on a drone or a high-rise building, or a laser radar device, a satellite, etc. The method for extracting the current BEV features from the BEV view can be to use a related feature extraction model to extract the BEV features, for example, a pre-trained convolutional neural network (CNN) model, a squeeze-and-excitation network (SENet) model, etc.

[0034] In some embodiments, Figure 4As shown, the method of obtaining the current BEV characteristics corresponding to the target area may also include the following steps.

[0035] S211. Acquire at least one original image corresponding to the target area in real time.

[0036] Each of the at least one original image is acquired by a different image acquisition device.

[0037] Specifically, the method of acquiring at least one original image corresponding to the target area in real time may be as follows: Figure 1 In the scenario shown, an image acquisition device disposed outside the vehicle 10 is used to acquire at least one original image corresponding to the target area in real time; Figure 2 In the scenario shown, an image acquisition device 11 disposed at the edge of the target area and facing the inside of the target area is used to acquire at least one original image corresponding to the target area in real time.

[0038] S212: Extract features from at least one original image to generate current BEV features.

[0039] Specifically, at least one original image is first passed through a convolutional neural network to obtain primary features corresponding to each original image, and then the primary features corresponding to each original image are subjected to BEV conversion processing to obtain current BEV features.

[0040] Among them, the convolutional neural network is pre-trained and used to extract the primary features of the original image.

[0041] The method of performing BEV conversion processing on the primary features corresponding to each original image may be to use a Transformer algorithm to perform BEV conversion processing on the primary features corresponding to each original image to obtain current BEV features.

[0042] In the above scheme, at least one original image corresponding to the target area is acquired in real time, and features are extracted from at least one original image to generate the current BEV features. It is possible to extract features from multiple original images corresponding to the target area to generate the current BEV features. Compared with the features extracted from one picture, it has more features corresponding to the target area, so that it has higher accuracy when detecting the target area.

[0043] Secondly, at least one historical BEV feature within the historical time period is obtained.

[0044] Specifically, the method for obtaining at least one historical BEV feature within the historical time period may be to obtain historical original images collected within the historical time period, and perform feature extraction on the historical original images to obtain at least one historical BEV feature within the historical time period; or it may be to directly obtain the BEV features corresponding to the historical time period from the storage space to obtain at least one historical BEV feature within the historical time period; or it may be to obtain the fused BEV features corresponding to the historical time period from the storage space to obtain at least one historical BEV feature within the historical time period.

[0045] S22: Integrate at least one historical BEV feature into the current BEV feature to obtain a fused BEV feature.

[0046] Specifically, in some embodiments, at least one historical BEV feature is integrated into the current BEV feature to obtain a fused BEV feature by performing a fusion operation on at least one historical BEV feature and the current BEV feature in sequence in chronological order to obtain the fused BEV feature.

[0047] The fusion operation includes the following steps A and B.

[0048] A. Integrate the first feature into the second BEV feature to obtain the fused second BEV feature.

[0049] Among them, the first feature is the first BEV feature or the fused first BEV feature; the first BEV feature is the BEV feature generated at the first moment, and the second BEV feature is the BEV feature generated at the second moment; the first moment is adjacent to the second moment, and the first moment is earlier than the second moment. For example, the current time is a hour, b minute, c second, and the historical time period is a hour, b minute, c-5 seconds to a hour, b minute, c-1 second. When the first moment is a hour, b minute, c-2 seconds, the second moment is a hour, b minute, c-1 second.

[0050] Specifically, Figure 5 As shown, the method of integrating the first feature into the second BEV feature to obtain the integrated second BEV feature may include the following steps.

[0051] S1. According to the second BEV feature, the first feature is position-transformed to obtain a conversion feature.

[0052] Among them, position transformation includes translation, rotation, scaling, shearing, etc.

[0053] Specifically, a feature point detection algorithm is first used to determine similar or identical feature points in the first feature and the second BEV feature. For example, the feature point detection algorithm may be a scale-invariant feature transform (SIFT) algorithm, a speeded-up robust features (SURF) algorithm, etc. Then, a conversion matrix between the first feature and the second BEV feature is calculated based on these similar or identical feature points, and finally, the position of each feature point in the first feature is adjusted according to the conversion matrix to obtain a conversion feature. Among them, the algorithm for calculating the conversion matrix may be a least squares method.

[0054] S2. Matrix concatenate the conversion feature with the current BEV feature to obtain the concatenated feature.

[0055] Specifically, the matrix corresponding to the conversion feature and the matrix corresponding to the current BEV feature can be concatenated to obtain the concatenated feature. The concatenation method can be horizontal concatenation, vertical concatenation, or block concatenation. In addition, the concatenation method can also be averaging the elements at the same position in the two matrices.

[0056] S3. Convolve the concatenated features with the target convolution kernel to obtain the convolution features.

[0057] Among them, the size, shape, and value of the target convolution kernel are all preset. For example, the target convolution kernel can be a 1×1 matrix with a value of 1.

[0058] S4. Input the convolutional features into the excitation network to obtain the fused second BEV features.

[0059] The excitation network is used to realize information interaction between features and output the fused second BEV features. For example, the excitation network can be a squeeze-and-excitation network (SENet).

[0060] In the above scheme, the first feature is transformed according to the second BEV feature to obtain a conversion feature, and the conversion feature is concatenated with the current BEV feature to obtain a concatenated feature. After that, the concatenated feature is convolved with the target convolution kernel to obtain a convolution feature, and the convolution feature is input into the excitation network to obtain a fused second BEV feature. The first feature can be integrated into the second BEV feature, so that the second BEV feature has more features corresponding to the target area, thereby having higher accuracy when detecting the target area.

[0061] B. When the second BEV feature is the current BEV feature, determining the fused second BEV feature as the fused BEV feature.

[0062] Specifically, when the second BEV feature is the current BEV feature, the fused second BEV feature is determined to be the fused BEV feature. When the second BEV feature is not the current BEV feature, the fused second BEV feature is integrated into the third BEV feature to obtain the fused third BEV feature, and it is determined whether it is the current BEV feature until the fused BEV feature is obtained.

[0063] The third BEV feature is a BEV feature generated at a third moment, the second moment is adjacent to the third moment, and the second moment is earlier than the third moment.

[0064] In the above scheme, the first feature is integrated into the second BEV feature to obtain the fused second BEV feature, and when the second BEV feature is the current BEV feature, the fused second BEV feature is determined to be the fused BEV feature. The BEV feature corresponding to the previous moment can be fused into the BEV feature corresponding to the next moment to obtain the fused BEV feature corresponding to the next moment, so that the BEV feature corresponding to the next moment has more features corresponding to the target area, thereby having higher accuracy when detecting the target area.

[0065] In some embodiments, at least one historical BEV feature may be integrated into the current BEV feature to obtain a fused BEV feature. Alternatively, a pre-trained neural network model may be used to integrate at least one historical BEV feature into the current BEV feature to obtain a fused BEV feature.

[0066] In some embodiments, after obtaining the fused BEV feature, the method further includes: storing at least one of the following: the current BEV feature, the fused BEV feature. Specifically, at least one of the current BEV feature and the fused BEV feature is stored in a storage space. Optionally, when storing the current BEV feature, the time corresponding to the current BEV feature is also stored. Similarly, when storing the fused BEV feature, the time corresponding to the fused BEV feature is also stored.

[0067] In the above scheme, after obtaining the fused BEV feature, at least one of the current BEV feature and the fused BEV feature is stored, so that when the BEV feature at a future moment is subsequently fused with the historical BEV feature, the historical BEV feature can be directly obtained, avoiding the need to extract the historical BEV feature from the historical original image and saving resources.

[0068] S23, performing target detection on the fused BEV features to obtain a detection result.

[0069] In some embodiments, Figure 6 As shown, the method of performing target detection on the fused BEV features and obtaining the detection results may include the following steps.

[0070] S231. Use an encoder to encode the fused BEV features to obtain an encoding result.

[0071] The encoder is a BEV space encoder, which is used to encode the features of the BEV space.

[0072] Specifically, the backbone network (backbone) + feature pyramid network (feature pyramid network, FPN) method can be used to encode the fused BEV features to obtain the encoding result.

[0073] S232: Use a target detection algorithm to detect the encoding result to obtain a detection result.

[0074] Specifically, the CenterPoint algorithm may be used to detect the encoding result to obtain a detection result.

[0075] In the above scheme, firstly, the current BEV feature corresponding to the target area and at least one historical BEV feature in the historical time period are obtained, and at least one historical BEV feature is integrated into the current BEV feature to obtain a fused BEV feature. After that, the fused BEV feature is subjected to target detection to obtain a detection result. In this way, the historical BEV features in the historical time period can be integrated into the current BEV feature to obtain a fused BEV feature, and then the fused BEV feature is detected to obtain the detection result, that is, instead of using a single current BEV feature to detect the target, the historical BEV feature is integrated into the current BEV feature before detecting the target, which increases the BEV features that can be detected, thereby improving the accuracy of the detection result. In addition, the current BEV feature is generated by extracting at least one original image, and the original image is the image corresponding to the target area, that is, instead of using the features corresponding to a single picture to detect the target, the features of multiple original images corresponding to the target area are extracted to detect the target, which further improves the accuracy of the detection result.

[0076] The embodiment of the present application can divide the functional modules of the target detection device according to the above method example. For example, each functional module can be divided according to each function, or two or more functions can be integrated into one processing unit. The above integrated module can be implemented in the form of hardware or in the form of software functional modules. It should be noted that the division of modules in the embodiment of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation.

[0077] like Figure 7 , which is a schematic diagram of the structure of a target detection device provided in an embodiment of the present application, wherein the target detection device includes an acquisition module 701 , a processing module 702 , and a detection module 703 .

[0078] The acquisition module 701 is used to acquire the current BEV feature corresponding to the target area and at least one historical BEV feature; the current BEV feature is generated by extracting at least one original image, and the original image is the image corresponding to the target area; the processing module 702 is used to integrate at least one historical BEV feature into the current BEV feature to obtain a fused BEV feature; the detection module 703 is used to perform target detection on the fused BEV feature to obtain a detection result.

[0079] In some embodiments, the processing module 702 is specifically used to: perform a fusion operation on at least one historical BEV feature and a current BEV feature in sequence in chronological order to obtain a fused BEV feature; wherein the fusion operation includes: integrating the first feature into the second BEV feature to obtain a fused second BEV feature; when the second BEV feature is the current BEV feature, determining that the fused second BEV feature is a fused BEV feature; the first feature is the first BEV feature or the fused first BEV feature; the first BEV feature is a BEV feature generated at a first moment, and the second BEV feature is a BEV feature generated at a second moment; the first moment is adjacent to the second moment, and the first moment is earlier than the second moment.

[0080] In some embodiments, the processing module 702 is specifically used to: transform the position of the first feature according to the second BEV feature to obtain a conversion feature; perform matrix splicing on the conversion feature and the current BEV feature to obtain a spliced ​​feature; perform a convolution operation on the spliced ​​feature and the target convolution kernel to obtain a convolution feature; input the convolution feature into the excitation network to obtain a fused second BEV feature.

[0081] In some embodiments, the acquisition module 701 is specifically used to: acquire at least one original image corresponding to the target area in real time; each of the at least one original image is acquired by a different image acquisition device; and perform feature extraction on the at least one original image to generate a current BEV feature.

[0082] In some embodiments, the acquisition module 701 is specifically used to: pass at least one original image through a convolutional neural network to obtain primary features corresponding to each original image; perform BEV conversion processing on the primary features corresponding to each original image to obtain current BEV features.

[0083] In some embodiments, the target detection device further includes a storage module 704, and the storage module 704 is used to store at least one of the following: current BEV features and fused BEV features.

[0084] In some embodiments, the detection module 703 is specifically used to: use an encoder to encode the fused BEV features to obtain an encoding result; use a target detection algorithm to detect the encoding result to obtain a detection result.

[0085] The target detection device provided in this embodiment can execute the target detection method provided in the above method embodiment. Its implementation principle and technical effect are similar to those of the above method and will not be repeated here.

[0086] Figure 8 An electronic device according to an exemplary embodiment may include a processor 802, and the processor 802 is used to execute application code to implement the target detection method in the present application.

[0087] The processor 802 may be a central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the present application.

[0088] like Figure 8 As shown, the electronic device may further include a memory 803. The memory 803 is used to store application program codes for executing the solution of the present application, and the execution is controlled by the processor 802.

[0089] The memory 803 may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 803 may exist independently and be connected to the processor 802 via the bus 804. The memory 803 may also be integrated with the processor 802.

[0090] like Figure 8As shown, the electronic device may further include a communication interface 801, wherein the communication interface 801, the processor 802, and the memory 803 may be coupled to each other, for example, via a bus 804. The communication interface 801 is used to exchange information with other devices, for example, to support information exchange between the electronic device and other devices.

[0091] It should be pointed out that Figure 8 The device structure shown in the figure does not constitute a limitation on the electronic device, except Figure 8 In addition to the components shown, the electronic device may include more or fewer components than shown, or combine certain components, or arrange the components differently. The electronic device provided in this embodiment can execute the target detection method provided in the above method embodiment, and its implementation principle and technical effect are similar to the above method, which will not be repeated here.

[0092] An embodiment of the present application provides a computer-readable storage medium, characterized in that a computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the various processes of the target detection method in the above method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0093] The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0094] An embodiment of the present application provides a computer program product, which stores a computer program. When the computer program is executed by a processor, the various processes of the target detection method in the above method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0095] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media that include computer-usable program code.

[0096] In the present application, the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0097] In this application, memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0098] In this application, computer-readable media includes permanent and non-permanent, removable and non-removable storage media. Storage media can be implemented by any method or technology to store information, and the information can be computer-readable instructions, data structures, modules of programs or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. According to the definition in this article, computer-readable media does not include temporary computer-readable media (transitory media), such as modulated data and carrier waves.

[0099] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0100] The above description is only a specific implementation of the present application, so that those skilled in the art can understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments described herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A target detection method, characterized in that: include: Acquire a current BEV feature corresponding to the target area and at least one historical BEV feature; the current BEV feature is generated by extracting at least one original image, and the original image is an image corresponding to the target area; integrating the at least one historical BEV feature into the current BEV feature to obtain a fused BEV feature; Perform target detection on the fused BEV features to obtain a detection result.

2. The target detection method according to claim 1, characterized in that: The step of integrating the at least one historical BEV feature into the current BEV feature to obtain a fused BEV feature includes: performing a fusion operation on the at least one historical BEV feature and the current BEV feature in sequence in chronological order to obtain a fused BEV feature; The fusion operation includes: integrating the first feature into the second BEV feature to obtain a fused second BEV feature; when the second BEV feature is the current BEV feature, determining the fused second BEV feature as a fused BEV feature; The first feature is the first BEV feature or the fused first BEV feature; the first BEV feature is the BEV feature generated at the first moment, and the second BEV feature is the BEV feature generated at the second moment; the first moment is adjacent to the second moment, and the first moment is earlier than the second moment.

3. The target detection method according to claim 2, characterized in that: The step of integrating the first feature into the second BEV feature to obtain the integrated second BEV feature includes: According to the second BEV feature, position-transforming the first feature to obtain a conversion feature; Performing matrix concatenation of the conversion feature and the current BEV feature to obtain a concatenated feature; Performing a convolution operation on the concatenated feature and the target convolution kernel to obtain a convolution feature; The convolutional features are input into the excitation network to obtain the fused second BEV features.

4. The target detection method according to claim 1, characterized in that: The obtaining of the current BEV characteristics corresponding to the target area includes: Acquire at least one original image corresponding to the target area in real time; each of the at least one original image is acquired by a different image acquisition device; Perform feature extraction on the at least one original image to generate a current BEV feature.

5. The target detection method according to claim 4, characterized in that: The step of extracting features from the at least one original image to generate current BEV features includes: Passing the at least one original image through a convolutional neural network to obtain primary features corresponding to each original image; The primary features corresponding to each original image are processed by BEV conversion to obtain the current BEV features.

6. The target detection method according to claim 1, characterized in that: After obtaining the fused BEV feature, the method further includes: At least one of the following is stored: the current BEV feature, the fused BEV feature.

7. The target detection method according to any one of claims 1 to 6, characterized in that: The performing target detection on the fused BEV feature to obtain a detection result includes: Encode the fused BEV features using an encoder to obtain an encoding result; The encoding result is detected using a target detection algorithm to obtain a detection result.

8. A target detection device, characterized in that: include: An acquisition module, configured to acquire a current BEV feature corresponding to a target area and at least one historical BEV feature; The current BEV feature is generated by extracting at least one original image, where the original image is an image corresponding to the target area; a processing module, configured to integrate the at least one historical BEV feature into the current BEV feature to obtain a fused BEV feature; The detection module is used to perform target detection on the fused BEV features to obtain a detection result.

9. An electronic device, characterized in that: include: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the target detection method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: include: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the target detection method according to any one of claims 1 to 7 is implemented.

Images