Human shape detection method and device based on event data, equipment and storage medium

Abstract

This application provides a method, apparatus, device, and storage medium for human detection based on event data. The method includes: aggregating multiple event data frames associated with the current detection task to obtain an event image to be detected; inputting the event image to be detected into a trained lightweight human detection model to output target key feature data; and obtaining human detection results based on the target key feature data. By implementing the solution of this application, human detection based on event data collected by an event camera is applicable to human detection in high-dynamic scenes. Furthermore, the lightweight model has relatively low hardware performance requirements and is easy to deploy on conventional devices, effectively improving its applicability.

Description

Technical Field

[0001] This application relates to the field of computer vision technology, and in particular to a method, apparatus, device and storage medium for human detection based on event data. Background Technology

[0002] With the continuous development of science and technology, human detection has been widely used in fields such as assisted driving, monitoring, and robotics. Current mainstream human detection algorithms are based on deep learning technology and are typically applied to the detection of images captured by traditional CMOS cameras. Furthermore, the deployment of human detection models requires high hardware performance, resulting in significant application limitations. Summary of the Invention

[0003] This application provides a human detection method, apparatus, device, and storage medium based on event data, which can at least solve the problem of the large application limitations of human detection algorithms based on CMOS cameras provided in related technologies.

[0004] The first aspect of this application provides a human detection method based on event data, comprising: aggregating multiple event data frames associated with the current detection task to obtain an event image to be detected; inputting the event image to be detected into a trained lightweight human detection model to output target key feature data; and obtaining human detection results based on the target key feature data.

[0005] A second aspect of this application provides a human detection device based on event data, comprising: an aggregation module for aggregating multiple event data frames associated with the current detection task to obtain an event image to be detected; a detection module for inputting the event image to be detected into a trained lightweight human detection model and outputting target key feature data; and an acquisition module for acquiring human detection results based on the target key feature data.

[0006] A third aspect of this application provides an electronic device, including a memory and a processor, wherein the processor is used to execute a computer program stored in the memory, and when the processor executes the computer program, it implements the steps of the human detection method provided in the first aspect of this application.

[0007] The fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it implements the steps of the human detection method provided in the first aspect of this application.

[0008] As can be seen from the above, the human detection method, apparatus, device, and storage medium based on event data provided in this application aggregate multiple event data frames associated with the current detection task to obtain an event image to be detected; the event image to be detected is input into a trained lightweight human detection model, which outputs key feature data of the target; and the human detection result is obtained based on the key feature data of the target. Through the implementation of this application, human detection based on event data collected by an event camera is applicable to human detection in high-dynamic scenes. Furthermore, the lightweight model has relatively low hardware performance requirements and is easy to deploy on conventional devices, effectively improving its applicability. Attached Figure Description

[0009] Figure 1 A schematic diagram illustrating an application scenario provided in one embodiment of this application;

[0010] Figure 2 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application;

[0011] Figure 3 This is a schematic diagram of the basic process of a human figure detection method provided in an embodiment of this application;

[0012] Figure 4 A schematic diagram of the network structure of a lightweight humanoid detection model provided in an embodiment of this application;

[0013] Figure 5 A detailed flowchart illustrating a human detection method provided in an embodiment of this application;

[0014] Figure 6 A schematic diagram of the program modules of a humanoid detection device provided in an embodiment of this application. Detailed Implementation

[0015] To make the inventive objectives, features, and advantages of this application more apparent and understandable, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0016] In the description of the embodiments of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0017] The following will describe in detail, with reference to the accompanying drawings, an embodiment of the present application of a human figure detection method, apparatus, device and storage medium based on event data.

[0018] To address the limitations of existing CMOS camera-based human detection algorithms, this application provides an embodiment of an event data-based human detection method, applicable to, for example... Figure 1 The scenario shown may include an event camera 10 and an electronic device 20.

[0019] It is worth noting that the Event-based Vision Sensor (EVS) configured in the event camera 10 is a novel sensor that mimics the human retina, responding to pixel pulses caused by brightness changes due to motion. In other words, the event camera 10 asynchronously records brightness changes at pixels. When the brightness change exceeds a certain threshold, it outputs an event including coordinates (x, y), a timestamp (t), and event polarity (p, with values ​​of +1 and -1, representing increases and decreases in brightness, respectively). Each event is represented in the form e = (t, x, y, p). Therefore, it can capture scene brightness changes (i.e., light intensity changes) at an extremely high frame rate, recording events at specific times and locations in the image, forming an event stream rather than a frame stream. This solves the problems of information redundancy, large data storage requirements, and high real-time processing demands associated with traditional cameras. Furthermore, the electronic device 20 includes various terminal devices with data processing capabilities, including but not limited to smartphones, tablets, laptops, desktop computers, vehicle terminals, and airborne terminals.

[0020] exist Figure 1 In the application scenario shown, event camera 10 can collect relevant event data for the actual application scenario, and then the event camera sends the event data to electronic device 20. Electronic device 20 executes the following human detection method for the received event data: First, it aggregates multiple event data frames associated with the current detection task to obtain the event image to be detected; then, it inputs the event image to be detected into the trained lightweight human detection model and outputs the target key feature data; finally, it obtains the human detection result based on the target key feature data.

[0021] like Figure 2The diagram shown is a schematic representation of an electronic device according to an embodiment of this application. The electronic device mainly includes a memory 201 and a processor 202. The number of processors 202 can be one or more. The memory 201 stores a computer program 203 that can run on the processor 202. The memory 201 and the processor 202 are communicatively connected. When the processor 202 executes the computer program 203, it implements the aforementioned human detection method.

[0022] It should be noted that the memory 201 can be an internal storage unit, such as a hard disk or RAM; the memory can also be an external storage device, such as an external hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, or a Flash Card. Furthermore, the memory can include both internal storage units and external storage devices, and can also be used to temporarily store data that has been output or will be output. It should be noted that when the processor 202 is a neural network chip, the electronic device may not include the memory 202; whether the electronic device needs to use the memory 202 to store the corresponding computer program depends on the type of processor 202.

[0023] Additionally, the processor 202 can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), neural network chips, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0024] One embodiment of this application also provides a computer-readable storage medium, which may be disposed in the aforementioned electronic device. The computer-readable storage medium may be as described above. Figure 2 The memory in the illustrated embodiment.

[0025] The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the aforementioned humanoid detection method. Furthermore, the computer-readable storage medium can also be a USB flash drive, external hard drive, read-only memory (ROM), RAM, magnetic disk, or optical disk, or any other medium capable of storing program code.

[0026] like Figure 3 This is a basic flowchart of a human figure detection method provided in an embodiment of this application. The human figure detection method can be... Figure 1 or Figure 2 The electronic device in the process executes the following steps:

[0027] Step 301: Aggregate multiple event data frames associated with the current detection task to obtain the event image to be detected.

[0028] In this embodiment, the event stream acquired by the event camera includes multiple event data frames. This embodiment can aggregate multiple event data frames based on the event representation data corresponding to pixels in each event data frame to obtain aggregated event data, and then generate corresponding event images based on the aggregated event data.

[0029] In one optional implementation of this embodiment, the step of aggregating multiple event data frames associated with the current detection task to obtain an event image to be detected includes: acquiring event representation data corresponding to the same pixel position in multiple event data frames associated with the current detection task; performing an OR operation on all event representation data at each pixel position to obtain aggregated event data; and generating an event image to be detected based on the aggregated event data.

[0030] Specifically, the event representation data in this embodiment includes 0 and 1, where 0 indicates no event is generated and 1 indicates an event is generated. In this embodiment, the event monitoring visual sensor configured in the event camera includes a pixel array composed of multiple pixels. Each pixel works independently, and only outputs an event when it detects a brightness change that reaches a preset brightness change threshold. In practical applications, the events generated by the pixels include positive and negative events. A positive event indicates that the brightness is stronger at the current moment compared to the previous moment, and a negative event indicates that the brightness is weaker at the current moment compared to the previous moment. In addition, no event is generated when there is no change in brightness.

[0031] In this embodiment, when aggregating event data, multiple event representation data corresponding to the same pixel position are obtained from multiple frames of event data, taking each pixel position in the overall pixel array as a unit. If at least one of the multiple event representation data includes a value of "1", then the pixel position is assigned a value of "1" in the aggregated event data. If all event representation data have values ​​of "0", then the pixel position is assigned a value of "0" in the aggregated event data. This operation is performed on all pixel positions in the overall pixel array to obtain the final aggregated event data. Next, a corresponding event image is generated based on the aggregated event data as the event image to be detected. It is worth noting that the above-described event data aggregation method is only one optional method. In actual application scenarios, other methods such as XOR operation can also be used. This embodiment does not limit this to a single method.

[0032] In one optional embodiment of this example, the step of generating an image of an event to be detected based on aggregated event data includes: determining a region of interest from the overall pixel array corresponding to the aggregated event data according to the task attributes of the current detection task; and generating an image of an event to be detected based on a portion of the event data corresponding to the region of interest in the aggregated event data.

[0033] Specifically, in practical applications, human detection scenarios vary, and different human detection scenarios have different detection task requirements. The reference value of event data in different pixel regions of aggregated event data for human detection varies, and the computational data volume of the global event image generated based on the overall event data is large. Based on this, this embodiment selects event data of pixel regions of interest from the overall event data according to the task attributes of the current detection task and generates the corresponding event image to be detected. This can improve the effectiveness of image details in the image to be detected and reduce the computational load of human detection.

[0034] Step 302: Input the image of the event to be detected into the trained lightweight human detection model and output the key feature data of the target.

[0035] Specifically, the lightweight human detection model in this embodiment is implemented based on a neural network, such as... Figure 4The diagram shown is a network structure diagram of a lightweight human detection model provided in this embodiment. The lightweight human detection model includes a first CBR module, a second CBR module, a third CBR module, a fourth CBR module, a fifth CBR module, a sixth CBR module, a max pooling layer, a concatenation module, an average pooling layer, a fully connected layer, and a normalization module. The first CBR module, the max pooling layer, the second CBR module, the third CBR module, the fourth CBR module, the fifth CBR module, the concatenation module, the average pooling layer, the fully connected layer, and the normalization module are cascaded in sequence. The input of the sixth CBR module is connected to the output of the third CBR module, and the output of the sixth CBR module is connected to the input of the concatenation module.

[0036] It should be noted that the sizes of the first CBR module, the max pooling layer, the second CBR module, the third CBR module, the fourth CBR module, the fifth CBR module, and the sixth CBR module are 8×8, 4×4, 4×4, 4×4, 4×4, 3×1, and 3×1, respectively. In addition, the number of channels of the first CBR module, the second CBR module, the third CBR module, the fourth CBR module, the fifth CBR module, and the sixth CBR module in this embodiment are 8, 12, 12, 16, 64, and 12, respectively. Furthermore, the first CBR module is downsampled by 4 times, the max pooling layer and the second CBR module are both downsampled by 2 times, and the third CBR module, the fourth CBR module, the fifth CBR module, and the sixth CBR module are not downsampled.

[0037] It should also be understood that the CBR module in this embodiment includes cascaded convolutional layers, batch normalization (BN) layers, and ReLU activation function layers. Furthermore, the size of the event image to be detected input to the lightweight human detection model in this embodiment is 192*192*1. Therefore, the model in this embodiment has fewer parameters and correspondingly lower computational load during application, resulting in lower hardware resource requirements, lower deployment costs, and ease of deployment on low-power devices such as MCUs.

[0038] In one optional embodiment of this example, before the step of inputting the image of the event to be detected into the trained lightweight human detection model and outputting the target key feature data, the method further includes: performing an affine transformation on each original APS image in a preset public APS dataset to generate a processed APS image; inputting the original APS image and the processed APS image into a preset event image simulation model to generate a simulated event image; and inputting a training sample set composed of all simulated event images into an initial lightweight human detection model for training to obtain a trained lightweight human detection model.

[0039] Specifically, the public dataset can be the Coco dataset, etc. In this embodiment, the original APS image in the dataset is subjected to affine transformation to adjust the position and angle of the image. The generated processed APS image has a certain position and angle offset relative to the original APS image. Then, the original APS image and the processed APS image are combined to simulate the event image. Finally, the initial lightweight human detection model is trained based on the training sample set composed of the simulated event images. When the loss value of the overall neural network no longer decreases, the model converges, and the trained lightweight human detection model is obtained.

[0040] In one optional implementation of this embodiment, the above-mentioned event image simulation model is represented as:

[0041]

[0042] Among them, image src Represents the original APS image, image dst Represents the processed APS image, Image dvs This represents a simulated event image.

[0043] Step 303: Obtain human detection results based on target key feature data.

[0044] Specifically, human detection can be a classification task. After extracting key features, the lightweight human detection model further classifies these features. In this embodiment, the human detection result includes whether someone is present or not. When the human detection result indicates that someone is present, this embodiment can further obtain the person recognition result based on the target key feature data. In practical applications, a preset skeleton key point feature library can be called. This feature library stores at least one legitimate skeleton key point feature. Then, the target key feature data is compared with the legitimate skeleton key point features in the skeleton key point feature library to obtain the actual feature matching degree. If the actual feature matching degree is higher than a preset matching degree threshold, the person recognition result is determined to be a legitimate person identified.

[0045] Furthermore, the task recognition results can be uploaded to the application platform, which can then implement corresponding control. Taking electronic door locks as an example, an unlocking command can be sent to the electronic lock control system based on the facial recognition results of a legitimate person. The electronic lock control system then controls the mechanical lock body of the door lock to perform the unlocking action according to the unlocking command.

[0046] In one optional implementation of this embodiment, after the above-mentioned step of obtaining human detection results based on target key feature data, the method further includes: if the human detection result indicates that a person has appeared, then obtaining the security application level required by the application task corresponding to the current detection task; and obtaining the person recognition result based on a portion of the key feature data corresponding to the security application level in the target key feature data.

[0047] Specifically, as mentioned above, when someone is detected, this embodiment can further perform a person recognition task, that is, identify whether the currently appearing person is a legitimate person. In this embodiment, the person recognition result is often used in identity verification application scenarios. For identity verification tasks with different security requirements, there are usually different security application level requirements. This embodiment can adapt to the security application level and obtain a portion of the key feature data from all detected key feature data for person recognition. While ensuring basic security application requirements, it reduces the computational load of person recognition and effectively improves recognition efficiency.

[0048] Figure 5 The method described in this application is a refined human detection method provided in one embodiment. The implementation process of this human detection method includes the following steps:

[0049] Step 501: Perform affine transformation on each original APS image in the preset public APS dataset to generate the processed APS image;

[0050] Step 502: Input the original APS image and the processed APS image into the preset event image simulation model to generate a simulated event image;

[0051] Step 503: Input the training sample set consisting of all simulated event images into the initial lightweight human detection model for training, and obtain the trained lightweight human detection model;

[0052] Step 504: Obtain event representation data corresponding to the same pixel position from multiple event data frames associated with the current detection task;

[0053] Step 505: Perform an OR operation on all event representation data for each pixel location to obtain aggregated event data;

[0054] Step 506: Based on the task attributes of the current detection task, determine the region of interest from the overall pixel array corresponding to the aggregated event data;

[0055] Step 507: Generate an image of the event to be detected based on the partial event data corresponding to the pixel region of interest in the aggregated event data;

[0056] Step 508: Input the image of the event to be detected into the trained lightweight human detection model and output the key feature data of the target.

[0057] Step 509: Obtain human detection results based on target key feature data.

[0058] It should be understood that the sequence number of each step in this embodiment does not imply the order in which the steps are executed. The execution order of each step should be determined by its function and internal logic, and should not constitute a unique limitation on the implementation process of this application embodiment.

[0059] Figure 6 A human detection device based on event data is provided as an embodiment of this application. This human detection device can be used to implement the human detection method in the foregoing embodiments. The human detection device mainly includes:

[0060] The aggregation module 601 is used to aggregate multiple event data frames associated with the current detection task to obtain an image of the event to be detected;

[0061] The detection module 602 is used to input the image of the event to be detected into the trained lightweight human detection model and output the key feature data of the target.

[0062] The acquisition module 603 is used to acquire human detection results based on the target's key feature data.

[0063] In some implementations of this embodiment, the aggregation module is specifically used to: acquire event representation data corresponding to the same pixel position in multiple event data frames associated with the current detection task; wherein, the event representation data includes 0 and 1, where 0 represents no event is generated and 1 represents an event is generated; perform OR operation on all event representation data at each pixel position to obtain aggregated event data; and generate an image of the event to be detected based on the aggregated event data.

[0064] Furthermore, in some embodiments of this example, when the aggregation module performs the function of generating an image of an event to be detected based on the aggregated event data, it is specifically used to: determine the region of interest from the overall pixel array corresponding to the aggregated event data according to the task attributes of the current detection task; and generate an image of an event to be detected based on the partial event data corresponding to the region of interest in the aggregated event data.

[0065] In some embodiments of this example, the lightweight humanoid detection model includes a first CBR module, a second CBR module, a third CBR module, a fourth CBR module, a fifth CBR module, a sixth CBR module, a max pooling layer, a fusion module, an average pooling layer, a fully connected layer, and a normalization module. The first CBR module, the max pooling layer, the second CBR module, the third CBR module, the fourth CBR module, the fifth CBR module, the fusion module, the average pooling layer, the fully connected layer, and the normalization module are cascaded in sequence. The input of the sixth CBR module is connected to the output of the third CBR module, and the output of the sixth CBR module is connected to the input of the fusion module. The dimensions of the first CBR module, the max pooling layer, the second CBR module, the third CBR module, the fourth CBR module, the fifth CBR module, and the sixth CBR module are 8×8, 4×4, 4×4, 4×4, 3×1, and 3×1, respectively.

[0066] In some embodiments of this example, the human detection device further includes a generation module and a training module. The generation module is used to: perform affine transformations on each original APS image in a preset public APS dataset to generate processed APS images; and input the original APS images and processed APS images into a preset event image simulation model to generate simulated event images. The training module is used to: input a training sample set composed of all simulated event images into an initial lightweight human detection model for training, and obtain a trained lightweight human detection model.

[0067] In some embodiments of this example, the acquisition module is further configured to: if the human detection result indicates that a person has appeared, acquire the security application level required by the application task corresponding to the current detection task; and acquire the person recognition result based on a portion of the key feature data corresponding to the security application level in the target key feature data.

[0068] It should be noted that the human detection methods in the foregoing embodiments can all be implemented based on the human detection device provided in this embodiment. Those skilled in the art can clearly understand that, for the sake of convenience and brevity, the specific working process of the human detection device described in this embodiment can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0069] Based on the technical solution of the above embodiments of this application, multiple event data frames associated with the current detection task are aggregated to obtain an event image to be detected; the event image to be detected is input into a trained lightweight human detection model, which outputs target key feature data; and the human detection result is obtained based on the target key feature data. Through the implementation of the solution of this application, human detection is performed based on event data collected by an event camera, which is suitable for human detection in high-dynamic scenes. Furthermore, the lightweight model has relatively low hardware performance requirements and is easy to deploy on conventional devices, effectively improving its applicability.

[0070] It should be noted that the apparatuses and methods disclosed in the several embodiments provided in this application can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0071] The modules described as separate components may or may not be physically separate. Similarly, the components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0072] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0073] If the integrated module is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a readable 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 of the various embodiments of this application. The aforementioned readable storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0074] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0075] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0076] The above is a description of the human detection method, apparatus, device and storage medium based on event data provided in this application. For those skilled in the art, based on the ideas of the embodiments of this application, there will be changes in the specific implementation and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A human detection method based on event data, characterized in that, include: Multiple event data frames associated with the current detection task are aggregated to obtain an image of the event to be detected; The image of the event to be detected is input into the trained lightweight human detection model, and the target key feature data is output. Human detection results are obtained based on the key feature data of the target; If the human detection result indicates that someone has appeared, then obtain the security application level required by the application task corresponding to the current detection task; Based on the target key feature data, the person recognition result is obtained from the partial key feature data corresponding to the security application level.

2. The human figure detection method according to claim 1, characterized in that, The step of aggregating multiple event data frames associated with the current detection task to obtain an image of the event to be detected includes: Acquire event representation data corresponding to the same pixel position from multiple event data frames associated with the current detection task; wherein, the event representation data includes 0 and 1, where 0 indicates no event occurred and 1 indicates an event occurred; Perform an OR operation on all the event representation data for each pixel location to obtain aggregated event data; An image of the event to be detected is generated based on the aggregated event data.

3. The human figure detection method according to claim 2, characterized in that, The step of generating an image of the event to be detected based on the aggregated event data includes: Based on the task attributes of the current detection task, determine the region of interest from the overall pixel array corresponding to the aggregated event data; An image of the event to be detected is generated based on a portion of the event data corresponding to the region of interest in the aggregated event data.

4. The human figure detection method according to claim 1, characterized in that, The lightweight humanoid detection model includes a first CBR module, a second CBR module, a third CBR module, a fourth CBR module, a fifth CBR module, a sixth CBR module, a max pooling layer, a fusion module, an average pooling layer, a fully connected layer, and a normalization module. The first CBR module, the max pooling layer, the second CBR module, the third CBR module, the fourth CBR module, the fifth CBR module, the fusion module, the average pooling layer, the fully connected layer, and the normalization module are cascaded in sequence. The input of the sixth CBR module is connected to the output of the third CBR module, and the output of the sixth CBR module is connected to the input of the fusion module. The dimensions of the first CBR module, the max pooling layer, the second CBR module, the third CBR module, the fourth CBR module, the fifth CBR module, and the sixth CBR module are 8×8, 4×4, 4×4, 4×4, 3×1, and 3×1, respectively.

5. The human figure detection method according to claim 1, characterized in that, Before the step of inputting the image of the event to be detected into the trained lightweight human detection model and outputting the key feature data of the target, the method further includes: Affine transformation is performed on each original APS image in the preset public APS dataset to generate a processed APS image. The original APS image and the processed APS image are input into a preset event image simulation model to generate a simulated event image. The training sample set consisting of all the simulated event images is input into the initial lightweight human detection model for training, and the trained lightweight human detection model is obtained.

6. The human figure detection method according to claim 5, characterized in that, The event image simulation model is represented as follows: ； in, This refers to the original APS image. This refers to the processed APS image. This represents the simulated event image.

7. A humanoid detection device based on event data, characterized in that, include: The aggregation module is used to aggregate multiple event data frames associated with the current detection task to obtain an image of the event to be detected. The detection module is used to input the image of the event to be detected into the trained lightweight human detection model and output the key feature data of the target. The acquisition module is used to acquire human detection results based on the target key feature data; if the human detection result indicates that a person has appeared, the security application level required by the application task corresponding to the current detection task is acquired. Based on the target key feature data, the person recognition result is obtained from the partial key feature data corresponding to the security application level.

8. An electronic device, characterized in that, Includes memory and processor, of which: The processor is used to execute computer programs stored in the memory; When the processor executes the computer program, it implements the steps in the human figure detection method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the human figure detection method according to any one of claims 1 to 6.

Images