Target detection method and device, electronic equipment and storage medium

Abstract

The invention relates to a target detection method and device, electronic equipment and a storage medium, and is applied to the technical field of image processing, and the method comprises the steps: carrying out the feature extraction of a to-be-detected image through a Backbone part, and obtaining a first image feature; processing the first image feature by using a Neck part to obtain a second image feature; the Neck part is obtained by sequentially connecting two convolutional layers, one Zoomcat module and a c2fODConv module, two convolutional layers, one Zoomcat module and a c2fODConv module, one convolutional layer and one connecting layer, one c2fODConv module, one convolutional layer and one connecting layer, one c2fODConv module, one convolutional layer and one connecting layer, one c2fODConv module, one ScalSeq module and one Add module; and processing the second image feature by using the Head part to obtain a target detection result. And the detection accuracy is improved.

Description

Technical Field

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

[0002] In recent years, with the continuous advancement of my country's urbanization, the number of domestic motor vehicles has continued to increase, resulting in an increase in road traffic congestion and accidents. Foggy weather is a common severe weather environment. In foggy weather, the recognition ability of vehicles and pedestrians is greatly reduced due to the influence of foggy weather, and the probability of traffic accidents will also increase. Affected by various suspended particles such as smoke, smog, and mist, the imaging process in this environment has problems such as color distortion, blur, and low contrast, making subsequent classification, tracking, pedestrian re-identification, and object detection tasks very difficult. Therefore, improving the accuracy of road target detection in foggy weather is of great significance to the development of unmanned driving technology.

[0003] At present, target detection algorithms based on deep learning models include: two-stage target detection algorithms and single-stage target detection algorithms. The two-stage target detection algorithm mainly divides target detection into two stages. The first stage generates a series of candidate regions containing the approximate location information of the target, and the second stage classifies and organizes the candidate regions and processes the location accurately. The single-stage target detection algorithm is represented by the Single Shot multibox Detector (SSD) and YOLO (You Only Look Once) series. However, the detection accuracy of the above two methods is low. Summary of the invention

[0004] In order to solve the above technical problems, the present application provides a target detection method, device, electronic device and storage medium.

[0005] According to a first aspect of the present application, a target detection method is provided, comprising:

[0006] Acquire the image to be detected;

[0007] Input the image to be detected into a pre-trained AAO-YOLO detection model; wherein the AAO-YOLO detection model includes: a target Backbone part, a target Neck part and a target Head part;

[0008] Using the target Backbone part, extracting features of the image to be detected to obtain a first image feature; wherein the target Backbone part includes the original Backbone part in the YOLOv8 model;

[0009] The first image feature is processed by using the target Neck part to obtain a second image feature; wherein the target Neck part is obtained by sequentially connecting two convolutional layers in the original Neck part of the YOLOv8 model, one added Zoom_cat module, one added c2f_ODConv module, two convolutional layers in the original Neck part, one added Zoom_cat module, one added c2f_ODConv module, one convolutional layer in the original Neck part, one connection layer in the original Neck part, one added c2f_ODConv module, one convolutional layer in the original Neck part, one connection layer in the original Neck part, one added c2f_ODConv module, one ScalSeq module in the original Neck part, and one Add module in the original Neck part;

[0010] The second image feature is processed using the target Head part to obtain a target detection result; wherein the target Head part is the original Head part in the YOLOv8 model.

[0011] Optionally, the target Backbone part is obtained by sequentially connecting the AOD-Net module and the original Backbone part in the YOLOv8 model.

[0012] Optionally, the training method of the AAO-YOLO detection model includes:

[0013] Add 2 Zoom_cat modules and 4 c2f_ODConv modules to the original Neck part of the YOLOv8 model to build the initial detection model;

[0014] Acquire a sample image, and determine a target detection result in the sample image;

[0015] The sample image is used as input and the target detection result in the sample image is used as label data to train the initial detection model to generate the AAO-YOLO detection model.

[0016] According to a second aspect of the present application, there is provided a target detection device, comprising:

[0017] An image acquisition module to be detected is used to acquire the image to be detected;

[0018] The image input module to be detected is used to input the image to be detected into a pre-trained AAO-YOLO detection model; wherein the AAO-YOLO detection model includes: a target Backbone part, a target Neck part and a target Head part;

[0019] A first processing module is used to extract features of the image to be detected by using the target Backbone part to obtain a first image feature; wherein the target Backbone part includes an original Backbone part in the YOLOv8 model;

[0020] a second processing module, configured to process the first image feature using the target Neck part to obtain a second image feature; wherein the target Neck part is obtained by sequentially connecting two convolutional layers in the original Neck part of the YOLOv8 model, one added Zoom_cat module, one added c2f_ODConv module, two convolutional layers in the original Neck part, one added Zoom_cat module, one added c2f_ODConv module, one convolutional layer in the original Neck part, one connection layer in the original Neck part, one added c2f_ODConv module, one convolutional layer in the original Neck part, one connection layer in the original Neck part, one added c2f_ODConv module, one ScalSeq module in the original Neck part, and one Add module in the original Neck part;

[0021] The third processing module is used to process the second image feature using the target Head part to obtain a target detection result; wherein the target Head part is the original Head part in the YOLOv8 model.

[0022] Optionally, the target Backbone part is obtained by sequentially connecting the AOD-Net module and the original Backbone part in the YOLOv8 model.

[0023] Optionally, the target detection device further includes:

[0024] Model architecture building module, which is used to add 2 Zoom_cat modules and 4 c2f_ODConv modules to the original Neck part in the YOLOv8 model to build the initial detection model;

[0025] A sample data acquisition module, used to acquire a sample image and determine a target detection result in the sample image;

[0026] The model training module is used to take the sample image as input and the target detection result in the sample image as label data, train the initial detection model, and generate the AAO-YOLO detection model.

[0027] According to a third aspect of the present application, an electronic device is provided, comprising: a processor, wherein the processor is configured to execute a computer program stored in a memory, wherein the computer program implements the method described in the first aspect when executed by the processor.

[0028] According to a fourth aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method described in the first aspect is implemented.

[0029] According to a fifth aspect of the present application, a computer program product is provided. When the computer program product is run on a computer, the computer is enabled to execute the method described in the first aspect.

[0030] Compared with the prior art, the technical solution provided by the embodiments of the present application has the following advantages:

[0031] The YOLO series uses a structure called spatial pyramid, which divides the feature map into pyramids of different scales. On the basis of the YOLOv8 model, two Zoom_cat modules and four c2f_ODConv modules are added. Among them, the Neck part introduces the Zoom_cat module that integrates spatial and scale features to enhance the multi-scale information extraction capability of the model and add detailed information by integrating feature maps of different scales. Since small targets may lose details on large-scale feature maps, the Zoom_cat module also introduces a channel and position attention mechanism, focusing on small objects related to information channels and spatial positions to improve detection and segmentation performance. In addition, the two Zoom_cat modules are set after two different convolutional layers, which can operate on the high-dimensional features of the previous layer to generate new outputs containing more position information and context features, enhancing the model's ability to handle complex scenes. The two c2f_ODConv modules are respectively set after the Zoom_cat module to make use of the feature map obtained by the previous convolution layer, so that the c2f_ODConv module can directly access more context information, improve the feature expression ability, and further optimize and extract features after fusing multi-scale information, so that it can handle more objects of different sizes. The c2f_ODConv module processes the already spliced ​​feature map. Therefore, the two c2f_ODConv modules are respectively set after the connection layer so that the feature map is directly spliced ​​in the channel dimension, so as to further extract and optimize features at the same scale. By setting the above model structure, the accuracy of target detection can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] 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.

[0033] 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.

[0034] Figure 1 A flow chart of a target detection method in an embodiment of the present application;

[0035] Figure 2 A schematic diagram of a network structure of the AAO-YOLO detection model in an embodiment of the present application;

[0036] Figure 3 A schematic diagram of the structure of a target detection device in an embodiment of the present application;

[0037] Figure 4 A schematic diagram of the structure of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0038] 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.

[0039] 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.

[0040] The YOLOv8 model has received extensive attention and application in the field of target detection due to its advantages such as fast detection speed and high accuracy. However, under foggy road conditions, due to poor visibility, the images taken by the visual sensor are blurred, resulting in a lack of accuracy and difficulty in meeting accuracy requirements. In response to the above problems, the embodiment of the present application uses YOLOv8 as the basic research model, improves and redesigns its network structure, and obtains a new AAO-YOLO detection model.

[0041] A Zoom_cat module that integrates spatial and scale features is added to the Neck part to enhance the model's multi-scale information extraction capability. By fusing feature maps of different scales, detailed information is added. The Zoom_cat module also introduces a channel and position attention mechanism, focusing on small objects related to information channels and spatial positions to improve detection and segmentation performance.

[0042] In addition, a convolution module, namely the c2f_ODConv module, is added to the Neck part. It can use a new multi-dimensional attention mechanism kernel parallel strategy to learn the complementary attention of the convolution kernel along all four dimensions of the kernel space in any convolution layer to improve the detection accuracy.

[0043] See also Figure 1 , Figure 1 A flow chart of a target detection method in an embodiment of the present application may include the following steps:

[0044] Step S102, obtaining an image to be detected.

[0045] The image to be detected may be an image taken in haze weather, and the image quality of the image to be detected is poor.

[0046] Step S104, inputting the image to be detected into a pre-trained AAO-YOLO detection model; wherein the AAO-YOLO detection model includes: a target Backbone part, a target Neck part and a target Head part.

[0047] The AAO-YOLO detection model is a model built on the basis of the YOLOv8 model. The YOLOv8 model includes the Backbone part, the Neck part, and the Head part. The Backbone part is the basis of the model and is used to extract features from the input image. The extracted features are the basis for target detection in the subsequent network layers. The Neck part is located between the Backbone part and the Head part and is used for feature fusion and enhancement. The Head part is the decision-making part of the YOLOv8 model and is used to generate the final detection result. As mentioned above, the Zoom_cat module and the c2f_ODConv module can be added to the Neck part to obtain the target Neck part to improve the detection accuracy.

[0048] Step S106, using the target Backbone part, extracting features of the image to be detected to obtain a first image feature; wherein the target Backbone part includes the original Backbone part in the YOLOv8 model.

[0049] The target Backbone part can be the same as the original Backbone part in the YOLOv8 model, or a defogging module can be added to the original Backbone part in the YOLOv8 model to improve the target detection performance of blurred images.

[0050] Optionally, the target Backbone part is obtained by sequentially connecting the AOD-Net module and the original Backbone part in the YOLOv8 model. That is, a dehazing module, namely the AOD-Net module, is added to the Backbone part. The AOD-Net module utilizes a multi-scale feature extraction strategy, captures different details of the image through convolution kernels of different sizes, and enhances the expression ability of the model through feature splicing. The purpose is to combine the input blurred image and the features extracted by the network to remove haze and improve the target detection performance of blurred images.

[0051] Step S108, using the target Neck part to process the first image feature to obtain the second image feature. The target Neck part is obtained by sequentially connecting 2 convolutional layers in the original Neck part of the YOLOv8 model, 1 added Zoom_cat module, 1 added c2f_ODConv module, 2 convolutional layers in the original Neck part, 1 added Zoom_cat module, 1 added c2f_ODConv module, 1 convolutional layer in the original Neck part, 1 connection layer in the original Neck part, 1 added c2f_ODConv module, 1 convolutional layer in the original Neck part, 1 connection layer in the original Neck part, 1 added c2f_ODConv module, 1 ScalSeq module in the original Neck part, and 1 Add module in the original Neck part.

[0052] See also Figure 2 , Figure 2 A schematic diagram of the network structure of the AAO-YOLO detection model in the embodiment of the present application. The YOLO series adopts a structure called a spatial pyramid, which divides the feature map into pyramids of different scales. The Zoom_cat module can capture targets of different sizes. Since small targets may lose details on large-scale feature maps, this is especially important for small target detection. It can be seen that the two added Zoom_cat modules are respectively set after the two convolutional layers, so that the high-dimensional features of the previous layer can be used to operate and generate new outputs containing more position information and context features, which enhances the model's ability to handle complex scenes.

[0053] The two c2f_ODConv modules are set after the Zoom_cat module to make use of the feature map obtained by the previous convolution layer, so that the c2f_ODConv module can directly access more context information, improve the feature expression ability, and further optimize and extract features after fusing multi-scale information, so that it can handle more objects of different sizes. The c2f_ODConv module processes the already spliced ​​feature map, so the two c2f_ODConv modules are set after the connection layer respectively, so that the feature map is directly spliced ​​in the channel dimension, so as to further extract and optimize features at the same scale.

[0054] After being processed by the c2f_ODConv module, multi-scale feature fusion and 3D convolution operations are performed through the ScalSeq module to enhance the model's ability to detect objects of different scales. The two feature maps are added together through the Add module to be suitable for a variety of feature fusion scenarios.

[0055] Step S110, using the target Head part to process the second image feature to obtain a target detection result; wherein the target Head part is the original Head part in the YOLOv8 model.

[0056] The target detection method of the embodiment of the present application introduces a Zoom_cat module that integrates spatial and scale features in the Neck part of the YOLOv8 model, which can enhance the multi-scale information extraction capability of the model and increase detailed information by integrating feature maps of different scales. Since small targets may lose details on large-scale feature maps, the Zoom_cat module also introduces a channel and position attention mechanism, focusing on small objects related to information channels and spatial positions to improve detection and segmentation performance. In addition, the two Zoom_cat modules are set after two different convolutional layers, and the high-dimensional features of the previous layer can be used for operation to generate new outputs containing more position information and context features, thereby enhancing the model's ability to handle complex scenes. The c2f_ODConv module is also introduced in the Neck part, and the two c2f_ODConv modules are respectively set after the Zoom_cat module, so as to utilize the feature maps obtained by the previous convolutional layer, so that the c2f_ODConv module can directly access more context information, improve feature expression capabilities, and further optimize and extract features after fusing multi-scale information, so that more objects of different sizes can be processed. The c2f_ODConv module processes the already concatenated feature maps. Therefore, two c2f_ODConv modules are respectively set after the connection layer so that the feature maps are directly concatenated in the channel dimension, thereby further extracting and optimizing features at the same scale. It can be seen that by setting the above model structure, the accuracy of target detection can be improved.

[0057] Optionally, the training method of the AAO-YOLO detection model includes:

[0058] Add 2 Zoom_cat modules and 4 c2f_ODConv modules to the original Neck part of the YOLOv8 model to build an initial detection model. The network structure of the initial detection model is the same as that of the trained AAO-YOLO detection model, but the network parameter values ​​are different.

[0059] A sample image is obtained, and an object detection result in the sample image is determined.

[0060] In an embodiment of the present application, an RTTS image dataset under real haze weather can be obtained. The dataset is derived from the RESIDE dataset and contains 4322 real foggy weather pictures, covering different meteorological conditions such as daytime, cloudy, foggy, and hazy days under different visibility conditions, as well as different road environments such as cities, rural areas, and highways. It involves a wide range of target objects with multiple categories, as well as different degrees of occlusion and truncation. The dataset sets 5 target detection categories, namely cars, people, buses, bicycles, and motorcycles, with a total of about 41,000 related targets. The present application intends to divide the dataset in a ratio of 6:3:1, and the final training set contains 2592 pictures, the verification set contains 1297 pictures, and the test set contains 433 pictures.

[0061] The sample image (i.e., sample set) is used as input and the target detection results in the sample image are used as label data to train the initial detection model and generate the AAO-YOLO detection model.

[0062] The validation set is used to test the performance of the model on unknown data and fine-tune the model parameters. The test set is used to test the performance of the finally trained AAO-YOLO detection model.

[0063] When the performance of the AAO-YOLO detection model meets the requirements, the model in this paper is compared with YOLOv3, YOLOv5, YOLOv7 and other models in the same experimental environment. The experimental results of each model are shown in Table 1.

[0064] Table 1

[0065]

[0066] From the experimental results in Table 1, we can see that compared with other models, the average precision mAP of this model has been significantly improved. Compared with YOLOv3, YOLOv5, YOLOv7 and other models, it has increased by 7.1%, 5.9%, and 3.6% respectively. Moreover, this model has also achieved certain improvement in the three experimental indicators of precision, recall, and mAP5-95. Through the above comparative experimental results, it is verified that this model can effectively improve various performance indicators to a certain extent, and its comprehensive ability is superior to other models.

[0067] Ablation experiments are used to verify the different effects of the various improvement methods proposed in this application on the detection performance of the original YOLOv8 algorithm. Taking the RTTS dataset as the main sample, 8 groups of ablation experiments are designed. The effects of different improvement methods on the model detection performance are shown in Table 2.

[0068] Table 2

[0069]

[0070] As can be seen from Table 2, the accuracy of the model in this paper reaches 78.1%, and the average precision mAP reaches 72.3%. In the ablation experiment, it can be seen that the presence or absence of the AOD-Net module has a significant impact on the experimental accuracy and has a great impact on subsequent improvements. After YOLOv8 merged the Zoom_cat module, the accuracy and recall rate increased by 1.9% and 1.4% respectively. After fusing the C2f_ODConv module, the parameters of the original model were reduced from 8.1GFLOPs (1 billion floating-point operations per second) to 7.9GFLOPs. Experiment 8 merged the three modules, which improved the accuracy, recall rate and average precision of the original model by 2%, 7% and 4.2% respectively. The final experiment proved that it has higher accuracy than the original model, verifying that the AAO-YOLO detection model can effectively improve the detection ability of foggy road scene pictures.

[0071] It should be noted that although the steps of the method in the present application are described in a specific order in the drawings, this does not require or imply that the steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps, etc.

[0072] Corresponding to the above method embodiment, the present application embodiment also provides a target detection device, see Figure 3 , the target detection device 300 includes:

[0073] The image acquisition module 302 is used to acquire the image to be detected;

[0074] The image input module 304 is used to input the image to be detected into the pre-trained AAO-YOLO detection model; wherein the AAO-YOLO detection model includes: a target Backbone part, a target Neck part and a target Head part;

[0075] The first processing module 306 is used to extract features of the image to be detected by using the target Backbone part to obtain a first image feature; wherein the target Backbone part includes the original Backbone part in the YOLOv8 model;

[0076] The second processing module 308 is used to process the first image feature by using the target Neck part to obtain the second image feature; wherein the target Neck part is obtained by sequentially connecting 2 convolutional layers in the original Neck part of the YOLOv8 model, 1 added Zoom_cat module, 1 added c2f_ODConv module, 2 convolutional layers in the original Neck part, 1 added Zoom_cat module, 1 added c2f_ODConv module, 1 convolutional layer in the original Neck part, 1 connection layer in the original Neck part, 1 added c2f_ODConv module, 1 convolutional layer in the original Neck part, 1 connection layer in the original Neck part, 1 added c2f_ODConv module, 1 ScalSeq module in the original Neck part, and 1 Add module in the original Neck part;

[0077] The third processing module 310 is used to process the second image feature using the target Head part to obtain a target detection result; wherein the target Head part is the original Head part in the YOLOv8 model.

[0078] Optionally, the target Backbone part is obtained by sequentially connecting the AOD-Net module and the original Backbone part in the YOLOv8 model.

[0079] Optionally, the target detection device 300 further includes:

[0080] Model architecture building module, which is used to add 2 Zoom_cat modules and 4 c2f_ODConv modules to the original Neck part in the YOLOv8 model to build the initial detection model;

[0081] A sample data acquisition module is used to acquire a sample image and determine the target detection result in the sample image;

[0082] The model training module is used to take the sample image as input and the target detection results in the sample image as label data, train the initial detection model, and generate the AAO-YOLO detection model.

[0083] The specific details of each module or unit in the above device have been described in detail in the corresponding method, so they will not be repeated here.

[0084] It should be noted that, although several modules or units of the equipment for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into being embodied by multiple modules or units.

[0085] The present application also provides an electronic device, see Figure 4 , Figure 4 The electronic device of the embodiment of the present application includes: a processor 401, a communication interface 402, a memory 403 and a communication bus 404, wherein the processor 401, the communication interface 402, and the memory 403 communicate with each other through the communication bus 404;

[0086] Memory 403, used for storing computer programs;

[0087] The processor 401 is used to implement the above-mentioned target detection method when executing the program stored in the memory 403.

[0088] It should be noted that the communication bus 404 mentioned in the above electronic device can be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The communication bus 404 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0089] The communication interface 402 is used for communication between the electronic device and other devices.

[0090] The memory 403 may include a RAM (Random Access Memory), or may include a non-volatile memory, such as at least one disk memory. Optionally, the memory 403 may also be at least one storage device located away from the aforementioned processor.

[0091] The above-mentioned processor 401 can be a general-purpose processor, including: CPU (Central Processing Unit), NP (Network Processor), etc.; it can also be DSP (Digital Signal Processing), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.

[0092] An embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned target detection method is implemented.

[0093] It should be noted that the computer-readable storage medium shown in the present application may 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 wires, a portable computer disk, a hard disk, a random access memory, a read-only memory, 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 the present application, 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, device or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, radio frequency, etc., or any suitable combination of the above.

[0094] In an embodiment of the present application, a computer program product is also provided. When the computer program product is run on a computer, the computer executes the above-mentioned target detection method.

[0095] 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.

[0096] 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 the image to be detected; Input the image to be detected into a pre-trained AAO-YOLO detection model; wherein the AAO-YOLO detection model includes: a target Backbone part, a target Neck part and a target Head part; Using the target Backbone part, extracting features of the image to be detected to obtain a first image feature; wherein the target Backbone part includes the original Backbone part in the YOLOv8 model; The first image feature is processed by using the target Neck part to obtain a second image feature; wherein the target Neck part is obtained by sequentially connecting two convolutional layers in the original Neck part of the YOLOv8 model, one added Zoom_cat module, one added c2f_ODConv module, two convolutional layers in the original Neck part, one added Zoom_cat module, one added c2f_ODConv module, one convolutional layer in the original Neck part, one connection layer in the original Neck part, one added c2f_ODConv module, one convolutional layer in the original Neck part, one connection layer in the original Neck part, one added c2f_ODConv module, one ScalSeq module in the original Neck part, and one Add module in the original Neck part; The second image feature is processed using the target Head part to obtain a target detection result; wherein the target Head part is the original Head part in the YOLOv8 model.

2. The method according to claim 1, characterized in that The target Backbone part is obtained by sequentially connecting the AOD-Net module and the original Backbone part in the YOLOv8 model.

3. The method according to claim 1, characterized in that The training method of the AAO-YOLO detection model includes: Add 2 Zoom_cat modules and 4 c2f_ODConv modules to the original Neck part of the YOLOv8 model to build the initial detection model; Acquire a sample image, and determine a target detection result in the sample image; The sample image is used as input and the target detection result in the sample image is used as label data to train the initial detection model to generate the AAO-YOLO detection model.

4. A target detection device, characterized in that: The device comprises: An image acquisition module to be detected is used to acquire the image to be detected; The image input module to be detected is used to input the image to be detected into a pre-trained AAO-YOLO detection model; wherein the AAO-YOLO detection model includes: a target Backbone part, a target Neck part and a target Head part; A first processing module is used to extract features of the image to be detected by using the target Backbone part to obtain a first image feature; wherein the target Backbone part includes an original Backbone part in the YOLOv8 model; a second processing module, configured to process the first image feature using the target Neck part to obtain a second image feature; wherein the target Neck part is obtained by sequentially connecting two convolutional layers in the original Neck part of the YOLOv8 model, one added Zoom_cat module, one added c2f_ODConv module, two convolutional layers in the original Neck part, one added Zoom_cat module, one added c2f_ODConv module, one convolutional layer in the original Neck part, one connection layer in the original Neck part, one added c2f_ODConv module, one convolutional layer in the original Neck part, one connection layer in the original Neck part, one added c2f_ODConv module, one ScalSeq module in the original Neck part, and one Add module in the original Neck part; The third processing module is used to process the second image feature using the target Head part to obtain a target detection result; wherein the target Head part is the original Head part in the YOLOv8 model.

5. The device according to claim 4, characterized in that The target Backbone part is obtained by sequentially connecting the AOD-Net module and the original Backbone part in the YOLOv8 model.

6. The device according to claim 4, characterized in that The device also includes: Model architecture building module, which is used to add 2 Zoom_cat modules and 4 c2f_ODConv modules to the original Neck part in the YOLOv8 model to build the initial detection model; A sample data acquisition module, used to acquire a sample image and determine a target detection result in the sample image; The model training module is used to take the sample image as input and the target detection result in the sample image as label data, train the initial detection model, and generate the AAO-YOLO detection model.

7. An electronic device, characterized in that: include: A processor, wherein the processor is used to execute a computer program stored in a memory, wherein the computer program, when executed by the processor, implements the method according to any one of claims 1 to 3.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 3 is implemented.

Images