Lightweight small target pest detection method and device and electronic equipment

Abstract

The invention provides a lightweight small target pest detection method and device and electronic equipment, and the method comprises the steps: carrying out the preprocessing of collected small target pest images, so as to construct a pest image data set; based on a YOLOv5 model, constructing a small target pest detection model; wherein a small target feature enhancement module is added in a trunk part of the YOLOv5s model, and the small target feature enhancement module is constructed based on a MobileNetv3 structure and a BiFomer structure; a VoV-GSCSP module is added to the neck part of the YOLOv5s model; training the small target pest detection model to obtain a trained small target pest detection model; and inputting the pest image data set into the trained small target pest detection model to obtain a detection result of the small target pest image. Through adoption of the method, the problems that the model complexity is high, the parameter quantity is large and the deployment of the model on edge equipment is not facilitated due to pursuing of relatively high recognition precision in a pest image detection method in the related technology are solved.

Description

Technical Field

[0001] The present invention relates to the field of image detection technology, and in particular to a lightweight small-target pest detection method, device and electronic equipment. Background Art

[0002] In agricultural production, many pests are small and hidden. When a large number of pests gather and overlap in a specific area, the pest detection model may have difficulty distinguishing the pests. In addition, pests are often distributed in a variety of complex environments, such as background shadows, species interference, etc., which leads to poor detection results of pest detection models. At the same time, existing pest image detection methods usually pursue higher recognition accuracy, with high model complexity and large number of parameters, which makes it more difficult to deploy the model on edge devices. Summary of the invention

[0003] The present invention provides a lightweight small-target pest detection method, device and electronic device to solve the problem that pest image detection methods in the prior art usually pursue higher recognition accuracy, have high model complexity and a large number of parameters, which increases the difficulty of deploying the model on edge devices.

[0004] The present invention provides a lightweight small target pest detection method, comprising the following steps: Preprocess the collected small target pest images to construct a pest image dataset; Based on the YOLOv5s model, a small target pest detection model is constructed; wherein a small target feature enhancement module is added to the trunk part of the YOLOv5s model, and the small target feature enhancement module is constructed based on the MobileNetv3 and BiFomer structures; a VoV-GSCSP module is added to the neck part of the YOLOv5s model; Training a small target pest detection model to obtain a trained small target pest detection model; The pest image dataset is input into the trained small target pest detection model to obtain the detection result of the small target pest image.

[0005] According to the lightweight small target pest detection method provided by the present invention, the feature processing process of the small target feature enhancement module includes: When the step size is 2, the feature information is subjected to the first ordinary convolution and the first depth convolution respectively; After the first ordinary convolution, the second depth convolution and BiFormer module operations are performed in sequence; After performing common convolution on the outputs after the BiFormer module operation and the first depth convolution, the output results are spliced; The spliced ​​output results are sequentially subjected to a deep convolution module and a common convolution to obtain the feature processing result of the small target feature enhancement module.

[0006] According to the lightweight small target pest detection method provided by the present invention, the feature processing process of the small target feature enhancement module includes: When the step size is 1, the feature information is split into channels to obtain feature information on both sides; After inputting the feature information on one side into the Mobile module, it is spliced ​​with the feature information on the other side; The spliced ​​output results are sorted by channels to obtain the feature processing results of the small target feature enhancement module.

[0007] According to the lightweight small target pest detection method provided by the present invention, the VoV-GSCSP module is added to the neck part of the YOLOv5s model, specifically including: The VoV-GSCSP module is constructed by combining the output of ordinary convolution with two layers of GSConv; The VoV-GSCSP module is added to the neck part of the YOLOv5s model.

[0008] The lightweight small target pest detection method provided by the present invention further includes: segmenting the pest image data set according to an optimal segmentation ratio to generate a training data set for training a small target pest detection model.

[0009] The lightweight small target pest detection method provided by the present invention is characterized in that the loss function for training the small target pest detection model is a NWD loss function.

[0010] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the lightweight small target pest detection method as described in any one of the above is implemented.

[0011] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the lightweight small target pest detection method as described in any one of the above is implemented.

[0012] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-described lightweight small-target pest detection methods.

[0013] The lightweight small target pest detection method, device and electronic device provided by the present invention increase the small target feature enhancement module based on MobileNetv3 and BiFomer structure in the main part of the YOLOv5s model; and increase the VoV-GSCSP module in the neck part of the YOLOv5s model. While improving the accuracy of small target pest detection, it also reduces the calculation amount, complexity and detection time of the model, making the model more convenient and quick to deploy on the mobile terminal, and realizing the detection of small target pests more accurately and efficiently. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0015] Figure 1 It is a schematic diagram of the process of the lightweight small target pest detection method provided by the present invention.

[0016] Figure 2 It is a schematic diagram of the principle of a small target pest detection model provided by the present invention.

[0017] Figure 3 It is a schematic diagram of the principle of the small target feature enhancement module under different step lengths provided by the present invention.

[0018] Figure 4 It is a structural schematic diagram of a lightweight small target pest detection device provided by the present invention.

[0019] Figure 5 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the descriptions used in this way can be interchanged where appropriate, so that the embodiments can be implemented in a sequence other than that illustrated or described in the present application. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules that are clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices. The naming or numbering of the steps that appear in the present application does not mean that the steps in the method flow must be executed in the time / logical sequence indicated by the naming or numbering. The process steps that have been named or numbered can change the execution order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved. The division of modules in this application is a logical division. There may be other division methods when it is implemented in actual applications. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection between units can be electrical or other similar forms, which are not limited in this application. In addition, the modules or submodules described as separate components may or may not be physically separated, may or may not be physical units, or may be distributed in multiple circuit units, and some or all of the units may be selected according to actual needs to achieve the purpose of the present application.

[0022] The present invention aims to solve the problems of difficulty in detecting small target pests in complex scenes in agricultural production fields, such as background shadows, target interference, dense pests, etc., as well as the high complexity and large computational complexity of existing small target detection models. The present invention proposes a lightweight small target pest detection method, device and electronic equipment for detecting small target pests in complex scenes, so as to ensure that the computational complexity, complexity and detection time of the model can be reduced while improving the detection accuracy of small target pests, so that the model can be deployed on the mobile terminal more conveniently and quickly, and the detection of small target pests can be achieved more accurately and efficiently.

[0023] Combine the following Figure 1-Figure 5 The specific contents of the present invention are described.

[0024] Figure 1 It is a flow chart of a lightweight small target pest detection method provided by the present invention, including steps S101-S104.

[0025] Step S101, preprocessing the collected small target pest images to construct a pest image dataset.

[0026] In an embodiment of the present invention, the uniformly collected small target pest images are resized to a size of 640×640 to adapt to the input size of the target pest detection model.

[0027] Exemplarily, small target pest images in different environments are collected by high-definition camera equipment, taking 60 pictures of indoor environments, 60 pictures of outdoor environments, and 60 pictures of indoor and outdoor mixed environments; The collected images are segmented, screened and divided, and cut according to a 5×5 ratio. Finally, they are divided into 1,500 indoor environment images, 1,500 outdoor environment images, and 1,500 indoor and outdoor mixed environment images. Corresponding training data sets, verification data sets and test data sets are produced, and preprocessing is performed on the three different environments.

[0028] In a possible implementation, the lightweight small target pest detection method further includes: selecting part of the image data from the pest image dataset to generate a training dataset for training the small target pest detection model.

[0029] In a possible implementation, a portion of image data is selected from a pest image dataset to generate a training dataset for training a small target pest detection model, including: The pest image dataset is segmented according to the optimal segmentation ratio to generate a training dataset for training a small target pest detection model.

[0030] Step S102, based on the YOLOv5s model, a small target pest detection model is constructed; wherein, a small target feature enhancement module is added to the trunk part of the YOLOv5s model, and the small target feature enhancement module is constructed based on the MobileNetv3 and BiFomer structures; and a VoV-GSCSP module is added to the neck part of the YOLOv5s model.

[0031] In an embodiment of the present invention, a small target feature enhancement module is based on MobileNetv3 and BiFomer structures, and is composed of ordinary convolution, depthwise convolution and point-by-point convolution arranged in a certain connection. Different convolution structures are designed by dividing different step lengths to achieve compression of the effective feature layer size, thereby integrating different information in the channel. The channels are sorted after output to achieve efficient extraction of fused features.

[0032] Specifically, if Figure 2 As shown, it is a schematic diagram of the principle of a small target pest detection model provided by the present invention.

[0033] like Figure 2As shown in the figure, Conv 3×3 represents ordinary convolution operation, Max-Pool 3×3 represents maximum pooling operation, EM Block represents small target feature enhancement module, GSConv convolution represents deep convolution + ordinary convolution operation, VoV-GSCSP represents GSConv + ordinary convolution operation, and Upsample represents upsampling operation.

[0034] In the embodiment of the present invention, the depthwise separable convolution mainly includes two specific operations: depthwise convolution and pointwise convolution. The depthwise convolution performs convolution on each input channel to obtain an output feature map with the same number of channels as the input feature map. The pointwise convolution uses a 1×1 convolution kernel for dimensionality increase. In an embodiment of the present invention, the Mobile module is used to expand the number of channels in the data before the data enters the deep convolution, use 3×3 depthwise separable convolution to perform channel convolution to achieve feature extraction, and then perform point-by-point convolution to reduce the number of channels.

[0035] In the embodiment of the present invention, the core idea of ​​the BiFormer structure is to introduce a dual-layer routing attention mechanism, with dynamic sparse attention and dual-layer routing as the core.

[0036] Among them, the BiFormer structure can capture global and local feature information at the same time. It is different from ordinary convolution, which is essentially a local operation. A key feature of attention is the global receptive field, which enables vision to capture long-distance dependencies and enables the model to better capture the internal correlation of the input data, thereby improving the performance of the model in visual tasks.

[0037] In an embodiment of the present invention, the GSConv structure combines the depth separation convolution with the ordinary convolution, because the feature map generated by the output channel of the shuffled DSC is still depth separation, thereby forming a new hybrid convolution GSConv.

[0038] In the embodiment of the present invention, the VoV-GSCSP structure: constructs GSbottleneck based on the combined output of one layer of ordinary convolution and two layers of GSConv, and combines a network layer structure with a higher reuse rate of features.

[0039] Among them, the VoV-GSCSP structure is constructed by combining the output of ordinary convolution and two layers of GSConv. The nonlinear expression ability is enhanced by adding DS layers and shuffle layers. These two structures are used at the neck of the model. Then, when the feature map reaches the neck, the channel dimension reaches the maximum value, the width and height dimensions reach the minimum value, and the transformation becomes moderate, so that the connected feature map does not need to be compressed, and there is less redundant and repeated information, so as to obtain better efficiency and performance.

[0040] Step S103: training a small target pest detection model to obtain a trained small target pest detection model.

[0041] Step S104, inputting the pest image dataset into the trained small target pest detection model to obtain the detection result of the small target pest image.

[0042] In a possible implementation, the feature processing process of the small target feature enhancement module includes: Figure 3 As shown in the figure, when the step size is 2, the feature information is subjected to the first ordinary convolution and the first deep convolution respectively; after the first ordinary convolution, the second deep convolution and the BiFormer module operation are performed in sequence; after the outputs after the BiFormer module operation and the first deep convolution are subjected to ordinary convolution respectively, the output results are spliced; the spliced ​​output results are subjected to the deep convolution module and the ordinary convolution in sequence to obtain the feature processing result of the small target feature enhancement module.

[0043] In a possible implementation, the feature processing process of the small target feature enhancement module includes: Figure 3 As shown in the figure, when the step size is 1, the feature information is split into channels to obtain feature information on both sides; after the feature information on one side is input into the Mobile module, it is spliced ​​with the feature information on the other side; the spliced ​​output results are sorted by channels to obtain the feature processing results of the small target feature enhancement module.

[0044] In a possible implementation, the loss function for training the small target pest detection model is a NWD loss function.

[0045] Specifically, the loss function in the model is modified, and the NWD loss function is used to improve the detection performance of the model for small target characteristics; the calculation formula of the NWD loss function is:

[0046] in Represents a new distance metric, the predicted bounding box is P and the true box is G, and the Gaussian distribution models of the predicted boxes P and G are .

[0047] The embodiment of the present invention uses Python language to lightweight small target pest detection model on a 64-bit Windows 10 operating system platform, and uses NVIDIA GTX 3090ti GPU to complete model training and testing. The batch on a single GPU is 32, 640×640 pixel images, the initial learning rate of the model is set to 0.01, the number of training rounds is 300 batches, the corresponding Python version is 3.9, the Pytorch deep learning framework is used, and the Pytorch version is 1.12.1.

[0048] The present invention can achieve the following beneficial effects by adopting the above method: 1. By adding a small target feature enhancement module based on MobileNetv3 and BiFomer structure to the main part of the YOLOv5s model and adding a VoV-GSCSP module to the neck part of the YOLOv5s model, the accuracy of small target pest detection is improved while reducing the model's computational complexity and detection time, making the model more convenient and quick to deploy on mobile terminals, and realizing small target pest detection more accurately and efficiently.

[0049] 2. The BiFormer structure is introduced, and the attention output is nonlinearly transformed and feature extracted through a multi-layer perceptron, which improves the expressiveness and performance of the model and improves the detection effect and efficiency of the model.

[0050] 3. Replacing the traditional IoU loss function with the NWD loss function makes the detection target more refined and conducts targeted small target pest detection, which helps to improve the model convergence speed.

[0051] The lightweight small target pest detection device provided by the present invention is described below. The lightweight small target pest detection device described below and the lightweight small target pest detection method described above can be referred to each other.

[0052] Figure 4 A schematic diagram of the structure of a lightweight small target pest detection device provided by the present invention includes the following parts: An image processing unit 410 is used to pre-process the collected small target pest images to construct a pest image dataset; The model building unit 420 is used to build a small target pest detection model based on the YOLOv5s model; wherein a small target feature enhancement module is added to the trunk part of the YOLOv5s model, and the small target feature enhancement module is built based on the MobileNetv3 and BiFomer structures; and a VoV-GSCSP module is added to the neck part of the YOLOv5s model; A model training unit 430 is used to train a small target pest detection model to obtain a trained small target pest detection model; The model application unit 440 is used to input the pest image dataset into the trained small target pest detection model to obtain the detection result of the small target pest image.

[0053] In a possible implementation, the feature processing process of the small target feature enhancement module includes: when the step size is 2, performing a first ordinary convolution and a first deep convolution on the feature information respectively; after the first ordinary convolution, performing a second deep convolution and a BiFormer module operation in sequence; performing ordinary convolution on the outputs after the BiFormer module operation and the first deep convolution respectively, and splicing the output results; performing a deep convolution module and an ordinary convolution on the spliced ​​output results in sequence to obtain the feature processing result of the small target feature enhancement module.

[0054] In a possible implementation, the feature processing process of the small target feature enhancement module includes: when the step size is 1, splitting the feature information into channels to obtain feature information on both sides; after inputting the feature information on one side into the Mobile module, splicing it with the feature information on the other side; and sorting the channels of the spliced ​​output results to obtain the feature processing results of the small target feature enhancement module.

[0055] In a possible implementation, the image processing unit 410 is further configured to select part of the image data from the pest image dataset to generate a training dataset for training a small target pest detection model.

[0056] In a possible implementation, the image processing unit 410 is specifically configured to segment the pest image dataset according to an optimal segmentation ratio to generate a training dataset for training a small target pest detection model.

[0057] In a possible implementation, the model building unit 420 is specifically configured to construct a VoV-GSCSP module by combining and outputting a common convolution and two layers of GSConv convolution; and to add the VoV-GSCSP module to the neck portion of the YOLOv5s model.

[0058] In a possible implementation, the loss function for training the small target pest detection model is a NWD loss function.

[0059] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5 As shown, the electronic device may include: a processor 510, a communications interface 520, a memory 530 and a communication bus 540, wherein the processor 510, the communications interface 520 and the memory 530 communicate with each other via the communication bus 540. The processor 510 may call the logic instructions in the memory 530 to execute a lightweight small target pest detection method, which includes: Preprocess the collected small target pest images to construct a pest image dataset; Based on the YOLOv5s model, a small target pest detection model is constructed; a small target feature enhancement module is added to the trunk of the YOLOv5s model, and the small target feature enhancement module is constructed based on the MobileNetv3 and BiFomer structures; a VoV-GSCSP module is added to the neck of the YOLOv5s model; Training a small target pest detection model to obtain a trained small target pest detection model; The pest image dataset is input into the trained small target pest detection model to obtain the detection results of small target pest images.

[0060] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0061] On the other hand, the present invention further provides a computer program product, the computer program product comprising a computer program, the computer program can be stored on a non-transitory computer-readable storage medium, and when the computer program is executed by a processor, the computer can execute a lightweight small target pest detection method provided by the above methods, the method comprising: Preprocess the collected small target pest images to construct a pest image dataset; Based on the YOLOv5s model, a small target pest detection model is constructed; a small target feature enhancement module is added to the trunk of the YOLOv5s model, and the small target feature enhancement module is constructed based on the MobileNetv3 and BiFomer structures; a VoV-GSCSP module is added to the neck of the YOLOv5s model; Training a small target pest detection model to obtain a trained small target pest detection model; The pest image dataset is input into the trained small target pest detection model to obtain the detection results of small target pest images.

[0062] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to perform a lightweight small target pest detection method provided by the above methods, the method comprising: Preprocess the collected small target pest images to construct a pest image dataset; Based on the YOLOv5s3 model, a small target pest detection model is constructed; a small target feature enhancement module is added to the trunk of the YOLOv5s model, and the small target feature enhancement module is constructed based on the MobileNetv3 and BiFomer structures; a VoV-GSCSP module is added to the neck of the YOLOv5s model; Training a small target pest detection model to obtain a trained small target pest detection model; The pest image dataset is input into the trained small target pest detection model to obtain the detection results of small target pest images.

[0063] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0064] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A lightweight small target pest detection method, characterized in that: include: Preprocess the collected small target pest images to construct a pest image dataset; Based on the YOLOv5s model, a small target pest detection model is constructed; wherein a small target feature enhancement module is added to the trunk part of the YOLOv5s model, and the small target feature enhancement module is constructed based on the MobileNetv3 and BiFomer structures; a VoV-GSCSP module is added to the neck part of the YOLOv5s model; Training a small target pest detection model to obtain a trained small target pest detection model; The pest image dataset is input into the trained small target pest detection model to obtain the detection result of the small target pest image.

2. The lightweight small target pest detection method according to claim 1, characterized in that: The feature processing process of the small target feature enhancement module includes: When the step size is 2, the feature information is subjected to the first ordinary convolution and the first depth convolution respectively; After the first ordinary convolution, the second depth convolution and BiFormer module operations are performed in sequence; After performing common convolution on the outputs after the BiFormer module operation and the first depth convolution, the output results are spliced; The spliced ​​output results are sequentially subjected to a deep convolution module and a common convolution to obtain the feature processing result of the small target feature enhancement module.

3. The lightweight small target pest detection method according to claim 1, characterized in that: The feature processing process of the small target feature enhancement module includes: When the step size is 1, the feature information is split into channels to obtain feature information on both sides; After inputting the feature information on one side into the Mobile module, it is spliced ​​with the feature information on the other side; The spliced ​​output results are sorted by channels to obtain the feature processing results of the small target feature enhancement module.

4. The lightweight small target pest detection method according to claim 1, characterized in that: The VoV-GSCSP module is added to the neck part of the YOLOv5s model, specifically including: The VoV-GSCSP module is constructed by combining the outputs of ordinary convolution and two layers of GSConv convolution. The VoV-GSCSP module is added to the neck part of the YOLOv5s model.

5. The lightweight small target pest detection method according to claim 1, characterized in that: Also includes: The pest image data set is segmented according to an optimal segmentation ratio to generate a training data set for training a small target pest detection model.

6. The lightweight small target pest detection method according to any one of claims 1 to 5, characterized in that: The loss function for training the small target pest detection model is the NWD loss function.

7. A lightweight small target pest detection device, characterized in that: include: An image processing unit, used for preprocessing the collected small target pest images to construct a pest image data set; A model building unit is used to build a small target pest detection model based on the YOLOv5s model; wherein a small target feature enhancement module is added to the trunk part of the YOLOv5s model, and the small target feature enhancement module is built based on the MobileNetv3 and BiFomer structures; and a VoV-GSCSP module is added to the neck part of the YOLOv5s model; A model training unit, used for training a small target pest detection model to obtain a trained small target pest detection model; The model application unit is used to input the pest image data set into the trained small target pest detection model to obtain the detection result of the small target pest image.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the lightweight small target pest detection method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the lightweight small target pest detection method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the lightweight small target pest detection method according to any one of claims 1 to 6 is implemented.

Images