A Lightweight Wildfire Detection Model Design Method and System Based on Auxiliary Training Branch

Abstract

This invention belongs to the field of deep learning technology and discloses a lightweight wildfire detection model design method and system based on auxiliary training branches. Wildfire image samples are collected by patrol drones and monitoring pan-tilt units to construct a wildfire dataset. The collected wildfire dataset is augmented using the Mosaic-4 data augmentation method. The complete dataset is preprocessed using a histogram equalization algorithm, and the preprocessed image samples are labeled. The labeled image samples are proportionally divided into training and testing sets. A lightweight convolutional neural network main model structure is designed. The lightweight wildfire detection model, including the auxiliary training branch design, is trained using the wildfire dataset, ultimately obtaining a lightweight wildfire detection model. This invention combines lightweight convolutional network design and auxiliary training branch design methods, Mosaic-4 data augmentation, and histogram equalization algorithms to significantly improve detection speed while maintaining detection accuracy.

Description

Technical Field

[0001] This invention belongs to, but is not limited to, the field of deep learning technology, and particularly relates to a lightweight wildfire detection model design method and system based on auxiliary training branches. Background Technology

[0002] Wildfires are highly destructive natural disasters that destroy vast areas of forest and wildlife habitats, endangering human lives, property, and the environment. Monitoring pan-tilt units and inspection drones are currently the mainstream edge devices for wildfire monitoring, capable of detecting fire sources and promptly disseminating early warning information using visible light and thermal imaging technologies. With the widespread application of target detection algorithms in wildfire detection, the demand for deploying efficient and accurate wildfire detection models on edge devices is increasing. Currently, most widely used wildfire detection methods rely on complex, parameter-intensive deep learning models, which face challenges in deploying on edge devices due to limited computing resources and storage capacity. Therefore, designing a lightweight wildfire detection model has become an urgent need.

[0003] The existing methods for detecting wildfires include the following: Currently, wildfire detection methods mainly include multi-sensor fusion detection, satellite remote sensing detection, and computer vision detection. Among these, multi-sensor fusion detection has a small detection range and is difficult to maintain, while satellite remote sensing detection technology is susceptible to weather conditions and has high startup costs. Thanks to the rapid development of electronic and imaging technologies, deep learning-based wildfire detection methods have become a reality, providing a low-cost, safe, and convenient solution for wildfire monitoring and detection. This can be used not only for ground-based monitoring platforms but also for aerial monitoring via drones. Deep learning-based forest fire detection algorithms establish fire detection models by training on large amounts of data. These methods acquire real-time forest images through drone inspections and tower video surveillance, detecting forest fires by identifying flames and smoke in the images. They are characterized by flexible deployment and wide coverage, making them suitable for use in open, large-space forest environments.

[0004] However, traditional deep learning models require a large amount of computing resources to run on edge devices, resulting in slow response speed and low computational efficiency. Secondly, large models have a large number of parameters and computational load, which exceeds the computing power and storage capacity of edge devices, greatly limiting the devices on which the models can be deployed. Most importantly, wildfire detection usually requires a real-time and rapid response in order to take timely measures, but the complexity of traditional models often leads to high detection latency, which cannot meet the requirements of real-time wildfire detection. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a lightweight wildfire detection model design method and system based on auxiliary training branches.

[0006] This invention is implemented as follows: a lightweight wildfire detection model design method based on an auxiliary training branch, the method comprising: S1: Wildfire image samples are collected by inspection drones and monitoring PTZ cameras to construct a wildfire dataset.

[0007] S2: The Mosaic-4 data augmentation method was used on the collected wildfire dataset. This method involves taking four images and stitching them together into one image using random cropping and random arrangement. At the same time, some Gaussian noise was introduced to enhance the robustness of the model.

[0008] S3: The collected complete dataset is preprocessed using a histogram equalization algorithm, and the preprocessed image samples are labeled.

[0009] S4: Divide the labeled image samples into training and test sets according to the proportions.

[0010] S5: Design a lightweight convolutional neural network main model structure. The designed lightweight convolutional neural network model compresses the model structure size. The proposed lightweight detection head using shared convolution can achieve the sharing of some convolution parameters, which improves the detection efficiency of the model.

[0011] S6: Design an auxiliary training branch structure. The designed auxiliary training branch structure can aggregate gradient information containing all target objects and retain information of deep features, thereby obtaining reliable gradients and alleviating the information loss problem in deep supervision.

[0012] S7: Use the wildfire dataset to train a lightweight wildfire detection model that includes auxiliary training branches, remove the auxiliary training branch structure from the trained model, and finally obtain a lightweight wildfire detection model.

[0013] Furthermore, in S5, the structure of the designed lightweight aircraft door detection model includes: (1) The backbone network is constructed using the real-time target detection network YOLOv8n to improve the real-time detection speed of the model.

[0014] (2) A feature extraction and fusion module for the neck network is constructed using partial convolution operators to improve the computational efficiency of the model.

[0015] (3) A shared convolutional kernel is used to construct the output detection head. By sharing parameters between modules, the number of parameters of the detection head is reduced, the real-time detection speed of the model is improved, and the identification and location of wildfires are realized.

[0016] Furthermore, in S6, the structure of the designed auxiliary training branch includes: (1) A normalization layer containing convolutional layers, batch normalization layers and activation functions is used to normalize the original image and the feature map input through the backbone network, and the output of the convolutional layer is standardized to speed up the training and improve the stability of the model.

[0017] (2) A feature fusion module is constructed using deformable convolution operators. By merging feature maps at different levels, their complementary information is used to enhance the model's ability to identify targets.

[0018] (3) Pass the merged feature map to the output detection head to provide additional auxiliary information for wildfire detection during model training and improve the detection accuracy of the model.

[0019] Furthermore, in S7, the lightweight wildfire detection model based on the auxiliary training branch includes: (1) The main structure, consisting of a backbone network, a neck network and an output detection head, is responsible for processing and detecting the input video images to realize the identification and location of wildfires.

[0020] (2) The auxiliary training branch, consisting of the normalization layer and the feature fusion module, is responsible for providing additional auxiliary information during model training, assisting the main structure in learning the features of image samples, and improving the detection accuracy of the model.

[0021] (3) The auxiliary training branch will be removed after the model training is completed. The final model will only have the main structure consisting of the backbone network, the neck network and the output detection head. This will reduce the number of model parameters and computational load and improve the real-time detection speed of the model without reducing the detection accuracy of the model.

[0022] Another objective of this invention is to provide a lightweight wildfire detection model design system based on the aforementioned lightweight wildfire detection model design method based on auxiliary training branches. This system specifically includes: The dataset construction module uses wildfire image samples collected by inspection drones and monitoring PTZ cameras to build a wildfire dataset.

[0023] The data augmentation module, connected to the dataset construction module, uses the Mosaic-4 data augmentation method on the collected wildfire dataset. This method uses four images and stitches them together into one image using random cropping and random arrangement. At the same time, it introduces some Gaussian noise to enhance the robustness of the model.

[0024] The preprocessing module, connected to the data augmentation module, preprocesses the collected complete dataset using a histogram equalization algorithm and labels the preprocessed image samples.

[0025] The partitioning module, connected to the preprocessing module, divides the labeled image samples into training and test sets proportionally.

[0026] The structural design module, connected to the partitioning module, designs a lightweight convolutional neural network main model structure. The designed lightweight convolutional neural network model compresses the model structure size. The proposed lightweight detection head using shared convolution can achieve the sharing of some convolution parameters, improving the detection efficiency of the model.

[0027] The branch structure design module, connected to the structure design module, designs an auxiliary training branch structure. The designed auxiliary training branch structure can aggregate gradient information containing all target objects and retain information of deep features, thereby obtaining reliable gradients and alleviating the information loss problem in deep supervision.

[0028] The model training module, connected to the branch structure design module, uses the wildfire dataset to train a lightweight wildfire detection model containing auxiliary training branches. The auxiliary training branch structure in the trained model is then removed to obtain the final lightweight wildfire detection model.

[0029] Another objective of this invention is to provide a computer device, characterized in that the computer device includes a memory and a processor, the memory storing a computer program, and when the computer program is executed by the processor, causing the processor to perform the steps of the lightweight wildfire detection model design method based on auxiliary training branches.

[0030] Another objective of this invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the lightweight wildfire detection model design method based on auxiliary training branches.

[0031] Another objective of this invention is to provide an information data processing terminal, which is used to implement the lightweight wildfire detection model design system based on auxiliary training branches.

[0032] Based on the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by this invention are as follows: First, this invention designs a lightweight wildfire detection model based on auxiliary training branches. This model combines lightweight model design methods such as lightweight convolutional network design and auxiliary training branch design, and introduces Mosaic-4 data augmentation and histogram equalization algorithms to significantly improve detection speed while ensuring detection accuracy. This invention emphasizes the real-time performance of the model, enabling fast and real-time response of wildfire detection on edge devices to meet the real-time task requirements of wildfire detection, so that prevention and control departments can take timely measures. The lightweight model design method and auxiliary training branches adopted in this invention can greatly reduce the number of model parameters and computational load, enabling effective deployment on edge devices and ensuring efficient wildfire detection tasks on edge devices.

[0033] Secondly, with the increasing frequency of wildfires caused by global climate change, the market demand for efficient and real-time wildfire detection systems is constantly growing. This invention can meet this demand and has broad market prospects.

[0034] This invention can significantly reduce the number of model parameters and computational load, enabling it to run efficiently on resource-constrained edge devices. This opens up the possibility of widespread deployment of wildfire detection systems, especially in remote and resource-limited areas.

[0035] The lightweight design used in this invention reduces the hardware requirements and energy consumption of the model, allowing for deployment using low-cost, low-performance edge devices, thereby lowering the overall system cost. This has significant economic implications for large-scale deployment and long-term operation.

[0036] Currently, the industry both domestically and internationally mainly uses traditional models with a large number of parameters for wildfire detection model design. However, traditional models require a large amount of computing resources when running on edge devices, resulting in slow response speed and low computational efficiency. Secondly, the large number of parameters and computational load of large models exceeds the computing processing capacity and storage capacity of edge devices, which greatly limits the devices on which the models can be deployed. Most importantly, wildfire detection usually requires real-time and rapid response in order to take timely measures, but the complexity of traditional models often leads to high detection latency, which cannot meet the requirements of real-time wildfire detection.

[0037] This invention addresses the high computational and memory demands of traditional deep convolutional neural networks by employing a lightweight convolutional network design method. This significantly reduces the number of model parameters and computational load, enabling the model to run efficiently on resource-constrained edge devices.

[0038] This invention employs an auxiliary training branch training method, which uses deformable convolution operators with strong feature extraction capabilities on the auxiliary training branch and removes the auxiliary training branch after model training is completed. This reduces the number of model parameters and computational load, and improves the real-time detection speed of the model without reducing the detection accuracy. Attached Figure Description

[0039] Figure 1 This is a flowchart of the lightweight wildfire detection model design method based on auxiliary training branches provided in this embodiment of the invention.

[0040] Figure 2 This is a structural diagram of the lightweight wildfire detection model provided in this embodiment of the invention.

[0041] Figure 3 This is an auxiliary training branch structure diagram provided in an embodiment of the present invention.

[0042] Figure 4 This is a diagram showing the overall structure of the lightweight wildfire detection model based on auxiliary training branches provided in this embodiment of the invention.

[0043] Figure 5 This is a structural diagram of a lightweight wildfire detection model design system based on auxiliary training branches provided by an embodiment of the present invention.

[0044] Figure 6 This is a smoke detection effect diagram provided by an embodiment of the present invention.

[0045] Figure 7 This is a diagram illustrating the flame detection effect provided in an embodiment of the present invention.

[0046] In the diagram: 1. Dataset construction module; 2. Data augmentation module; 3. Preprocessing module; 4. Partitioning module; 5. Structure design module; 6. Branch structure design module; 7. Model training module. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0048] like Figure 1 As shown, this embodiment of the invention provides a lightweight wildfire detection model design method based on an auxiliary training branch, the method comprising: S1: Wildfire image samples are collected by inspection drones and monitoring PTZ cameras to construct a wildfire dataset.

[0049] S2: The Mosaic-4 data augmentation method was used on the collected wildfire dataset. This method involves taking four images and stitching them together into one image using random cropping and random arrangement. At the same time, some Gaussian noise was introduced to enhance the robustness of the model.

[0050] S3: The collected complete dataset is preprocessed using a histogram equalization algorithm, and the preprocessed image samples are labeled.

[0051] S4: Divide the labeled image samples into training and test sets according to the proportions.

[0052] S5: Design a lightweight convolutional neural network main model structure. The designed lightweight convolutional neural network model compresses the model structure size. The proposed lightweight detection head using shared convolution can achieve the sharing of some convolution parameters, which improves the detection efficiency of the model.

[0053] S6: Design an auxiliary training branch structure. The designed auxiliary training branch structure can aggregate gradient information containing all target objects and retain information of deep features, thereby obtaining reliable gradients and alleviating the information loss problem in deep supervision.

[0054] S7: Use the wildfire dataset to train a lightweight wildfire detection model that includes auxiliary training branches, remove the auxiliary training branch structure from the trained model, and finally obtain a lightweight wildfire detection model.

[0055] This invention provides a lightweight wildfire detection model design method based on an auxiliary training branch. Through the collaborative work of multiple steps, it achieves efficient and accurate wildfire detection. First, in step S1, wildfire image samples are collected using patrol drones and monitoring pan-tilt-zoom (PTZ) cameras, constructing a wildfire dataset covering various scenarios and fire states. These collected and organized sample data provide sufficient diversity and representativeness, offering a solid data foundation for subsequent model training and detection.

[0056] Next, in step S2, the collected wildfire dataset is augmented using Mosaic-4 data augmentation. The Mosaic-4 method enhances the robustness of the model by randomly cropping and stitching four images into a single image and introducing Gaussian noise. This step effectively increases the diversity and complexity of the data, enabling the model to more stably identify fire characteristics in real-world wildfire scenarios and reducing detection errors caused by noise and environmental changes.

[0057] In step S3, the dataset undergoes preprocessing, including using a histogram equalization algorithm to balance image contrast and brightness, thereby improving image clarity and detail. Subsequently, the preprocessed image samples are labeled to ensure that each sample in the dataset possesses clear target information. The labeled data samples are further divided into training and testing sets (step S4) to ensure that the model training and testing process is sufficiently representative and feasible.

[0058] In step S5, a lightweight convolutional neural network main model was designed. By compressing the size of the model structure, the computational load was reduced, and the detection speed of the model was improved. In particular, the lightweight detection head using shared convolutions effectively improved the model's detection efficiency by sharing some convolution parameters, ensuring that the detection accuracy of the model was maintained while reducing the number of parameters. This design is especially important in resource-constrained environments, such as in drones or other embedded systems.

[0059] Next, in step S6, an auxiliary training branch structure was designed. This structure was introduced to aggregate gradient information containing all target objects during model training, while retaining deep feature information, ensuring the model obtains reliable gradient signals during training. The auxiliary training branch can alleviate the information loss problem in deep supervision, thereby improving the model's training effect and detection accuracy. After training (step S7), the final model removes the auxiliary training branch, retaining the lightweight detection model. This design ensures the model's lightweight and efficiency, providing an efficient and accurate solution for wildfire detection in practical applications.

[0060] like Figure 2 As shown, the structure of the lightweight aircraft door inspection model designed in S5 includes: (1) The backbone network is constructed using the real-time target detection network YOLOv8n to improve the real-time detection speed of the model.

[0061] (2) A feature extraction and fusion module for the neck network is constructed using partial convolution operators to improve the computational efficiency of the model.

[0062] (3) A shared convolutional kernel is used to construct the output detection head. By sharing parameters between modules, the number of parameters of the detection head is reduced, the real-time detection speed of the model is improved, and the identification and location of wildfires are realized.

[0063] like Figure 3 As shown, the structure of the auxiliary training branch designed in S6 includes: (1) A normalization layer containing convolutional layers, batch normalization layers and activation functions is used to normalize the original image and the feature map input through the backbone network, and the output of the convolutional layer is standardized to speed up the training and improve the stability of the model.

[0064] (2) A feature fusion module is constructed using deformable convolution operators. By merging feature maps at different levels, their complementary information is used to enhance the model's ability to identify targets.

[0065] (3) Pass the merged feature map to the output detection head to provide additional auxiliary information for wildfire detection during model training and improve the detection accuracy of the model.

[0066] like Figure 4 As shown, in S7, the lightweight wildfire detection model based on the auxiliary training branch includes: (1) The main structure, consisting of a backbone network, a neck network and an output detection head, is responsible for processing and detecting the input video images to realize the identification and location of wildfires.

[0067] (2) The auxiliary training branch, consisting of the normalization layer and the feature fusion module, is responsible for providing additional auxiliary information during model training, assisting the main structure in learning the features of image samples, and improving the detection accuracy of the model.

[0068] (3) The auxiliary training branch will be removed after the model training is completed. The final model will only have the main structure consisting of the backbone network, the neck network and the output detection head. This will reduce the number of model parameters and computational load and improve the real-time detection speed of the model without reducing the detection accuracy of the model.

[0069] like Figure 5 As shown, this embodiment of the invention provides a lightweight wildfire detection model design system based on the aforementioned lightweight wildfire detection model design method based on auxiliary training branches. The system specifically includes: Dataset construction module 1 uses inspection drones and monitoring PTZ cameras to collect wildfire image samples and construct a wildfire dataset.

[0070] Data augmentation module 2, connected to dataset construction module 1, uses the Mosaic-4 data augmentation method on the collected wildfire dataset. This method uses four images and stitches them together into one image by random cropping and random arrangement, while also introducing some Gaussian noise to enhance the robustness of the model.

[0071] Preprocessing module 3, connected to data augmentation module 2, preprocesses the collected complete dataset using a histogram equalization algorithm and labels the preprocessed image samples.

[0072] The partitioning module 4, connected to the preprocessing module 3, divides the labeled image samples into training and test sets according to a certain ratio.

[0073] The structural design module 5, connected to the partitioning module 4, designs a lightweight convolutional neural network main model structure. The designed lightweight convolutional neural network model compresses the model structure size. The proposed lightweight detection head using shared convolution can share some convolution parameters, improving the detection efficiency of the model.

[0074] Branch structure design module 6, connected to structure design module 5, designs auxiliary training branch structure. The designed auxiliary training branch structure can aggregate gradient information containing all target objects and retain information of deep features, thereby obtaining reliable gradients and alleviating the information loss problem in deep supervision.

[0075] Model training module 7, connected to branch structure design module 6, uses the wildfire dataset to train a lightweight wildfire detection model containing auxiliary training branches. The auxiliary training branch structure in the trained model is then removed to obtain the lightweight wildfire detection model.

[0076] This invention provides a computer device, characterized in that the computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the lightweight wildfire detection model design method based on auxiliary training branches.

[0077] This invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the lightweight wildfire detection model design method based on auxiliary training branches.

[0078] This invention provides an information data processing terminal, which is used to implement the lightweight wildfire detection model design system based on auxiliary training branches.

[0079] Evidence related to the technical effects obtained by the embodiments of the present invention.

[0080] The performance of the lightweight wildfire detection model based on auxiliary training branches designed in this invention is as follows: (1) The number of parameters is 2.3M, which is 23% less than the original model (3.0M).

[0081] (2) The computational cost is 6.3G, which is 22% less than the original model (8.1G).

[0082] (3) The average recognition accuracy was 77%, which was 5% higher than the original model (72%).

[0083] (4) The detection speed is 80 frames, which is 29% faster than the original model (57 frames).

[0084] The model's detection performance is as follows Figure 6 and Figure 7 As shown, where Figure 6 For better smoke detection performance, Figure 7 This is for flame detection results.

[0085] It should be noted that embodiments of the present invention can be implemented in hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented by hardware circuitry such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., or by software executed by various types of processors, or by a combination of the above-described hardware circuitry and software, such as firmware.

[0086] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A lightweight wildfire detection model design method based on auxiliary training branches, characterized in that, The method includes: S1: Wildfire image samples are collected by inspection drones and monitoring PTZ cameras to construct a wildfire dataset; S2: The Mosaic-4 data augmentation method is used on the collected wildfire dataset. This involves taking four images and stitching them together into one image using random cropping and random arrangement, while also introducing some Gaussian noise appropriately. S3: The collected complete dataset is preprocessed using a histogram equalization algorithm, and the preprocessed image samples are labeled. S4: Divide the labeled image samples into training and test sets according to the proportions; S5: Design a lightweight convolutional neural network main model structure. The designed lightweight convolutional neural network model compresses the model structure size. The lightweight detection head using shared convolution achieves the sharing of some convolution parameters. S6: Design an auxiliary training branch structure. The designed auxiliary training branch structure aggregates gradient information of all target objects and retains information of deep features, thereby obtaining reliable gradients and alleviating the information loss problem in deep supervision. S7: Use the wildfire dataset to train a lightweight wildfire detection model that includes auxiliary training branches, remove the auxiliary training branch structure from the trained model, and finally obtain a lightweight wildfire detection model. In S5, the design of the lightweight convolutional neural network main model structure includes: (1) The backbone network is constructed using the real-time target detection network YOLOv8n to improve the real-time detection speed of the model; (2) A feature extraction and fusion module for the neck network is constructed using partial convolution operators to improve the computational efficiency of the model; (3) A shared convolutional kernel is used to construct the output detection head. By sharing parameters between modules, the number of parameters of the detection head is reduced, the real-time detection speed of the model is improved, and the identification and location of wildfires are realized. In S6, the structure of the designed auxiliary training branch includes: (1) A normalization layer containing convolutional layers, batch normalization layers and activation functions is used to normalize the original image and the feature map input through the backbone network, and the output of the convolutional layer is standardized to speed up the training and improve the stability of the model. (2) A feature fusion module is constructed using deformable convolution operators. By merging feature maps at different levels, their complementary information is used to enhance the model's ability to identify targets. (3) Pass the merged feature map to the output detection head to provide additional auxiliary information for wildfire detection during model training and improve the detection accuracy of the model; In S7, the lightweight wildfire detection model based on the auxiliary training branch includes: (1) The main structure, consisting of a backbone network, a neck network and an output detection head, is responsible for processing and detecting the input video images to realize the identification and location of wildfires; (2) The auxiliary training branch, consisting of the normalization layer and the feature fusion module, is responsible for providing additional auxiliary information during model training, assisting the main structure in learning the features of image samples, and improving the detection accuracy of the model. (3) The auxiliary training branch will be removed after the model training is completed. The final model will only have the main structure consisting of the backbone network, the neck network and the output detection head. This will reduce the number of model parameters and computational load and improve the real-time detection speed of the model without reducing the detection accuracy of the model.

2. A lightweight wildfire detection model design system based on auxiliary training branches, based on the lightweight wildfire detection model design method described in claim 1, characterized in that, The system specifically includes: The dataset construction module uses wildfire image samples collected by inspection drones and monitoring PTZ cameras to build a wildfire dataset. The data augmentation module is connected to the dataset construction module. It uses the Mosaic-4 data augmentation method on the collected wildfire dataset, which uses four images and stitches them together into one image by random cropping and random arrangement, while reasonably introducing some Gaussian noise. The preprocessing module, connected to the data augmentation module, preprocesses the collected complete dataset using a histogram equalization algorithm and labels the preprocessed image samples. The partitioning module, connected to the preprocessing module, divides the labeled image samples into training and test sets according to a certain ratio. The structural design module, connected to the partitioning module, designs a lightweight convolutional neural network main model structure. The designed lightweight convolutional neural network model compresses the model structure size. The lightweight detection head using shared convolutions achieves the sharing of some convolution parameters. A branch structure design module, connected to the structure design module, designs an auxiliary training branch structure. The designed auxiliary training branch structure aggregates gradient information of all target objects and retains information of deep features, thereby obtaining reliable gradients and alleviating the information loss problem in deep supervision. The model training module, connected to the branch structure design module, uses the wildfire dataset to train a lightweight wildfire detection model containing auxiliary training branches. The auxiliary training branch structure in the trained model is then removed to obtain the final lightweight wildfire detection model.

3. A computer device, characterized in that, The computer device includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, causes the processor to perform the steps of the lightweight wildfire detection model design method based on auxiliary training branches as described in claim 1.

4. A computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the lightweight wildfire detection model design method based on auxiliary training branches as described in claim 1.

5. An information data processing terminal, characterized in that, The information data processing terminal is used to implement the steps of the lightweight wildfire detection model design method based on auxiliary training branches as described in claim 1.

Images