An intelligent data processing method for forestry ecological monitoring

Abstract

The present invention discloses an intelligent data processing method for forestry ecological monitoring, which relates to the technical field of forestry ecological monitoring. The method includes S1, acquiring multi-source image data of a forest area, S2, preprocessing the image data based on an environmental perception model, and correcting image overexposure or underexposure caused by smoke and fire, S3, extracting features from the image data, S4, analyzing the features using a deep learning model, and identifying potential fire risks, S5, generating corresponding early warning information and sending it to a monitoring center in real time, and storing the processing results and early warning information; the intelligent data processing method for forestry ecological monitoring solves the problem of image overexposure or underexposure caused by the high dynamic range of smoke and fire in forest fire monitoring in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of forestry ecological monitoring, and in particular to an intelligent data processing method for forestry ecological monitoring. Background Art

[0002] Forest fire monitoring is a crucial component of modern forestry ecological monitoring. Existing monitoring methods typically rely on image sensors and related algorithms to achieve early detection and dynamic monitoring of fire outbreaks. However, due to the complexity of the monitoring environment and the characteristics of fires, existing technologies have significant limitations when processing high dynamic range images.

[0003] Specifically, the presence of smoke and flames during forest fire monitoring prevents image sensors from properly balancing light intensity, leading to over- or underexposure of images. This phenomenon makes it difficult for monitoring systems to accurately extract key fire characteristics, which in turn affects fire identification accuracy and monitoring efficiency. Furthermore, traditional image processing algorithms are generally less adaptable to dynamic lighting changes and have limited processing capabilities in complex lighting environments, making them incapable of meeting the needs of real-time monitoring. Furthermore, existing solutions to image overexposure and underexposure often rely on hardware upgrades, significantly increasing the cost of monitoring systems and limiting their promotion and application in resource-limited areas. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent data processing method for forestry ecological monitoring to solve the problem of overexposure or underexposure of images caused by the high dynamic range of smoke and fire light in the existing technology in forest fire monitoring.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent data processing method for forestry ecological monitoring, the method comprising:

[0006] S1. Obtain multi-source image data of the forest area;

[0007] S2. Preprocess the image data based on the environmental perception model to correct for overexposure or underexposure caused by smoke and fire. This includes analyzing the brightness distribution of smoke and fire areas in the image, determining the initial brightness value, performing contrast enhancement on the brightness, and compressing the brightness of the overexposed areas. The specific formula for brightness contrast enhancement is: C = k[A] n ;

[0008] Where C represents the enhanced contrast, k represents the contrast adjustment ratio parameter, A represents the initial brightness value of the image, and n represents the exponential parameter of the adjustment intensity;

[0009] S3, extracting features from image data;

[0010] S4. Use deep learning models to analyze features and identify potential fire risks, including simulating the error dynamics of the deep learning model and optimizing the analysis process. The specific formula is:

[0011] dN / dt=rN-aNP, dP / dt=bNP-mP;

[0012] Where N represents the number of fire risk characteristics, P represents the number of misclassifications, r represents the growth rate of fire risk characteristics, a represents the suppression rate of misclassification on fire characteristics, b represents the rate at which fire characteristics promote misclassification, m represents the natural reduction rate of misclassification, and t represents time.

[0013] Output fire risk level and risk area location;

[0014] S5. Generate corresponding warning information and send it to the monitoring center in real time, and store the processing results and warning information.

[0015] Preferably, the S1 includes:

[0016] Collect multi-source image data, including drone images, satellite remote sensing images, and ground monitoring data. Calculate the value index of each image source based on its resolution and timeliness. Use the resource allocation model to determine the weight of each data source and generate a multi-source comprehensive image. The specific formula is:

[0017] Among them, R i Indicates the weight of the i-th image source in multi-source image data fusion, V i represents the value index of the i-th data source, g represents the number of image data sources used in forest monitoring, V j It represents the value index of the j-th data source, i represents the number of the current image data source, and j represents a number among all image data sources.

[0018] Preferably, the S3 includes:

[0019] The gradient value of the image is input into the fitness function to generate the feature fitness distribution. The flame edge and smoke contour area are located according to the output of the fitness function, and the shape, size and concentration characteristics of the area are extracted. The specific formula of the fitness function is:

[0020]

[0021] Among them, F(x) represents feature fitness, x represents the image gradient value, and e represents a natural constant.

[0022] Preferably, the S5 includes:

[0023] Generate early warning information based on the fire analysis results, encrypt the early warning information, send the encrypted information to the monitoring center through the wireless communication network, and store it in the cloud for subsequent review and analysis. The specific formula for encrypting the early warning information is: D = M ⊕ K;

[0024] Among them, D represents the ciphertext information generated after the encryption operation, M represents the unencrypted warning information, K represents the key used in the encryption and decryption process, and ⊕ represents the bitwise exclusive OR operation.

[0025] Preferably, the S2 further includes performing image data correction based on a local histogram, applying a grayscale stretching algorithm to smoke and fire areas, optimizing image quality using an adaptive contrast enhancement method, and removing image noise through a noise filter.

[0026] Preferably, the image data correction based on the local histogram includes dividing the image data into a plurality of non-overlapping sub-regions, calculating a histogram for each sub-region, and applying a histogram equalization formula H(i)=G×(n1+n2+...+n i ) / E, where H(i) represents the cumulative histogram, G is a constant, and n1 to n i Represents the number of pixels in the histogram, E is the total number of pixels, and the sub-regions after histogram equalization are merged into the complete image data.

[0027] Preferably, the calculation of the histogram for each sub-region includes extracting the pixel value range [a, b] of the sub-region, calculating the frequency of each pixel value, and mapping the pixel value range [a, b] to [0, 255] using linear interpolation. The specific formula is: f(x) = 255 × (y-a) / (b-a), where f(x) is the mapped pixel value, y is the original pixel value, and a and b are the minimum and maximum pixel values, respectively.

[0028] Preferably, the image data in S1 includes multispectral images, thermal infrared images and visible light images.

[0029] Preferably, the features in the image data in S3 include texture features, shape features and spectral features.

[0030] Preferably, the warning information generated in S5 includes the fire level and the location of the risk area.

[0031] It can be seen from the above technical solution that the present invention has the following beneficial effects:

[0032] This intelligent data processing method for forestry ecological monitoring obtains multi-source image data of the forest area, pre-processes the image data based on the environmental perception model, corrects image overexposure or underexposure caused by smoke and fire, extracts features in the image data, analyzes the features using a deep learning model, identifies potential fire risks, generates corresponding warning information and sends it to the monitoring center in real time, stores the processing results and warning information, and can achieve more accurate fire feature extraction in complex lighting environments, improve the recognition accuracy of the fire monitoring system, and the improved image processing technology can quickly respond to dynamic lighting changes, adapt to the changing monitoring environment of forest fires, meet real-time monitoring needs, reduce the hardware requirements and operating costs of the system, and make it easier to promote and apply in areas with limited resources. It can output optimized monitoring images in real time, effectively reduce data processing delays, and improve the overall efficiency of fire monitoring. It solves the problem of image overexposure or underexposure caused by the high dynamic range of smoke and fire in forest fire monitoring in the existing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0034] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only 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 making creative efforts are within the scope of protection of the present invention.

[0035] like Figure 1 As shown, the present invention provides a technical solution: an intelligent data processing method for forestry ecological monitoring, the method comprising:

[0036] S1. Obtain multi-source image data of the forest area;

[0037] S2. Preprocess the image data based on the environmental perception model to correct for overexposure or underexposure caused by smoke and fire. This includes analyzing the brightness distribution of smoke and fire areas in the image, determining the initial brightness value, performing contrast enhancement on the brightness, and compressing the brightness of the overexposed areas. The specific formula for brightness contrast enhancement is: C = k[A] n ;

[0038] Where C represents the enhanced contrast, k represents the contrast adjustment ratio parameter, A represents the initial brightness value of the image, and n represents the exponential parameter of the adjustment intensity;

[0039] S3, extracting features from image data;

[0040] S4. Use deep learning models to analyze features and identify potential fire risks, including simulating the error dynamics of the deep learning model and optimizing the analysis process. The specific formula is:

[0041] dN / dt=rN-aNP, dP / dt=bNP-mP;

[0042] Where N represents the number of fire risk characteristics, P represents the number of misclassifications, r represents the growth rate of fire risk characteristics, a represents the suppression rate of misclassification on fire characteristics, b represents the rate at which fire characteristics promote misclassification, m represents the natural reduction rate of misclassification, and t represents time.

[0043] Output fire risk level and risk area location;

[0044] S5. Generate corresponding warning information and send it to the monitoring center in real time, and store the processing results and warning information.

[0045] The core of this implementation method is to achieve intelligent monitoring of forest fire risks through the acquisition, preprocessing, feature extraction and deep learning analysis of multi-source image data. Through image correction technology based on environmental perception models, the image distortion problem caused by smoke and fire is solved, ensuring the accuracy of subsequent feature extraction. Further, through the error dynamic suppression mechanism of the deep learning model, combined with mathematical formulas to optimize the analysis process, the fire risk characteristics can be accurately identified, and the risk level and risk area location can be predicted. Finally, combined with the generated warning information, real-time feedback is provided to the monitoring center to achieve risk monitoring and decision support. Through contrast enhancement and deep learning analysis, the recognition accuracy and sensitivity of fire risk characteristics are improved, ensuring the reliability of the prediction results. The optimized dynamic suppression mechanism shortens the time for risk feature analysis and warning generation, improves data processing efficiency, generates and transmits warning information in real time, effectively reduces forest fire risks, enhances the protection capabilities of the ecological monitoring system, reduces the ecological damage caused by fire, and provides technical support for the sustainable management of forest resources.

[0046] S1 involves collecting multi-source image data, including drone images, satellite remote sensing images, and ground monitoring data. The value index of each image source is calculated based on its resolution and timeliness. The resource allocation model is used to determine the weight of each data source to generate a multi-source integrated image. The specific formula is:

[0047] Among them, R i In the fusion of multi-source image data, the weight of the i-th image source, V i represents the value index of the i-th data source, g represents the number of image data sources used in forest monitoring, V jIt represents the value index of the j-th data source, i represents the number of the current image data source, and j represents a number among all image data sources.

[0048] In this implementation, the fusion and optimization of multi-source images are achieved through the comprehensive utilization of drone images, satellite remote sensing images and ground monitoring data. Each image data source has different resolutions and timeliness, so the value index V is introduced for quantitative evaluation. According to the resource allocation model, the weight formula R i A weighted calculation is performed on each image source to generate a composite image. This method effectively integrates the information advantages of multiple image data sources, providing more accurate and real-time forestry ecological monitoring data. By allocating weights based on value indicators, the advantages of different image sources are maximized, significantly improving the overall quality and accuracy of image data. Combining drone imagery, satellite remote sensing imagery, and ground monitoring data, it achieves multidimensional monitoring capabilities across large forest areas. The introduction of timeliness indicators ensures the priority of key data sources, optimizes real-time requirements, and enables rapid response to ecological changes in forest areas. The resource allocation model is applicable to different forest monitoring scenarios, and the allocation strategy can be flexibly adjusted according to specific needs.

[0049] S3 includes inputting the gradient value of the image into the fitness function to generate a feature fitness distribution, locating the flame edge and smoke contour area according to the output of the fitness function, and extracting the shape, size and concentration characteristics of the area. The specific formula of the fitness function is:

[0050]

[0051] Among them, F(x) represents feature fitness, x represents the image gradient value, and e represents a natural constant.

[0052] This embodiment calculates the feature fitness by extracting the image gradient value and inputting the fitness function F(x). The fitness function can effectively distinguish the flame edge from the smoke contour area. Specifically, the larger the gradient value x, the closer F(x) is to 1, indicating that the regional features are more significant; the smaller the gradient value, the closer F(x) is to 0, indicating that the regional features are not significant. Using this fitness function, the shape, size and waviness characteristics of the flame and smoke areas can be accurately extracted, thereby providing reliable data support for subsequent fire risk analysis. Through the fitness function based on the gradient value, the positioning accuracy of the flame edge and smoke contour is improved, and the influence of noise interference on feature extraction is avoided. The fitness function is simple in form and low in computational cost. It is suitable for real-time processing of large-scale forest monitoring data. By extracting the shape, size and waviness characteristics, it provides multi-dimensional reliable data support for fire risk assessment. The fitness function can adjust the input parameters according to different application scenarios to ensure that the method has wide adaptability.

[0053] S5 includes generating early warning information based on the fire analysis results, encrypting the early warning information, sending the encrypted information to the monitoring center through the wireless communication network, and storing it in the cloud for subsequent review and analysis. The specific formula for encrypting the early warning information is: D = M ⊕ K;

[0054] Among them, D represents the ciphertext information generated after the encryption operation, M represents the unencrypted warning information, K represents the key used in the encryption and decryption process, and ⊕ represents the bitwise exclusive OR operation.

[0055] In this embodiment, the warning information generated by the fire analysis results is processed as content to be sent. In order to ensure the security and privacy of the warning information, a bitwise XOR operation is used for encryption. The unencrypted warning information M is subjected to a bitwise XOR operation with the key K to generate a ciphertext information D. This ciphertext information is sent to the monitoring center via the wireless communication network and stored in the cloud at the same time. The encryption method is based on a simple and efficient XOR operation, which ensures that the original information can be restored only when the key is known, thereby improving the security of data transmission and storage. The confidentiality of the warning information is protected by an encryption mechanism to avoid potential risks caused by information leakage during wireless network transmission. The encrypted information is sent to the monitoring center in real time and stored in the cloud to ensure that the warning information can respond immediately and provide reliable data support for subsequent analysis. The bitwise XOR operation encryption has high computational efficiency and is suitable for the encryption needs of large-scale warning information. The key K can be dynamically updated to further improve the system's anti-attack capabilities and adapt to different communication and storage security requirements.

[0056] S2 also includes image data correction based on local histograms, applying a grayscale stretching algorithm to smoke and fire areas, optimizing image quality using an adaptive contrast enhancement method, and removing image noise through a noise filter.

[0057] When correcting image data, this embodiment first analyzes the grayscale distribution characteristics of the image based on the local histogram to correct the image quality degradation caused by local overexposure or underexposure. For smoke and fire areas, the grayscale stretching algorithm is applied to expand the grayscale range and enhance the detail expression of the image. Subsequently, the adaptive contrast enhancement method is used to improve the clarity and layering of the image by adjusting the contrast differences between different areas. Finally, the high-frequency noise and artifacts in the image are removed by the noise filter to ensure that the image data quality is suitable for subsequent feature extraction and risk analysis processing. The brightness uniformity and detail expression of the image are improved, and the visualization effect of the smoke and fire areas is optimized. The application of the adaptive contrast enhancement method improves the layering of the image and provides a clearer data basis for the subsequent extraction of fire features. The use of the noise filter effectively removes random noise and artifacts in the image, ensuring the stability and reliability of the image data. This method can flexibly respond to image data under different lighting conditions and imaging equipment, and is suitable for a variety of monitoring environments.

[0058] Image data correction based on local histogram includes dividing the image data into multiple non-overlapping sub-regions, calculating the histogram for each sub-region, and applying the histogram equalization formula H(i)=G×(n1+n2+...+n i ) / E, where H(i) represents the cumulative histogram, G is a constant, and n1 to n i Represents the number of pixels in the histogram, E is the total number of pixels, and the sub-regions after histogram equalization are merged into the complete image data.

[0059] In this embodiment, when correcting image data based on local histograms, the entire image is first divided into several non-overlapping subregions, each of which is processed independently. By calculating the histogram of each subregion and applying the histogram equalization formula, the grayscale distribution is adjusted to enhance the contrast of the subregion. In this formula, the cumulative histogram H(i) is calculated based on the cumulative pixel ratio, and the adjusted grayscale is mapped to the target range using the constant G, thereby improving the local image quality. After processing, all equalized subregions are re-merged into the complete image data to ensure global brightness and detail consistency. By processing the image subregions separately, the grayscale contrast of the local regions is significantly enhanced, adapting to the brightness unevenness existing in the image. The equalized image is fully optimized in terms of brightness distribution and detail expression, laying a good foundation for subsequent feature extraction and analysis. The image division and processing are highly flexible, and the subregion size can be adjusted according to the image resolution or content complexity to meet different correction requirements. This method can be widely used for image correction in complex lighting environments, such as forestry monitoring data processing interference from heavy smoke and fire.

[0060] Calculating the histogram for each sub-region includes extracting the pixel value range [a, b] of the sub-region, calculating the frequency of each pixel value, and mapping the pixel value range [a, b] to [0, 255] using linear interpolation. The specific formula is: f(x) = 255 × (y-a) / (b-a), where f(x) is the mapped pixel value, y is the original pixel value, and a and b are the minimum and maximum pixel values, respectively.

[0061] In this embodiment, to calculate the histogram for subregions of image data, the pixel value range [a, b] is first extracted from each subregion, specifically the minimum a and maximum b of the pixel values ​​within the region are determined. Next, a histogram is generated by statistically analyzing the frequency distribution of each pixel value. To enhance image contrast, linear interpolation is used to map the original pixel value y to the target range [0, 255]. In the formula, the mapped pixel value f(x) is calculated based on the relative position of the pixel values, thereby achieving a linear stretch of the grayscale. This method effectively balances the grayscale distribution of the subregions, improving image contrast and visualization. By mapping the pixel values ​​of the subregions to the full grayscale range [0, 255], the uniformity and contrast of the grayscale distribution are enhanced, making details easier to discern. Linear interpolation is computationally simple and can quickly process large-scale image data, meeting the needs of real-time monitoring. The grayscale mapping is dynamically adjusted based on the pixel value range of each subregion, adapting to image correction requirements under varying lighting conditions. The mapped grayscale value distribution is more suitable for subsequent feature extraction and analysis algorithms, improving the accuracy and reliability of the overall monitoring system.

[0062] The image data in S1 includes multispectral images, thermal infrared images and visible light images. In this embodiment, the acquisition of forest image data includes three different types of images: multispectral images, thermal infrared images and visible light images. Multispectral images: By capturing information from multiple spectral bands, they can reflect the spectral characteristics of vegetation in the forest area, help identify different vegetation types, monitor health status, and detect abnormal changes in vegetation. Thermal infrared images: used to perceive surface temperature distribution, especially in fire monitoring, can directly capture information in high-temperature areas, and provide reliable data for locating fire hotspots. Visible light images: capture visual information of the forest area, provide clear and intuitive images, and provide background and reference for other image types. The combination of the three image data not only provides rich multi-dimensional information, but also enhances the accuracy of fire monitoring and ecological assessment through comparison and fusion.

[0063] The features in the image data in S3 include texture features, shape features and spectral features. In this embodiment, the feature extraction of image data includes three main features: texture features, shape features and spectral features. Texture features: By calculating the distribution and change pattern of pixel values ​​in the image, the roughness, directionality and repeatability information of the area are obtained. For example, the use of gray-level co-occurrence matrix (GLCM) or wavelet transform to extract texture features can identify the fine structure of vegetation or smoke areas. Shape features: Analyze the geometric contours and spatial distribution information in the image, such as the shape of the flame edge or smoke contour, and identify the geometric characteristics of the target by calculating parameters such as perimeter, area, aspect ratio, etc. Spectral features: Based on the reflectance data of different bands in multispectral or hyperspectral images, the spectral response information of the target is extracted for analyzing vegetation health, smoke density and flame characteristics. These three features provide multi-dimensional support for fire risk assessment and ecological status monitoring through joint analysis, which helps to improve the accuracy and stability of the monitoring system.

[0064] The warning information generated in S5 includes the fire level and the location of the risk area. This embodiment generates warning information by integrating the results of the fire risk analysis, which includes two key contents: fire level and risk area location: Fire level: Based on the features extracted during the fire analysis process (such as flame intensity, smoke density, regional temperature, etc.), combined with the model calculation results, the severity of the fire is evaluated. Fire levels are usually divided into different levels (such as level I, level II, and level III), and each level corresponds to a different fire intensity and degree of hazard, providing a clear reference for the monitoring center. Risk area location: By analyzing the geographical distribution information of fire features in the image, the specific area where the fire occurs is located. Using a geographic coordinate system (such as GPS or geographic registration of remote sensing images), the boundaries of the risk area are marked to provide accurate location data for emergency response and resource deployment. The generated warning information is displayed in digital form and can be transmitted to the monitoring center in real time through a wireless communication network and stored in the cloud for subsequent analysis and scheduling decisions.

[0065] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent data processing method for forestry ecological monitoring, characterized in that: The method comprises: S1. Obtain multi-source image data of the forest area; S2. Preprocess the image data based on the environmental perception model to correct for overexposure or underexposure caused by smoke and fire. This includes analyzing the brightness distribution of smoke and fire areas in the image, determining the initial brightness value, performing contrast enhancement on the brightness, and compressing the brightness of the overexposed areas. The specific formula for brightness contrast enhancement is: C = k[A] n ; Where C represents the enhanced contrast, k represents the contrast adjustment ratio parameter, A represents the initial brightness value of the image, and n represents the exponential parameter of the adjustment intensity; S3, extracting features from image data; S4. Use deep learning models to analyze features and identify potential fire risks, including simulating the error dynamics of the deep learning model and optimizing the analysis process. The specific formula is: dN / dt=rN-aNP, dP / dt=bNP-mP; Where N represents the number of fire risk characteristics, P represents the number of misclassifications, r represents the growth rate of fire risk characteristics, a represents the suppression rate of misclassification on fire characteristics, b represents the rate at which fire characteristics promote misclassification, m represents the natural reduction rate of misclassification, and t represents time. Output fire risk level and risk area location; S5. Generate corresponding warning information and send it to the monitoring center in real time, and store the processing results and warning information.

2. The intelligent data processing method for forestry ecological monitoring according to claim 1, characterized in that: Said S1 comprises: Collect multi-source image data, including drone images, satellite remote sensing images, and ground monitoring data. Calculate the value index of each image source based on its resolution and timeliness. Use the resource allocation model to determine the weight of each data source and generate a multi-source comprehensive image. The specific formula is: Among them, R i In the fusion of multi-source image data, the weight of the i-th image source, V i represents the value index of the i-th data source, g represents the number of image data sources used in forest monitoring, V j It represents the value index of the j-th data source, i represents the number of the current image data source, and j represents a number among all image data sources.

3. The intelligent data processing method for forestry ecological monitoring according to claim 1, characterized in that: The S3 includes: The gradient value of the image is input into the fitness function to generate the feature fitness distribution. The flame edge and smoke contour area are located according to the output of the fitness function, and the shape, size and concentration characteristics of the area are extracted. The specific formula of the fitness function is: Among them, F(x) represents feature fitness, x represents the image gradient value, and e represents a natural constant.

4. The intelligent data processing method for forestry ecological monitoring according to claim 1, characterized in that: The S5 includes: Generate early warning information based on the fire analysis results, encrypt the early warning information, send the encrypted information to the monitoring center through the wireless communication network, and store it in the cloud for subsequent review and analysis. The specific formula for encrypting the early warning information is: D = M ⊕ K; Among them, D represents the ciphertext information generated after the encryption operation, M represents the unencrypted warning information, K represents the key used in the encryption and decryption process, and ⊕ represents the bitwise exclusive OR operation.

5. The intelligent data processing method for forestry ecological monitoring according to claim 1, characterized in that: The S2 also includes image data correction based on local histogram, applying grayscale stretching algorithm to smoke and fire areas, optimizing image quality using adaptive contrast enhancement method, and removing image noise through noise filter.

6. The intelligent data processing method for forestry ecological monitoring according to claim 5, characterized in that: The image data correction based on the local histogram includes dividing the image data into a plurality of non-overlapping sub-regions, calculating a histogram for each sub-region, and applying a histogram equalization formula H(i)=G×(n1+n2+...+n i ) / E, where H(i) represents the cumulative histogram, G is a constant, and n1 to n i Represents the number of pixels in the histogram, E is the total number of pixels, and the sub-regions after histogram equalization are merged into the complete image data.

7. The intelligent data processing method for forestry ecological monitoring according to claim 6, characterized in that: The calculation of the histogram for each sub-region includes extracting the pixel value range [a, b] of the sub-region, calculating the frequency of each pixel value, and mapping the pixel value range [a, b] to [0, 255] using linear interpolation. The specific formula is: f(x) = 255 × (y-a) / (b-a), where f(x) is the mapped pixel value, y is the original pixel value, and a and b are the minimum and maximum pixel values, respectively.

8. The intelligent data processing method for forestry ecological monitoring according to claim 1, characterized in that: The image data in S1 includes multispectral images, thermal infrared images and visible light images.

9. The intelligent data processing method for forestry ecological monitoring according to claim 1, characterized in that: The features of the image data in S3 include texture features, shape features and spectral features.

10. The intelligent data processing method for forestry ecological monitoring according to claim 1, characterized in that: The warning information generated in S5 includes the fire level and the location of the risk area.

Images