Forest monitoring method and system based on image recognition

Abstract

The present application belongs to the technical field of identification monitoring, and particularly relates to a forest monitoring method and system based on image recognition, which comprises the following steps: setting a forest monitoring area, constructing a monitoring network through unmanned aerial vehicle remote sensing and Internet of Things communication based on the set monitoring points; performing data processing on historical image data and real-time image data to obtain processed historical image data and real-time image data; analyzing the processed historical image data and constructing a forest monitoring level; training and extracting the processed real-time image data through image training extraction, comparing the feature data of the real-time image data with the feature data of the historical image data, and adjusting the forest monitoring frequency according to the comparison result and the constructed forest monitoring level. The present application realizes real-time, efficient and accurate monitoring of forests, and can dynamically adjust the monitoring frequency according to the comparison result of real-time images and historical images, thereby significantly improving the efficiency and safety of forest monitoring.

Description

Technical Field

[0001] This invention belongs to the field of identification and monitoring technology, specifically relating to a forest monitoring method and system based on image recognition. Background Technology

[0002] Current forest monitoring technologies mainly rely on manual patrols, video surveillance, and some rudimentary image recognition-based methods. While manual patrols are intuitive, they are inefficient and struggle to cover large areas. Video surveillance can automate monitoring, but its monitoring range is limited and it is poorly adaptable to complex environments. In recent years, image recognition-based forest monitoring technologies have gradually emerged, improving monitoring efficiency through video frame filtering and stitching, but these methods still have certain limitations.

[0003] However, existing forest monitoring technologies have many shortcomings. Manual patrols rely on manpower, are easily limited by weather, terrain, and other natural conditions, and struggle to achieve continuous monitoring. Video surveillance equipment has limited coverage, making comprehensive forest monitoring impossible, and is prone to misjudgments in complex environments. Furthermore, while image recognition-based monitoring methods (such as CN114863350A) have improved monitoring efficiency to some extent, their image data processing capabilities still need improvement, particularly in image resolution, noise reduction, and feature extraction, making them insufficient for rapid response to emergencies such as forest fires. Summary of the Invention

[0004] The purpose of this invention is to provide a forest monitoring method and system based on image recognition. By constructing a monitoring network through UAV remote sensing and Internet of Things communication, and combining technologies such as image enhancement, noise reduction, feature extraction and data clustering, real-time, efficient and accurate monitoring of forests can be achieved. The monitoring frequency can be dynamically adjusted according to the comparison results of real-time images and historical images, which significantly improves the efficiency and security of forest monitoring.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0006] The image recognition-based forest monitoring method includes the following steps:

[0007] S1. Define a forest monitoring area and set up q monitoring points around the forest monitoring area. Based on the set monitoring points, construct a monitoring network through UAV remote sensing and Internet of Things communication.

[0008] S2. Based on the constructed monitoring network, real-time image data of the forest monitoring area is collected and historical image data is obtained. The historical image data and real-time image data are processed to obtain processed historical image data and real-time image data.

[0009] S3. Analyze the processed historical image data and construct forest monitoring levels. The process includes:

[0010] S3.1. The processed historical image data is trained and extracted using an image training extraction method to obtain the feature data of the historical image data;

[0011] S3.2 Classify the feature data of historical image data through data clustering to obtain classified historical image data, and construct forest monitoring levels based on the classified historical image data;

[0012] S4. Train and extract the processed real-time image data through image training extraction, compare the feature data of the obtained real-time image data with the feature data of historical image data, and adjust the forest monitoring frequency according to the comparison results and the constructed forest monitoring level.

[0013] Preferably, in step S1, constructing a monitoring network using UAV remote sensing and IoT communication includes the following steps:

[0014] S1.1 Based on q monitoring points set up around the forest monitoring area, a three-dimensional laser imaging device is installed at each monitoring point, and image data of the forest monitoring area is collected through the installed three-dimensional laser imaging device;

[0015] Based on the designated forest monitoring area, drones are used to collect image data within the forest monitoring area.

[0016] Number the q monitoring points and the drones. Set number Number the drone. Number the q monitoring points;

[0017] S1.2 Setting monitoring points The coordinates are taken as the origin, according to Establish a three-dimensional monitoring coordinate system based on monitoring points. Update and determine the coordinates of all other monitoring points and drones;

[0018] S1.3 Save the coordinates of each monitoring point and build a monitoring network based on the connection status of the 3D laser imaging equipment and the drone. Each monitoring point is set to connect individually with the drone.

[0019] Preferably, in step S2, obtaining the processed historical image data and real-time image data includes the following steps:

[0020] S2.1 The acquired historical image data is processed using data processing methods to obtain processed historical image data;

[0021] S2.2 The acquired real-time image data is processed using data processing methods to obtain processed real-time image data.

[0022] Preferably, in step S2.1, the process of obtaining the processed historical image data is as follows:

[0023] The acquired historical image data is enhanced using image enhancement algorithms;

[0024] Based on the grayscale value distribution in historical image data, calculate the pixel probability distribution of the acquired historical image data:

[0025] ;

[0026] Where P represents the pixel probability of the acquired historical image data. This represents the number of pixels with a grayscale value of u, where N is the total number of pixels.

[0027] Construct an image cumulative distribution function based on the calculated pixel probability distribution of the acquired historical image data:

[0028] ;

[0029] Where L represents the total number of gray values ​​in the image. This represents the gray value of the r-th gray value after being mapped by the cumulative distribution function;

[0030] Based on the constructed image cumulative distribution function, image enhancement is performed on the acquired historical image data through data mapping.

[0031] Set the historical image data after image enhancement to the processed historical image data.

[0032] Preferably, in step S2.2, the process of obtaining the processed real-time image data is as follows:

[0033] The noise reduction algorithm is used to denoise the acquired real-time image data, as shown below:

[0034] ;

[0035] in, For noisy images, The image after noise reduction. Noise in the image;

[0036] With the image frame count set to M, the denoised real-time image data is synthesized using a multi-frame averaging algorithm. The formula for the synthesized image is shown below:

[0037] ;

[0038] in, Indicates the coordinate position of a pixel. This represents the average grayscale value of the pixel after synthesis. For the image in frame l at position grayscale value; For an ideal noise-free image at position The grayscale value below, For the image in frame l at position The grayscale value of the noise;

[0039] The real-time image data after averaging multiple frames is set as the processed real-time image data.

[0040] Preferably, in step S3.1, obtaining the feature data of the historical image data includes the following steps:

[0041] S3.1.1 Divide the processed historical image data evenly into m image data blocks;

[0042] S3.1.2. The m divided image data blocks are input into the convolutional neural network. After receiving the input image data blocks, the convolutional layers in the convolutional neural network perform convolution calculations on the input image data blocks based on the set convolutional kernel size, convolutional stride, and initial weights, extract the corresponding feature values, and perform multiplication and accumulation. The convolution calculation formula is as follows:

[0043] ;

[0044] Where I represents the input image data block, Q represents the weight of the corresponding convolution kernel, b represents the bias value, and F represents the feature value obtained by convolution calculation;

[0045] S3.1.3. The convolutional layers and pooling layers in the convolutional neural network are stacked continuously. After the initial extraction by the convolutional layer is completed, the obtained feature values ​​are processed by the activation function and input into the pooling layer.

[0046] S3.1.4 Optimize the features of image data blocks using the SSE algorithm during the continuous stacking of convolution and pooling processes;

[0047] S3.1.5. Input the optimized image data block features into the fully connected layer of the convolutional neural network;

[0048] S3.1.6. Expand and combine the feature data through a fully connected layer to obtain a feature array, and save it;

[0049] S3.1.7 Set the feature array obtained from each historical image data as the feature data of the historical image data.

[0050] Preferably, in S3.1.4, optimizing the features of the image data block using the SSE algorithm includes the following steps:

[0051] The SSE algorithm reduces the channel dimension to one dimension by performing 1×1 convolution on the features of image data blocks, and then... The function normalizes the data and multiplies it with the features of the original image data block to complete the filtering of the feature space information of the image data block;

[0052] The calculation formula for the SSE algorithm is as follows:

[0053] ;

[0054] The function is represented as:

[0055] ;

[0056] The characteristics of the input image data block are as follows: , The output is the input features after pooling. Let J be the k-th input feature in Y, J be the number of input features, and H and W represent the height and width of the image data block, respectively. , This represents the coordinates of image data block features within the channels of a convolutional neural network. Indicating image data block features in eigenvalues ​​at that location for The function, y represents The input to the function is e, which is a natural constant.

[0057] Preferably, in step S3.2, constructing the forest monitoring level based on the classified historical image data includes the following steps:

[0058] Randomly select c feature arrays from the saved feature arrays as cluster centers, with each cluster center representing a type;

[0059] Calculate the distance between the other saved feature arrays and the cluster centers, and add the other feature arrays to the nearest cluster centers based on the calculated distances, until all feature arrays are classified;

[0060] After classification, the cluster centers in each type are recalculated using the mean algorithm, and iterative clustering is performed based on the recalculated cluster centers until all feature arrays converge, and the classified historical image data is output.

[0061] Experts score the categorized historical image data, and forest monitoring levels are constructed based on the scoring results.

[0062] Preferably, in step S4, adjusting the forest monitoring frequency based on the comparison results and the constructed forest monitoring level includes the following steps:

[0063] The feature data of the real-time image data is compared with the feature data of the corresponding historical image data using a similarity comparison method. The similarity comparison formula is as follows:

[0064] ;

[0065] in, The similarity between feature data D of real-time image data and feature data E of corresponding classified historical image data;

[0066] A similarity threshold is set. When the calculation result is greater than or equal to the set threshold, it indicates that the feature data of the real-time image data is similar to the feature data of the corresponding historical image data.

[0067] When the calculation result is less than the set threshold, it means that the feature data of the real-time image data is not similar to the feature data of the corresponding historical image data.

[0068] The scores of historical image data with similar classifications to the real-time image data are used as the scores of the real-time image data, and the forest monitoring frequency is adjusted according to the forest monitoring level corresponding to the scores.

[0069] An image recognition-based forest monitoring system, used to implement the above methods, includes:

[0070] The data acquisition module is used to acquire real-time image data and historical image data;

[0071] The data processing module is used to process the acquired real-time image data and historical image data;

[0072] The data analysis module is used to analyze the processed real-time image data and historical image data;

[0073] The comparison module is used to compare the analyzed real-time image data with historical image data;

[0074] The monitoring adjustment module is used to adjust the monitoring frequency based on the comparison results.

[0075] The beneficial effects of this invention are as follows:

[0076] This invention establishes a forest monitoring area and sets up q monitoring points around it. A monitoring network is constructed based on these monitoring points using UAV remote sensing and IoT communication. Once constructed, real-time and historical image data are collected and processed in real time. After processing, the processed real-time and historical image data are analyzed and feature extracted using image analysis methods. Finally, the feature data of the real-time image data is compared with the feature data of the historical image data, and the forest monitoring frequency is adjusted based on the comparison results, thereby increasing the frequency of forest monitoring.

[0077] This invention constructs a monitoring network using drone remote sensing and Internet of Things communication, solving the problems of limited and insufficient video monitoring range, and improving the security of forest monitoring.

[0078] This invention processes the acquired real-time image data and historical image data separately. First, it processes the acquired historical image data through image enhancement, which improves the accuracy of historical image data analysis. Then, it processes the acquired real-time image data through noise reduction and multi-frame synthesis, which improves the real-time performance of forest monitoring and can meet the needs of rapid response to emergencies such as forest fires.

[0079] This invention trains and extracts processed historical image data through image training. After extraction, the feature data of the historical image data is classified through data clustering. After classification, a forest monitoring level is constructed based on the classified historical image data, thereby improving the rationality of forest monitoring.

[0080] This invention compares the feature data of real-time image data with the feature data of historical image data using a similarity comparison method, which improves the accuracy of real-time image analysis and helps to more accurately detect early signs of forest fires or illegal intrusions, thereby better achieving the purpose of forest monitoring. Attached Figure Description

[0081] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0082] Figure 2 This is a schematic diagram of the system modules of the present invention. Detailed Implementation

[0083] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0084] Example 1: As Figure 1 As shown, this invention provides a forest monitoring method based on image recognition, comprising:

[0085] S1. Define a forest monitoring area and set up q monitoring points around the forest monitoring area. Based on the set monitoring points, construct a monitoring network through UAV remote sensing and Internet of Things communication.

[0086] S2. Based on the constructed monitoring network, real-time image data of the forest monitoring area is collected and historical image data is obtained. The historical image data and real-time image data are processed to obtain processed historical image data and real-time image data.

[0087] S3. Analyze the processed historical image data and construct forest monitoring levels. The process includes:

[0088] S3.1. The processed historical image data is trained and extracted using an image training extraction method to obtain the feature data of the historical image data;

[0089] S3.2 Classify the feature data of historical image data through data clustering to obtain classified historical image data, and construct forest monitoring levels based on the classified historical image data;

[0090] S4. Train and extract the processed real-time image data through image training extraction, compare the feature data of the obtained real-time image data with the feature data of historical image data, and adjust the forest monitoring frequency according to the comparison results and the constructed forest monitoring level.

[0091] In S1, the construction of a monitoring network using drone remote sensing and IoT communication includes the following steps:

[0092] S1.1 Based on q monitoring points set up around the forest monitoring area, a three-dimensional laser imaging device is installed at each monitoring point, and image data of the forest monitoring area is collected through the installed three-dimensional laser imaging device;

[0093] Based on the designated forest monitoring area, drones are used to collect image data within the forest monitoring area.

[0094] Number the q monitoring points and the drones. Set number Number the drone. Number the q monitoring points;

[0095] S1.2 Setting monitoring points The coordinates are taken as the origin, according to Establish a three-dimensional monitoring coordinate system based on monitoring points. Update and determine the coordinates of all other monitoring points and drones;

[0096] S1.3 Save the coordinates of each monitoring point and build a monitoring network based on the connection status of the 3D laser imaging equipment and the drone. Each monitoring point is set to connect individually with the drone.

[0097] In S2, obtaining the processed historical image data and real-time image data includes the following steps:

[0098] S2.1. The acquired historical image data is processed using data processing methods to obtain processed historical image data:

[0099] The acquired historical image data is enhanced using image enhancement algorithms;

[0100] Based on the grayscale value distribution in historical image data, calculate the pixel probability distribution of the acquired historical image data:

[0101] ;

[0102] Where P represents the pixel probability of the acquired historical image data. This represents the number of pixels with a grayscale value of u, where N is the total number of pixels.

[0103] Construct an image cumulative distribution function based on the calculated pixel probability distribution of the acquired historical image data:

[0104] ;

[0105] Where L represents the total number of gray values ​​in the image. This represents the gray value of the r-th gray value after being mapped by the cumulative distribution function;

[0106] Based on the constructed image cumulative distribution function, image enhancement is performed on the acquired historical image data through data mapping.

[0107] Set the historical image data after image enhancement to the processed historical image data.

[0108] S2.2. The acquired real-time image data is processed using data processing methods to obtain processed real-time image data:

[0109] The noise reduction algorithm is used to denoise the acquired real-time image data, as shown below:

[0110] ;

[0111] in, For noisy images, The image after noise reduction. Noise in the image;

[0112] With the image frame count set to M, the denoised real-time image data is synthesized using a multi-frame averaging algorithm. The formula for the synthesized image is shown below:

[0113] ;

[0114] in, Indicates the coordinate position of a pixel. This represents the average grayscale value of the pixel after synthesis. For the image in frame l at position grayscale value; For an ideal noise-free image at position The grayscale value below, For the image in frame l at position The grayscale value of the noise;

[0115] The real-time image data after averaging multiple frames is set as the processed real-time image data.

[0116] In S3.1, obtaining the feature data of historical image data includes the following steps:

[0117] S3.1.1 Divide the processed historical image data evenly into m image data blocks;

[0118] S3.1.2. The m divided image data blocks are input into the convolutional neural network. After receiving the input image data blocks, the convolutional layers in the convolutional neural network perform convolution calculations on the input image data blocks based on the set convolutional kernel size, convolutional stride, and initial weights, extract the corresponding feature values, and perform multiplication and accumulation. The convolution calculation formula is as follows:

[0119] ;

[0120] Where I represents the input image data block, Q represents the weight of the corresponding convolution kernel, b represents the bias value, and F represents the feature value obtained by convolution calculation;

[0121] S3.1.3. The convolutional layers and pooling layers in the convolutional neural network are stacked continuously. After the initial extraction by the convolutional layer is completed, the obtained feature values ​​are processed by the activation function and input into the pooling layer.

[0122] Set the feature values ​​after the convolutional layer has extracted the feature values. V represents the number of eigenvalues, and each eigenvalue... The corresponding weights are ;

[0123] The extracted feature values ​​are processed by an activation function to obtain the processed feature values.

[0124] The activation function formula is shown below:

[0125] ;

[0126] in, is the corresponding activation function, and y is the processed feature value;

[0127] S3.1.4 The SSE algorithm is used to optimize the features of image data blocks during the continuous stacking of convolutions and pooling processes.

[0128] The SSE algorithm reduces the channel dimension to one dimension by performing 1×1 convolution on the features of image data blocks, and then... The function normalizes the data and multiplies it with the features of the original image data block to complete the filtering of the feature space information of the image data block;

[0129] The calculation formula for the SSE algorithm is as follows:

[0130] ;

[0131] The function is represented as:

[0132] ;

[0133] The characteristics of the input image data block are as follows: , The output is the input features after pooling. Let J be the k-th input feature in Y, J be the number of input features, and H and W represent the height and width of the image data block, respectively. , This represents the coordinates of image data block features within the channels of a convolutional neural network. Indicating image data block features in eigenvalues ​​at that location for The function, y represents The input to the function, e is the natural constant;

[0134] S3.1.5. Input the optimized image data block features into the fully connected layer of the convolutional neural network;

[0135] S3.1.6. Expand and combine the feature data through a fully connected layer to obtain a feature array, and save it;

[0136] S3.1.7 Set the feature array obtained from each historical image data as the feature data of the historical image data.

[0137] In S3.2, the process of constructing forest monitoring levels based on classified historical image data includes the following steps:

[0138] Randomly select c feature arrays from the saved feature arrays as cluster centers, with each cluster center representing a type;

[0139] Calculate the distance between the other saved feature arrays and the cluster centers, and add the other feature arrays to the nearest cluster centers based on the calculated distances, until all feature arrays are classified;

[0140] After classification, the cluster centers in each type are recalculated using the mean algorithm, and iterative clustering is performed based on the recalculated cluster centers until all feature arrays converge, and the classified historical image data is output.

[0141] Experts score the categorized historical image data, and forest monitoring levels are constructed based on the scoring results.

[0142] For example, the rating range is set to 0-100, Level 1 is 80-100 and marked as excellent, Level 2 is 70-80 and marked as good, Level 3 is 60-70 and marked as passable, and Level 4 is 0-60 and marked as fail.

[0143] The monitoring frequency for Level 1 drones is set to once every 24 hours, for Level 2 drones to once every 12 hours, for Level 3 drones to once every 6 hours, and for Level 4 drones to once every 2 hours.

[0144] In S4, adjusting the forest monitoring frequency based on the comparison results and the constructed forest monitoring level includes the following steps:

[0145] The feature data of the real-time image data is compared with the feature data of the corresponding historical image data using a similarity comparison method. The similarity comparison formula is as follows:

[0146] ;

[0147] in, The similarity between feature data D of real-time image data and feature data E of corresponding classified historical image data;

[0148] A similarity threshold is set. When the calculation result is greater than or equal to the set threshold, it indicates that the feature data of the real-time image data is similar to the feature data of the corresponding historical image data.

[0149] When the calculation result is less than the set threshold, it means that the feature data of the real-time image data is not similar to the feature data of the corresponding historical image data, and real-time monitoring and comparison calculation will continue.

[0150] The scores of historical image data with similar classifications to the real-time image data are used as the scores of the real-time image data, and the forest monitoring frequency is adjusted according to the forest monitoring level corresponding to the scores.

[0151] Example 2: This example is based on Example 1 and makes the following improvements: First, a light intensity sensor is set up in the monitoring area. Then, in S2.1, before performing image enhancement processing on the historical image data, the historical light intensity information of the forest monitoring area is obtained through the light intensity sensor.

[0152] Based on historical illumination intensity information, the contrast stretching parameters in the image enhancement algorithm are dynamically adjusted. When the illumination intensity is low, the intensity of contrast stretching is increased to improve image brightness and detail; when the illumination intensity is high, the intensity of contrast stretching is appropriately reduced to avoid image overexposure.

[0153] An improved histogram equalization algorithm is introduced, which combines illumination intensity information to perform segmented equalization processing on the image's histogram. The grayscale value range of the image is divided into multiple intervals, and the equalization parameters of each interval are dynamically adjusted according to the illumination intensity to better preserve details in both dark and bright areas of the image.

[0154] During image enhancement, the image is divided into blocks based on local illumination changes. More aggressive enhancement strategies are used for areas with significant illumination changes, while gentler strategies are employed for areas with minor changes, achieving adaptive enhancement both globally and locally.

[0155] For real-time image data, the same image enhancement processing can be performed based on real-time illumination intensity information before noise reduction to improve image quality.

[0156] Example 3: This example improves upon Example 1 as follows: In S2.2, when denoising the acquired real-time image data, not only is a denoising algorithm used, but an adaptive filtering algorithm is also combined to process the real-time image data. The specific steps are as follows:

[0157] Obtain noise intensity information from real-time image data, and dynamically adjust the parameters of the noise reduction algorithm based on the noise intensity to adapt to image noise reduction requirements under different noise intensities;

[0158] Meanwhile, the real-time image data is further processed based on the adaptive filtering algorithm. The filter parameters of the adaptive filtering algorithm are adaptively adjusted according to the local features of the image to better preserve image details and remove noise.

[0159] The real-time image data processed by the noise reduction algorithm and the adaptive filtering algorithm are fused to obtain the final processed real-time image data.

[0160] This improved method can more effectively remove noise from real-time image data while better preserving image detail, thereby improving the accuracy and reliability of subsequent image recognition-based forest monitoring.

[0161] Example 4: Figure 2 As shown, the image recognition-based forest monitoring system, used to implement the methods in Examples 1-3, includes:

[0162] The data acquisition module is used to acquire real-time image data and historical image data;

[0163] The data processing module is used to process the acquired real-time image data and historical image data;

[0164] The data analysis module is used to analyze the processed real-time image data and historical image data;

[0165] The comparison module is used to compare the analyzed real-time image data with historical image data;

[0166] The monitoring adjustment module is used to adjust the monitoring frequency based on the comparison results.

[0167] Example 5: A forest monitoring device based on image recognition, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the computer program to implement the methods in Examples 1-3.

Claims

1. A forest monitoring method based on image recognition, characterized in that, Includes the following steps: S1. Define a forest monitoring area and set up q monitoring points around the forest monitoring area. Based on the set monitoring points, construct a monitoring network through UAV remote sensing and Internet of Things communication. S2. Based on the constructed monitoring network, real-time image data of the forest monitoring area is collected and historical image data is obtained. The historical image data and real-time image data are processed to obtain processed historical image data and real-time image data. S3. Analyze the processed historical image data and construct forest monitoring levels. The process includes: S3.1 The processed historical image data is trained and extracted using an image training extraction method to obtain the feature data of the historical image data, including the following steps: S3.1.1 Divide the processed historical image data evenly into m image data blocks; S3.1.

2. The m divided image data blocks are input into the convolutional neural network. After receiving the input image data blocks, the convolutional layers in the convolutional neural network perform convolution calculations on the input image data blocks based on the set convolutional kernel size, convolutional stride, and initial weights, extract the corresponding feature values, and perform multiplication and accumulation. The convolution calculation formula is as follows: ； Where I represents the input image data block, Q represents the weight of the corresponding convolution kernel, b represents the bias value, and F represents the feature value obtained by convolution calculation; S3.1.

3. The convolutional layers and pooling layers in the convolutional neural network are stacked continuously. After the initial extraction by the convolutional layer is completed, the obtained feature values ​​are processed by the activation function and input into the pooling layer. S3.1.4 Optimize the features of image data blocks using the SSE algorithm during the continuous stacking of convolution and pooling processes; S3.1.

5. Input the optimized image data block features into the fully connected layer of the convolutional neural network; S3.1.

6. Expand and combine the feature data through a fully connected layer to obtain a feature array, and save it; S3.1.

7. Set the feature array obtained from each historical image data as the feature data of the historical image data; S3.2 Classify the feature data of historical image data through data clustering to obtain classified historical image data, and construct forest monitoring levels based on the classified historical image data; S4. Train and extract the processed real-time image data through image training extraction, compare the feature data of the obtained real-time image data with the feature data of historical image data, and adjust the forest monitoring frequency according to the comparison results and the constructed forest monitoring level.

2. The forest monitoring method based on image recognition according to claim 1, characterized in that, In S1, the construction of a monitoring network through UAV remote sensing and IoT communication includes the following steps: S1.1 Based on q monitoring points set up around the forest monitoring area, a three-dimensional laser imaging device is installed at each monitoring point, and image data of the forest monitoring area is collected through the installed three-dimensional laser imaging device; Based on the designated forest monitoring area, drones are used to collect image data within the forest monitoring area. Number the q monitoring points and the drones. Set number Number the drone. Number the q monitoring points; S1.2 Setting monitoring points The coordinates are taken as the origin, according to Establish a three-dimensional monitoring coordinate system based on monitoring points. Update and determine the coordinates of all other monitoring points and drones; S1.3 Save the coordinates of each monitoring point and build a monitoring network based on the connection status of the 3D laser imaging equipment and the drone. Each monitoring point is set to connect individually with the drone.

3. The forest monitoring method based on image recognition according to claim 1, characterized in that, In step S2, obtaining the processed historical image data and real-time image data includes the following steps: S2.1 The acquired historical image data is processed using data processing methods to obtain processed historical image data; S2.2 The acquired real-time image data is processed using data processing methods to obtain processed real-time image data.

4. The forest monitoring method based on image recognition according to claim 3, characterized in that, In step S2.1, the process of obtaining the processed historical image data is as follows: The acquired historical image data is enhanced using image enhancement algorithms; Based on the grayscale value distribution in historical image data, calculate the pixel probability distribution of the acquired historical image data: ； Where P represents the pixel probability of the acquired historical image data. This represents the number of pixels with a grayscale value of u, where N is the total number of pixels. Construct an image cumulative distribution function based on the calculated pixel probability distribution of the acquired historical image data: ； Where L represents the total number of gray values ​​in the image. This represents the gray value of the r-th gray value after being mapped by the cumulative distribution function; Based on the constructed image cumulative distribution function, image enhancement is performed on the acquired historical image data through data mapping. Set the historical image data after image enhancement to the processed historical image data.

5. The forest monitoring method based on image recognition according to claim 3, characterized in that, In step S2.2, the process of obtaining the processed real-time image data is as follows: The noise reduction algorithm is used to denoise the acquired real-time image data, as shown below: ； in, For noisy images, The image after noise reduction. Noise in the image; With the image frame count set to M, the denoised real-time image data is synthesized using a multi-frame averaging algorithm. The formula for the synthesized image is shown below: ； in, Indicates the coordinate position of a pixel. This represents the average grayscale value of the pixel after synthesis. For the image in frame l at position grayscale value; For an ideal noise-free image at position The grayscale value below, For the image in frame l at position The grayscale value of the noise; The real-time image data after averaging multiple frames is set as the processed real-time image data.

6. The forest monitoring method based on image recognition according to claim 1, characterized in that, In S3.1.4, optimizing the features of image data blocks using the SSE algorithm includes the following steps: The SSE algorithm reduces the channel dimension to one dimension by performing 1×1 convolution on the features of image data blocks, and then... The function normalizes the data and multiplies it with the features of the original image data block to complete the filtering of the feature space information of the image data block; The calculation formula for the SSE algorithm is as follows: ； The function is represented as: ； The characteristics of the input image data block are as follows: , The output is the input features after pooling. Let J be the k-th input feature in Y, J be the number of input features, and H and W represent the height and width of the image data block, respectively. , This represents the coordinates of image data block features within the channels of a convolutional neural network. Indicating image data block features in eigenvalues ​​at that location for The function, y represents The input to the function is e, which is a natural constant.

7. The forest monitoring method based on image recognition according to claim 1, characterized in that, In step S3.2, constructing the forest monitoring level based on the classified historical image data includes the following steps: Randomly select c feature arrays from the saved feature arrays as cluster centers, with each cluster center representing a type; Calculate the distance between the other saved feature arrays and the cluster centers, and add the other feature arrays to the nearest cluster centers based on the calculated distances, until all feature arrays are classified; After classification, the cluster centers in each type are recalculated using the mean algorithm, and iterative clustering is performed based on the recalculated cluster centers until all feature arrays converge, and the classified historical image data is output. Experts score the categorized historical image data, and forest monitoring levels are constructed based on the scoring results.

8. The forest monitoring method based on image recognition according to claim 7, characterized in that, In step S4, adjusting the forest monitoring frequency based on the comparison results and the constructed forest monitoring level includes the following steps: The feature data of the real-time image data is compared with the feature data of the corresponding historical image data using a similarity comparison method. The similarity comparison formula is as follows: ； in, The similarity between feature data D of real-time image data and feature data E of corresponding classified historical image data; A similarity threshold is set. When the calculation result is greater than or equal to the set threshold, it indicates that the feature data of the real-time image data is similar to the feature data of the corresponding historical image data. When the calculation result is less than the set threshold, it means that the feature data of the real-time image data is not similar to the feature data of the corresponding historical image data. The scores of historical image data with similar classifications to the real-time image data are used as the scores of the real-time image data, and the forest monitoring frequency is adjusted according to the forest monitoring level corresponding to the scores.

9. A forest monitoring system based on image recognition, used to implement the method according to any one of claims 1-8, characterized in that, include: The data acquisition module is used to acquire real-time image data and historical image data; The data processing module is used to process the acquired real-time image data and historical image data; The data analysis module is used to analyze the processed real-time image data and historical image data; The comparison module is used to compare the analyzed real-time image data with historical image data; The monitoring adjustment module is used to adjust the monitoring frequency based on the comparison results.

Images