Remote sensing image detail enhancement method and system based on target detection

Abstract

The invention discloses a remote sensing image detail enhancement method and system based on target detection. The method comprises the following steps: acquiring remote sensing image data; dividing a remote sensing image into a plurality of parts based on ground feature distribution in the remote sensing image data; on the basis of ground feature images in the remote sensing images of all the parts, detail features and detail distribution rules of the ground feature images are obtained; on the basis of the detail features and the detail distribution rule of the remote sensing image surface features of each part, adding the detail features into the image, and obtaining the enhanced remote sensing image of each part; and based on the plurality of enhanced remote sensing images, performing target detection to obtain a plurality of groups of recognition result probabilities, and obtaining a recognition result according to the comprehensive probability. The method has the advantages that the feature distribution probability is obtained by performing region division on the remote sensing image and extracting the detail features, so that the target detection precision is improved, especially on complex ground features and detail expressions.

Description

Technical Field

[0001] The present invention relates to target detection technology, and in particular to a remote sensing image detail enhancement method and system based on target detection. Background Art

[0002] The remote sensing image detail enhancement method and system based on target detection obtains remote sensing image data and divides the image into multiple parts according to the distribution of objects, ensuring that the object features in each area are analyzed independently. Then, by analyzing the detail features and distribution laws of each part of the object image, the image is further enhanced in detail to highlight the edges and weak features and improve the recognizability of the target. The enhanced parts of the image are processed by target detection to obtain multiple groups of recognition result probabilities. Finally, by combining the recognition result probabilities of each area, a more accurate and reliable overall recognition result is calculated. This method effectively improves the detection accuracy of weak or complex objects in the image by performing detail enhancement and target detection on different areas, especially in complex scenes, which can reduce errors and improve the robustness of detection.

[0003] The current target detection-based remote sensing image detail enhancement methods on the market use target detection algorithms to identify target areas in images, and then perform fine detail enhancement processing on these areas. Convolutional neural networks are usually used for feature extraction, and the recognizability and detail clarity of targets in images are improved by enhancing the contrast of target areas and sharpening edges. Some methods use adaptive enhancement strategies to perform local optimization based on the characteristics of different areas, thereby improving the analysis capabilities of images in complex environments. Summary of the invention

[0004] In order to improve the existing remote sensing image detail enhancement method, a remote sensing image detail enhancement method and system based on target detection are provided. The method improves the accuracy of target detection by dividing the remote sensing image into multiple regions and enhancing the details according to the ground object features in each region. The comprehensive recognition result probability of each part can effectively reduce errors and improve the accuracy and robustness of the overall detection.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0006] Remote sensing image detail enhancement methods based on target detection include:

[0007] Acquisition of remote sensing image data;

[0008] Based on the distribution of objects in the remote sensing image data, the remote sensing image is divided into several parts;

[0009] Based on the ground object images in the remote sensing images of each part, its detail features and detail distribution rules are obtained;

[0010] Based on the detail features and detail distribution rules of the objects in each part of the remote sensing image, detail features are added to the image to obtain enhanced remote sensing images of each part;

[0011] Based on multiple enhanced remote sensing images of each part, target detection is performed, multiple groups of recognition result probabilities are obtained, and recognition results are obtained according to the comprehensive probability.

[0012] Preferably, dividing the remote sensing image into several parts based on the distribution of objects in the remote sensing image data specifically includes:

[0013] Obtain the gradient information of various objects in the remote sensing image, and divide the remote sensing image based on the watershed algorithm;

[0014] Based on the divided remote sensing images, the portion containing the ground object image is obtained.

[0015] Preferably, the obtaining of detail features and detail distribution rules of the ground object images in the remote sensing images of each part specifically includes:

[0016] Perform image preprocessing based on remote sensing images of each part;

[0017] Based on the preprocessed remote sensing images, the detailed features in the remote sensing images are obtained by the box counting method in the typing analysis method;

[0018] Based on the acquired detail features of remote sensing images, the distribution law of details in the image is obtained by combining image-based color space conversion with self-organizing mapping.

[0019] Preferably, the method of obtaining detailed features in the remote sensing image by a box counting method in the typing analysis method based on the preprocessed remote sensing image specifically includes:

[0020] Based on the rule from large to small, select box sizes ∈ of different scales;

[0021] Based on the box size of each scale, the image is divided into ∈×∈ grids, and each box is counted to see whether it contains the image features. If so, the box count is 1, otherwise it is 0;

[0022] For different ∈ values, get the number of boxes N(∈) at each scale;

[0023] Based on the relationship between the number of boxes N(∈) and the box size ∈, the fractal dimension D is obtained, and the formula is:

[0024]

[0025] The specific steps are to perform logarithmic transformation on the number of boxes N(∈) and the box size ∈ of different scales, draw a curve of log N(∈) versus log N(∈), and calculate the slope of the curve, which is the fractal dimension D. The fractal dimension reflects the complexity of the image at different scales. The larger the value, the more complex the image structure and the richer the details.

[0026] Based on the obtained fractal dimension D, the region to which each part of the remote sensing image belongs is determined. For example, regions with high fractal dimensions are usually cities and forests, and regions with low fractal dimensions are usually water bodies and agricultural areas.

[0027] Based on the determined regions to which each part of the remote sensing image belongs, the image detail features of each region are obtained through the spectral information contained in each pixel of the remote sensing image.

[0028] Preferably, the method of obtaining the detail distribution law in the image based on the acquired remote sensing image detail features by combining the image-based color space conversion with the self-organizing map specifically includes:

[0029] Based on the conversion of each part of the remote sensing image from RGB space to HSV color space;

[0030] Based on the acquired image detail features of each region, it is converted into a high-dimensional feature vector containing various parameters in the color space;

[0031] The self-organizing map network is trained based on the high-dimensional feature vector, the Euclidean distance between the high-dimensional feature vector and all nodes in the network is calculated, and the best matching unit is found;

[0032] According to the weight of the best matching unit, the weights of the best matching unit and its neighboring nodes are updated to map the input high-dimensional data to a low-dimensional grid;

[0033] Based on the self-organizing mapping network after training, the mapping position of each part of the image area in the low-dimensional grid is obtained, and the distribution law of the detailed features of the remote sensing image is obtained through the mapping position.

[0034] Preferably, the adding of detail features to the image based on the detail features and detail distribution rules of the objects in each part of the remote sensing image to obtain the enhanced remote sensing images of each part specifically comprises:

[0035] Based on the acquired detail features, detail features are added to the remote sensing images of each part according to the distribution law of detail features to obtain a detail enhancement distribution map;

[0036] Based on the acquired detail enhancement distribution map and the original remote sensing image, high-frequency components of the detail enhancement distribution map and the original remote sensing image are obtained through high-frequency filtering, and low-frequency components of the original remote sensing image are obtained through low-frequency filtering;

[0037] The high-frequency part of the detail enhancement distribution map is weightedly fused with the high-frequency part of the original remote sensing image;

[0038] The fused high-frequency part is fused with the low-frequency component of the original remote sensing image to obtain enhanced remote sensing images of each part;

[0039] Based on the enhanced remote sensing images of each part, post-processing is performed to enhance the image effect.

[0040] Preferably, the target detection is performed based on the multiple enhanced remote sensing images, and multiple groups of recognition result probabilities are obtained, and the recognition results are obtained according to the comprehensive probability, which specifically includes:

[0041] Acquire multiple enhanced remote sensing images of each part based on detail enhancement method;

[0042] Based on multiple enhanced remote sensing images of each part, the recognition result probability of the ground object in each part of the remote sensing image is obtained through an intelligent recognition algorithm;

[0043] Based on the probabilities of multiple object recognition results obtained from multiple enhanced remote sensing images of each part, weighted calculation is performed according to the number of detail features in the detail enhancement distribution map, and the one with the highest comprehensive probability is obtained as the recognition result of the current part. The formula is:

[0044] P=α 1 ·P 1 +α 2 ·P 2 +...+α n ·P n

[0045] Among them, P is the comprehensive probability, α 1 , α 2 , ..., α n is the weighted value obtained according to the number of detail features in the detail enhancement distribution map, P 1 , P 2 , ..., P n The probability of recognition results for each enhanced remote sensing image.

[0046] Furthermore, a remote sensing image detail enhancement system based on target detection is proposed, comprising:

[0047] Remote sensing image acquisition module: The remote sensing image acquisition module is mainly used to acquire remote sensing image data of each area;

[0048] Image division module: The image division module is mainly used to divide the remote sensing image into several parts based on the distribution of objects in the remote sensing image data;

[0049] Image preprocessing module: The image preprocessing module mainly includes denoising, normalization, smoothing, etc.;

[0050] Box counting module: The box counting module is mainly used to count fractal dimensions and obtain detailed features in remote sensing images;

[0051] Self-organizing map network training module: The self-organizing map network training module is mainly used to obtain the distribution law of detail features by training high-dimensional feature vectors;

[0052] Image enhancement module: The image enhancement module is mainly used to add detail features to the image based on the detail features and detail distribution rules of each part of the remote sensing image objects;

[0053] Image object recognition module: The image object recognition module is mainly used to perform target recognition based on the enhanced remote sensing images and obtain multiple recognition result probabilities;

[0054] Comprehensive probability analysis module: The comprehensive probability analysis module is mainly used to perform weighted comprehensive calculation on the probability of results obtained from multiple remote sensing images to obtain the recognition result with the highest probability;

[0055] Processor: The processor is mainly used for calculating the formulas in the detail features and distribution rules and the probability calculation of the recognition results.

[0056] Compared with the prior art, the advantages of the present invention are:

[0057] By dividing the image into multiple parts, the detailed features of each area can be enhanced independently, thus avoiding the information loss that may be caused by unified processing of the entire image. Enhancement based on the detailed features and distribution patterns of each area can effectively improve the boundaries of weak or blurred objects in the image and enhance the recognizability of the target. After target detection, the combination of recognition results and probabilities of each area can reduce the error of a single detection result, thereby improving the overall detection accuracy. In addition, comprehensive analysis based on the recognition result probabilities of multiple enhanced images helps to eliminate erroneous recognition caused by noise or insufficient local features, making the overall recognition more robust and accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 A schematic diagram of the analysis method of the method and system proposed in the present invention;

[0059] Figure 2 A schematic diagram of remote sensing image division of the method and system proposed in the present invention;

[0060] Figure 3 A schematic diagram of image processing of the method and system proposed in the present invention;

[0061] Figure 4 A schematic diagram for obtaining detailed features of the method and system proposed in the present invention;

[0062] Figure 5 A schematic diagram of obtaining the detailed distribution law of the method and system proposed by the present invention;

[0063] Figure 6 Add schematic diagrams for detailed features of the method and system proposed in the present invention;

[0064] Figure 7 A schematic diagram of image detection of the method and system proposed in the present invention;

[0065] Figure 8 This is a schematic diagram of the electronic device in this solution;

[0066] Fig. 9 This is a schematic diagram of the computer-readable storage medium structure in this solution. DETAILED DESCRIPTION

[0067] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0068] Remote sensing image detail enhancement system based on target detection, including:

[0069] Remote sensing image acquisition module: The remote sensing image acquisition module is mainly used to acquire remote sensing image data of each area;

[0070] Image division module: The image division module is mainly used to divide the remote sensing image into several parts based on the distribution of objects in the remote sensing image data;

[0071] Image preprocessing module: The image preprocessing module mainly includes denoising, normalization, smoothing, etc.;

[0072] Box counting module: The box counting module is mainly used to count fractal dimensions and obtain detailed features in remote sensing images;

[0073] Self-organizing map network training module: The self-organizing map network training module is mainly used to obtain the distribution law of detail features by training high-dimensional feature vectors;

[0074] Image enhancement module: The image enhancement module is mainly used to add detail features to the image based on the detail features and detail distribution rules of each part of the remote sensing image objects;

[0075] Image object recognition module: The image object recognition module is mainly used to perform target recognition based on the enhanced remote sensing images and obtain multiple recognition result probabilities;

[0076] Comprehensive probability analysis module: The comprehensive probability analysis module is mainly used to perform weighted comprehensive calculation on the probability of results obtained from multiple remote sensing images to obtain the recognition result with the highest probability;

[0077] Processor: The processor is mainly used for calculating the formulas in the detail features and distribution rules and the probability calculation of the recognition results.

[0078] See also Figure 1 As shown, based on the above system, the present invention also proposes a remote sensing image detail enhancement method based on target detection, including:

[0079] Step 1: Obtain remote sensing image data;

[0080] Step 2: Divide the remote sensing image into several parts based on the distribution of objects in the remote sensing image data;

[0081] Step 3: Based on the ground feature images in the remote sensing images of each part, obtain their detail features and detail distribution rules;

[0082] Step 4: Based on the detail features and detail distribution rules of the objects in each part of the remote sensing image, add detail features to the image to obtain enhanced remote sensing images of each part;

[0083] Step 5: Based on multiple enhanced remote sensing images, perform target detection, obtain multiple groups of recognition result probabilities, and obtain recognition results based on the comprehensive probability.

[0084] See also Figure 2 As shown in the figure, based on the distribution of objects in the remote sensing image data, the remote sensing image is divided into several parts, including:

[0085] Obtain the gradient information of various objects in the remote sensing image, and divide the remote sensing image based on the watershed algorithm;

[0086] Based on the divided remote sensing images, the portion containing the ground object image is obtained.

[0087] Specifically, the watershed algorithm is usually based on the gradient information of the image, and regards the image as a terrain, simulating water gradually filling up the highlands (areas with higher grayscale values) starting from the valleys (areas with lower grayscale values). The water flow will eventually form "dams" between different valleys, i.e., dividing lines. The different areas finally formed correspond to different parts of the image, thereby achieving image segmentation.

[0088] See also Figure 3 As shown in the figure, based on the ground object images in the remote sensing images of each part, obtaining its detail features and detail distribution laws specifically includes:

[0089] Perform image preprocessing based on remote sensing images of each part;

[0090] Based on the preprocessed remote sensing images, the detailed features in the remote sensing images are obtained by the box counting method in the typing analysis method;

[0091] Based on the acquired detail features of remote sensing images, the distribution law of details in the image is obtained by combining image-based color space conversion with self-organizing mapping.

[0092] See also Figure 4 As shown in the figure, based on the preprocessed remote sensing image, the detailed features in the remote sensing image are obtained by the box counting method in the typing analysis method, including:

[0093] Based on the rule from large to small, select box sizes ∈ of different scales;

[0094] Based on the box size of each scale, the image is divided into ∈×∈ grids, and each box is counted to see whether it contains the image features. If so, the box count is 1, otherwise it is 0;

[0095] For different ∈ values, get the number of boxes N(∈) at each scale;

[0096] Based on the relationship between the number of boxes N(∈) and the box size ∈, the fractal dimension D is obtained, and the formula is:

[0097]

[0098] The specific steps are to perform logarithmic transformation on the number of boxes N(∈) and the box size ∈ of different scales, draw a curve of log N(∈) versus log N(∈), and calculate the slope of the curve, which is the fractal dimension D. The fractal dimension reflects the complexity of the image at different scales. The larger the value, the more complex the image structure and the richer the details.

[0099] Based on the obtained fractal dimension D, the region to which each part of the remote sensing image belongs is determined. For example, regions with high fractal dimensions are usually cities and forests, and regions with low fractal dimensions are usually water bodies and agricultural areas.

[0100] Based on the determined regions to which each part of the remote sensing image belongs, the image detail features of each region are obtained through the spectral information contained in each pixel of the remote sensing image.

[0101] It is understandable that the border of an image is usually simpler than the internal area, which can easily lead to inaccurate calculation of the fractal dimension of the box counting method. The box counting method may be affected by the image boundary effect, especially at a smaller scale, the boundary may appear too simple, which in turn affects the calculation of the fractal dimension. The mirror extension technique can be used to reduce the impact of the boundary effect by mirroring the pixels at the image boundary to maintain a more realistic texture structure, or the boundary area can be removed in image processing to focus on the core area of ​​the image for box counting calculations.

[0102] See also Figure 5 As shown, based on the acquired remote sensing image detail features, the image-based color space conversion and self-organizing map combination method are used to obtain the detail distribution law in the image, including:

[0103] Based on the conversion of each part of the remote sensing image from RGB space to HSV color space;

[0104] Based on the acquired image detail features of each region, it is converted into a high-dimensional feature vector containing various parameters in the color space;

[0105] The self-organizing map network is trained based on the high-dimensional feature vector, the Euclidean distance between the high-dimensional feature vector and all nodes in the network is calculated, and the best matching unit is found;

[0106] According to the weight of the best matching unit, the weights of the best matching unit and its neighboring nodes are updated to map the input high-dimensional data to a low-dimensional grid;

[0107] Based on the self-organizing mapping network after training, the mapping position of each part of the image area in the low-dimensional grid is obtained, and the distribution law of the detailed features of the remote sensing image is obtained through the mapping position.

[0108] See also Figure 6 Based on the detail features and detail distribution rules of each part of the remote sensing image objects, detail features are added to the image to obtain enhanced remote sensing images of each part, specifically including:

[0109] Based on the acquired detail features, detail features are added to the remote sensing images of each part according to the distribution law of detail features to obtain a detail enhancement distribution map;

[0110] Based on the acquired detail enhancement distribution map and the original remote sensing image, high-frequency components of the detail enhancement distribution map and the original remote sensing image are obtained through high-frequency filtering, and low-frequency components of the original remote sensing image are obtained through low-frequency filtering;

[0111] The high-frequency part of the detail enhancement distribution map is weightedly fused with the high-frequency part of the original remote sensing image;

[0112] The fused high-frequency part is fused with the low-frequency component of the original remote sensing image to obtain enhanced remote sensing images of each part;

[0113] Based on the enhanced remote sensing images of each part, post-processing is performed to enhance the image effect.

[0114] See also Figure 7 , based on multiple enhanced remote sensing images of each part, target detection is performed to obtain multiple groups of recognition result probabilities, and the recognition results obtained according to the comprehensive probability include:

[0115] Acquire multiple enhanced remote sensing images of each part based on detail enhancement method;

[0116] Based on multiple enhanced remote sensing images of each part, the recognition result probability of the ground object in each part of the remote sensing image is obtained through an intelligent recognition algorithm;

[0117] Based on the probabilities of multiple object recognition results obtained from multiple enhanced remote sensing images of each part, weighted calculation is performed according to the number of detail features in the detail enhancement distribution map, and the one with the highest comprehensive probability is obtained as the recognition result of the current part. The formula is:

[0118] P=α 1 ·P 1 +α 2 ·P 2 +...+α n ·P n

[0119] Among them, P is the comprehensive probability, α 1 , α 2 , ..., α n is the weighted value obtained according to the number of detail features in the detail enhancement distribution map, P 1 , P 2 , ..., P n The probability of recognition results for each enhanced remote sensing image.

[0120] Furthermore, the method according to the embodiment of the present application can also be performed by Figure 8 The electronic device architecture shown in FIG. Figure 8 As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, a read-only memory (ROM) 503, a random access memory (RAM) 504, a communication port 505 connected to a network, an input / output component 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, may store the remote sensing image detail enhancement method and system based on target detection provided by the present application. The electronic device 500 may also include a terminal interface 508. Of course, Figure 8The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 8 One or more components of an electronic device are shown.

[0121] Fig. 9 Schematic diagram of a computer-readable storage medium structure provided by an embodiment of the present application. Fig. 9 As shown, a computer-readable storage medium 600 according to an embodiment of the present application is shown. Computer-readable instructions are stored on the computer-readable storage medium 600. When the computer-readable instructions are executed by the processor, the method and system for enhancing remote sensing image details based on target detection according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium 600 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0122] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The above is a description of a specific embodiment of this specification. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0123] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0124] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A remote sensing image detail enhancement method based on target detection, characterized in that: include: Acquisition of remote sensing image data; Based on the distribution of objects in the remote sensing image data, the remote sensing image is divided into several parts; Based on the ground object images in the remote sensing images of each part, its detail features and detail distribution rules are obtained; Based on the detail features and detail distribution rules of the objects in each part of the remote sensing image, detail features are added to the image to obtain enhanced remote sensing images of each part; Based on multiple enhanced remote sensing images of each part, target detection is performed, multiple groups of recognition result probabilities are obtained, and recognition results are obtained according to the comprehensive probability.

2. The method for enhancing remote sensing image details based on target detection according to claim 1, characterized in that: The method of dividing the remote sensing image into several parts based on the distribution of objects in the remote sensing image data specifically includes: Obtain the gradient information of various objects in the remote sensing image, and divide the remote sensing image based on the watershed algorithm; Based on the divided remote sensing images, the portion containing the ground object image is obtained.

3. The method for enhancing remote sensing image details based on target detection according to claim 2, characterized in that: The obtaining of the detailed features and detailed distribution rules of the ground object images in the remote sensing images of each part specifically includes: Perform image preprocessing based on remote sensing images of each part; Based on the preprocessed remote sensing images, the detailed features in the remote sensing images are obtained by the box counting method in the typing analysis method; Based on the acquired detail features of remote sensing images, the distribution law of details in the image is obtained by combining image-based color space conversion with self-organizing mapping.

4. The method for enhancing remote sensing image details based on target detection according to claim 3, characterized in that: The method of obtaining detailed features in the remote sensing image by the box counting method in the typing analysis method based on the preprocessed remote sensing image specifically includes: Based on the rule from large to small, select box sizes ∈ of different scales; Based on the box size of each scale, the image is divided into ∈×∈ grids, and each box is counted to see whether it contains the image features. If so, the box count is 1, otherwise it is 0; For different ∈ values, get the number of boxes N(∈) at each scale; Based on the relationship between the number of boxes N(∈) and the box size ∈, the fractal dimension D is obtained, and the formula is: The specific steps are to perform logarithmic transformation on the number of boxes N(∈) and the box size ∈ of different scales, draw a curve of log N(∈) versus log N(∈), and calculate the slope of the curve, which is the fractal dimension D. The fractal dimension reflects the complexity of the image at different scales. The larger the value, the more complex the image structure and the richer the details. Based on the obtained fractal dimension D, the region to which each part of the remote sensing image belongs is determined. For example, regions with high fractal dimensions are usually cities and forests, and regions with low fractal dimensions are usually water bodies and agricultural areas. Based on the determined regions to which each part of the remote sensing image belongs, the image detail features of each region are obtained through the spectral information contained in each pixel of the remote sensing image.

5. The method for enhancing remote sensing image details based on target detection according to claim 4, characterized in that: The method of obtaining the detail distribution law in the image based on the acquired remote sensing image detail features by combining the image-based color space conversion with the self-organizing map specifically includes: Based on the conversion of each part of the remote sensing image from RGB space to HSV color space; Based on the acquired image detail features of each region, it is converted into a high-dimensional feature vector containing various parameters in the color space; The self-organizing map network is trained based on the high-dimensional feature vector, the Euclidean distance between the high-dimensional feature vector and all nodes in the network is calculated, and the best matching unit is found; According to the weight of the best matching unit, the weights of the best matching unit and its neighboring nodes are updated to map the input high-dimensional data to a low-dimensional grid; Based on the self-organizing mapping network after training, the mapping position of each part of the image area in the low-dimensional grid is obtained, and the distribution law of the detailed features of the remote sensing image is obtained through the mapping position.

6. The method for remote sensing image detail enhancement based on target detection according to claim 5, characterized in that: The adding of detail features to the image based on the detail features and detail distribution rules of the objects in the remote sensing image of each part to obtain the enhanced remote sensing image of each part specifically includes: Based on the acquired detail features, detail features are added to the remote sensing images of each part according to the distribution law of detail features to obtain a detail enhancement distribution map; Based on the acquired detail enhancement distribution map and the original remote sensing image, high-frequency components of the detail enhancement distribution map and the original remote sensing image are obtained through high-frequency filtering, and low-frequency components of the original remote sensing image are obtained through low-frequency filtering; The high-frequency part of the detail enhancement distribution map is weightedly fused with the high-frequency part of the original remote sensing image; The fused high-frequency part is fused with the low-frequency component of the original remote sensing image to obtain enhanced remote sensing images of each part; Based on the enhanced remote sensing images of each part, post-processing is performed to enhance the image effect.

7. The method for enhancing remote sensing image details based on target detection according to claim 6, characterized in that: The target detection is performed based on the multiple enhanced remote sensing images, and multiple groups of recognition result probabilities are obtained. The recognition results are obtained according to the comprehensive probabilities, which specifically include: Acquire multiple enhanced remote sensing images of each part based on detail enhancement method; Based on multiple enhanced remote sensing images of each part, the recognition result probability of the ground object in each part of the remote sensing image is obtained through an intelligent recognition algorithm; Based on the probabilities of multiple ground object recognition results obtained from multiple enhanced remote sensing images of each part, weighted calculation is performed according to the number of detail features in the detail enhancement distribution map, and the one with the highest comprehensive probability is obtained as the recognition result of the current part. The formula is: P=α1·P1+α2·P2+...+α n ·P n Among them, P is the comprehensive probability, α1, α2, ..., α n is the weighted value obtained according to the number of detail features in the detail enhancement distribution map, P1, P2, ..., P n The recognition result probability of each enhanced remote sensing image.

8. A remote sensing image detail enhancement system based on target detection, used to implement the remote sensing image detail enhancement method based on target detection as claimed in any one of claims 1 to 7, characterized in that: include: Remote sensing image acquisition module: The remote sensing image acquisition module is mainly used to acquire remote sensing image data of each area; Image division module: The image division module is mainly used to divide the remote sensing image into several parts based on the distribution of objects in the remote sensing image data; Image preprocessing module: The image preprocessing module mainly includes denoising, normalization, smoothing, etc.; Box counting module: The box counting module is mainly used to count fractal dimensions and obtain detailed features in remote sensing images; Self-organizing map network training module: The self-organizing map network training module is mainly used to obtain the distribution law of detail features by training high-dimensional feature vectors; Image enhancement module: The image enhancement module is mainly used to add detail features to the image based on the detail features and detail distribution rules of each part of the remote sensing image objects; Image object recognition module: The image object recognition module is mainly used to perform target recognition based on the enhanced remote sensing images and obtain multiple recognition result probabilities; Comprehensive probability analysis module: The comprehensive probability analysis module is mainly used to perform weighted comprehensive calculation on the probability of results obtained from multiple remote sensing images to obtain the recognition result with the highest probability; Processor: The processor is mainly used for calculating the formulas in the detail features and distribution rules and the probability calculation of the recognition results.

9. An electronic device, characterized in that: include: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the remote sensing image detail enhancement method based on target detection as described in any one of claims 1-7.

10. A computer-readable storage medium storing computer-readable instructions, characterized in that: When the computer-readable instructions are executed by a processor, the remote sensing image detail enhancement method based on target detection according to any one of claims 1 to 7 is implemented.

Images