River and lake long-term flood prevention monitoring image processing method based on artificial intelligence

Abstract

The invention belongs to the field of computer image processing, and particularly relates to a river and lake long-term flood prevention monitoring image processing method based on artificial intelligence. Cameras are arranged at key positions of rivers and lakes, an unmanned aerial vehicle is used for collecting images, and the images are preprocessed by improving dark light enhancement and a noise reduction algorithm so as to improve the image quality in low-light and severe weather. A superpixel segmentation technology and an image change detection algorithm are applied, image gradient information self-adaptive adjustment weight coefficients are calculated, then feature classification is extracted, different time point regions are compared to judge changes, and flood risk is analyzed and predicted in combination with a time sequence. And finally, making emergency plans for different risk levels through a big data platform in combination with real-time monitoring information, providing decision support such as resource allocation and public early warning suggestions, and performing assessment and summarization after flood situations so as to improve flood prevention strategies. The method can effectively solve many problems of existing flood prevention monitoring, improves the intelligent level and efficiency, and guarantees the life and property safety of people and the stable development of regions.

Description

Technical Field

[0001] The present invention belongs to the field of computer image processing, and in particular relates to an artificial intelligence-based river and lake chief system flood control monitoring image processing method. Background Art

[0002] In the flood control work of the river and lake chief system, accurate and timely monitoring of flood conditions is crucial to protecting the lives and property of the people. Existing flood control monitoring mainly relies on manual inspections and simple monitoring equipment, which has many shortcomings. On the one hand, manual inspections are difficult to achieve all-weather, all-round real-time monitoring, and are prone to blind inspections and missed inspections. On the other hand, the images collected by conventional monitoring equipment under bad weather conditions such as low light, heavy rain, and dense fog are often blurred and noisy, which seriously affects the acquisition of key flood information. Moreover, for massive monitoring image data, the existing analysis and processing methods mostly rely on manual experience for naked eye observation and judgment, which is inefficient and difficult to guarantee accuracy. This manual-based method is difficult to quickly and accurately identify subtle changes in key flood control locations such as rivers, lakes, and dams, and cannot detect potential flood risks in time, making it difficult to make effective flood control deployments in advance. In addition, in the process of flood control decision-making, there is a lack of big data support and intelligent analysis methods, making it difficult to quickly generate scientific and reasonable command plans based on real-time conditions. To overcome these problems, an intelligent and efficient flood control monitoring image processing method is urgently needed. Summary of the invention

[0003] In view of the above-mentioned technical problems of flood control monitoring, the present invention proposes an artificial intelligence-based flood control monitoring image processing method of the river and lake chief system.

[0004] In order to achieve the above object, the technical solution adopted by the present invention comprises the following steps:

[0005] S1. First, deploy multiple surveillance cameras at key locations of rivers and lakes to collect surveillance images in real time, and use drones for regular patrols to collect high-angle hydrological images;

[0006] S2, secondly, the image quality under low light and bad weather conditions is improved by using improved dark light enhancement algorithm and noise reduction algorithm for image preprocessing;

[0007] S3, then directly process the collected image data, using superpixel segmentation technology and image change detection algorithm to divide the image into different areas, pay special attention to the key flood control locations of rivers, lakes, and dams, and identify changes in key areas, so as to detect possible flood conditions in advance;

[0008] S4. Finally, the big data platform is combined with real-time monitoring information to provide decision-making suggestions to help river and lake chiefs make accurate flood prevention decisions.

[0009] Preferably, the step S3 uses superpixel segmentation technology and image change detection algorithm to divide the image into different areas, and the implementation steps of identifying changes in key areas are:

[0010] S31, first calculate the gradient information of the input image in Represents the gradient of the image in the x and y directions, and adaptively adjusts the weight coefficients of the space and color distance based on the gradient information. The adaptive weight parameter λ(x,y) is introduced, which is adjusted according to the local gradient size: Among them, α is an adjustment factor used to control the influence of gradient on weight. When the gradient is large, the weight λ(x,y) will decrease, so that the weight of spatial distance is higher. The improved distance calculation formula is: where d spatial (p,q) is the distance between pixels p and q in space, d color (p,q) is their distance in color space;

[0011] S32, for each super pixel region R i , extract the color, texture, and edge features of the image, and use these feature information to classify each area into different categories. The extraction formula is: Feature(R i )=[Color(R i ),Texture(R i ),Edge(R i )], where Color(R i ),Texture(R i ),Edge(R i ) are color features, texture features and edge features respectively; then based on the extracted features, the superpixel area is classified by a support vector machine or other classification algorithm;

[0012] S33, finally, by calculating the super pixel area at each time point t The area at the previous time point t-1 Compare and determine whether the area has changed. The change detection formula is: When the difference value exceeds the threshold of 0.15, it is considered that the area has changed significantly;

[0013] S34. Finally, once changes in key areas are detected, further analysis is performed in combination with time series data to generate a trend chart. By tracking changes in multiple time periods, possible floods or other risk areas in the future can be predicted.

[0014] Preferably, the method for implementing image preprocessing using the improved dark light enhancement algorithm in step S2 is:

[0015] S211, first use the image light intensity histogram to calculate the global and local light intensity distribution, and set the image global brightness gain coefficient G L and the local brightness gain coefficient G R (x,y), the formula is as follows: Among them I max is the maximum brightness value of the image, I avg is the global average brightness, I max (x, y) is the maximum brightness of the local area, I local (x,y) is the average brightness of the local area;

[0016] S212, using an adaptive color balance algorithm to restore the color of the image and eliminate the color cast caused by insufficient light. Using the proportional relationship of the RGB channels, adjust the brightness of each channel to make it balanced. The formula is as follows: Where R avg , G avg , B avg is the average value of the RGB channels of the image;

[0017] S213. Finally, the Laplace operator is used to enhance the edge of the image in combination with the local contrast enhancement method. The formula is as follows: in is the Laplace operator, α is the coefficient for adjusting contrast enhancement, and I′(x, y) is the enhanced image.

[0018] Preferably, the method for implementing image preprocessing using the improved noise reduction algorithm in step S2 is:

[0019] S221, first perform wavelet transform on the image to decompose the image into a low-frequency sub-band L(x, y) and multiple high-frequency sub-bands H i (x,y), the formula is:

[0020] S222, for high frequency sub-band H i (x, y) performs adaptive threshold denoising and sets the denoising threshold T i , the formula is as follows: Where T i Adaptive adjustment according to the intensity of high-frequency noise;

[0021] S223, the processed high frequency sub-band H i ′(x,y) and the low-frequency subband L(x,y) are inversely transformed to reconstruct the denoised image. The formula is as follows:

[0022] S224. Finally, in order to further improve the denoising effect, the non-local means algorithm is used to refine the denoised image and maintain the texture details of the image. The formula is as follows: Where C(x,y) is the normalization constant, h is the smoothing parameter, Ω(x) is the search window, and j is the displacement of the pixel in the search window relative to the center pixel (x,y), which is used to calculate the similarity with the center pixel and determine the final denoised pixel value based on it.

[0023] Preferably, the decision-making suggestions provided in step S4 include formulating corresponding emergency plans for different risk levels, including personnel evacuation, material deployment, and equipment preparation, and providing resource deployment suggestions, such as mobilizing rescue personnel, materials, mechanical equipment, etc., so as to quickly respond to possible flood events. , Propose public warning suggestions, guide river and lake chiefs on how to issue alerts to the public, notify evacuation or other necessary protective measures, and recommend follow-up evaluation and summary after the flood situation ends, so as to improve future flood control strategies and emergency management measures.

[0024] Compared with the prior art, the advantages and positive effects of the present invention are as follows: first, the present invention realizes all-round and multi-angle image acquisition by combining the deployment of surveillance cameras at key locations of rivers and lakes with drone cruises, thereby obtaining more comprehensive information. Second, by utilizing improved dark light enhancement and noise reduction algorithms, the image quality in low light and severe weather is significantly improved, making key information clearer. Third, superpixel segmentation technology and image change detection algorithms can accurately identify changes in key flood control areas and are applicable to a variety of scenarios. Fourth, the decision-making recommendations provided by the big data platform are scientific and reasonable, which effectively guarantees the efficiency, accuracy and timeliness of flood control work and greatly improves the overall flood control capabilities. DETAILED DESCRIPTION

[0025] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described below in conjunction with embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0026] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways than those described herein. Therefore, the present invention is not limited to the specific embodiments of the following disclosure.

[0027] Embodiment: In past flood prevention practices, the monitoring model has faced many severe challenges. The manual inspection method consumes a lot of manpower and time. The inspectors work in complex terrain and severe weather conditions. It is not only inefficient, but also difficult to achieve a no-dead-angle, uninterrupted inspection of the entire basin, which often leads to gaps in inspections in some hidden areas or in extreme weather. The artificial intelligence-based river and lake chief system flood control monitoring image processing method of the present invention is implemented in this area, hoping to effectively solve the many problems existing in existing flood control monitoring methods, improve the intelligence level and efficiency of flood control monitoring in the basin, and effectively protect the safety of people’s lives and property and the stable development of the regional economy and society.

[0028] First, according to the geographical characteristics of the basin and the distribution of key flood control areas, the deployment locations of surveillance cameras were carefully planned to ensure that they could fully cover the key nodes of the river, the weak sections of the embankment, and the areas around the lakes that are prone to dangerous situations. At the same time, advanced drone equipment was equipped to perform patrol missions regularly to obtain high-resolution, multi-angle hydrological images, providing a rich source of data for subsequent precise image analysis.

[0029] The improved dark light enhancement algorithm and noise reduction algorithm are used for image preprocessing to improve the image quality in low light and bad weather conditions. The present invention combines local contrast enhancement with adaptive brightness enhancement and color restoration technology to improve the brightness and visible details of images in low light environments, avoiding overexposure or image distortion caused by traditional enhancement methods. First, the image light intensity histogram is used to calculate the global and local light intensity distribution, and the global brightness gain coefficient G of the image is set. L and the local brightness gain coefficient G R (x,y), the formula is as follows: Among them I max is the maximum brightness value of the image, I avg is the global average brightness, I max (x, y) is the maximum brightness of the local area, I local (x, y) is the average brightness of the local area; the adaptive color balance algorithm is used to restore the color of the image and eliminate the color cast caused by insufficient light. The proportional relationship of the RGB channels is used to adjust the brightness of each channel to make it balanced. The formula is as follows: Where R avg , G avg , B avg is the average value of the RGB channels of the image; finally, combined with the local contrast enhancement method, the Laplace operator is used to enhance the edge of the image. The formula is as follows: in is the Laplace operator, α is the coefficient for adjusting contrast enhancement, and I′(x, y) is the enhanced image.

[0030] Then, a denoising algorithm based on multi-scale analysis is introduced. The image is decomposed by wavelet transform, and the noise is concentrated in the high-frequency part for processing, while retaining the image details in the low-frequency part to avoid the loss of details caused by traditional denoising algorithms. First, the image is transformed by wavelet transform to decompose the image into a low-frequency subband L(x, y) and multiple high-frequency subbands H i (x,y), the formula is: For the high frequency subband H i (x, y) performs adaptive threshold denoising and sets the denoising threshold T i , the formula is as follows: Where T i Adaptively adjust according to the high-frequency noise intensity; i ′(x,y) and the low-frequency subband L(x,y) are inversely transformed to reconstruct the denoised image. The formula is as follows: Finally, in order to further improve the denoising effect, the non-local means algorithm is used to refine the denoised image and maintain the texture details of the image. The formula is as follows: Where C(x,y) is the normalization constant, h is the smoothing parameter, Ω(x) is the search window, and j is the displacement of the pixel in the search window relative to the center pixel (x,y), which is used to calculate the similarity with the center pixel and determine the final denoised pixel value based on it.

[0031] In order to achieve the segmentation of monitoring images and the changes in key areas of rivers and lakes, the present invention uses superpixel segmentation technology to locally divide the image, and uses feature extraction combined with classifiers to classify the regions, and finally detects regional changes through time series analysis. This method avoids the use of CNN models and is suitable for use in scenarios with limited computing resources or insufficient deep learning training data. First, calculate the gradient information of the input image in Represents the gradient of the image in the x and y directions, and adaptively adjusts the weight coefficients of the space and color distance based on the gradient information. The adaptive weight parameter λ(x,y) is introduced, which is adjusted according to the local gradient size: Among them, α is an adjustment factor used to control the influence of gradient on weight. When the gradient is large, the weight λ(x,y) will decrease, so that the weight of spatial distance is higher. The improved distance calculation formula is: where d spatial (p,q) is the distance between pixels p and q in space, d color (p, q) is their distance in color space; for each superpixel region R i , extract the color, texture, and edge features of the image, and use these feature information to classify each area into different categories. The extraction formula is: Feature(Ri )=[Color(R i ),Texture(R i ),Edge(R i )], where Color(R i ),Texture(R i ),Edge(R i ) are color features, texture features and edge features respectively; then, based on the extracted features, the super-pixel regions are classified by support vector machine or other classification algorithms; finally, the super-pixel regions at each time point t are classified by The area at the previous time point t-1 Compare and determine whether the area has changed. The change detection formula is: When the difference value exceeds the threshold of 0.15, it is considered that the area has changed significantly; finally, once the changes in the key areas are detected, further analysis is performed in combination with the time series data to generate a change trend chart. By tracking changes in multiple time periods, possible floods or other risk areas in the future can be predicted.

[0032] Finally, the big data platform is combined with real-time monitoring information to provide decision-making suggestions. According to different risk levels, corresponding emergency plans are formulated, including personnel evacuation, material deployment, and equipment preparation, and resource deployment suggestions are provided, such as mobilizing rescue personnel, materials, and mechanical equipment, so as to quickly respond to possible flood events. , Public warning suggestions are put forward to guide river and lake chiefs on how to issue alerts to the public, notify evacuations or other necessary protective measures. After the flood situation is over, follow-up evaluations and summaries are recommended to improve future flood control strategies and emergency management measures.

[0033] The above description is only a preferred embodiment of the present invention and does not limit the present invention in other forms. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A flood control monitoring image processing method based on artificial intelligence for river and lake chief system, characterized in that: The following steps are involved: S1. First, deploy multiple surveillance cameras at key locations of rivers and lakes to collect surveillance images in real time, and use drones for regular patrols to collect high-angle hydrological images; S2, secondly, the image quality under low light and bad weather conditions is improved by using improved dark light enhancement algorithm and noise reduction algorithm for image preprocessing; S3, then directly process the collected image data, use superpixel segmentation technology and image change detection algorithm to divide the image into different areas, pay special attention to the key flood control locations of rivers, lakes, and dams, and identify changes in key areas, so as to detect possible flood conditions in advance; S4. Finally, the big data platform is combined with real-time monitoring information to provide decision-making suggestions to help river and lake chiefs make accurate flood prevention decisions; The step S3 uses superpixel segmentation technology and image change detection algorithm to divide the image into different areas, and the implementation steps of identifying changes in key areas are as follows: S31, first calculate the gradient information of the input image in Represents the gradient of the image in the x and y directions, and adaptively adjusts the weight coefficients of the space and color distance based on the gradient information. The adaptive weight parameter λ(x,y) is introduced, which is adjusted according to the local gradient size: Among them, α is an adjustment factor used to control the influence of gradient on weight. When the gradient is large, the weight λ(x,y) will decrease, so that the weight of spatial distance is higher. The improved distance calculation formula is: where d spatial (p,q) is the distance between pixels p and q in space, d color (p,q) is their distance in color space; S32, for each super pixel region R i , extract the color, texture, and edge features of the image, and use these feature information to classify each area into different categories. The extraction formula is: Feature(R i )=[Color(R i ),Texture(R i ),Edge(R i )], where Color(R i ),Texture(R i ),Edge(R i ) are color features, texture features and edge features respectively; Then, based on the extracted features, the superpixel regions are classified by support vector machines or other classification algorithms; S33, finally, by calculating the super pixel area at each time point t The area at the previous time point t-1 Compare and determine whether the area has changed. The change detection formula is: When the difference value exceeds the threshold of 0.15, it is considered that the area has changed significantly; S34. Finally, once changes in key areas are detected, further analysis is performed in combination with time series data to generate a trend graph. By tracking changes in multiple time periods, possible floods or other risk areas in the future can be predicted.

2. According to the artificial intelligence-based flood control monitoring image processing method of river and lake chief system according to claim 1, it is characterized in that: The implementation method of performing image preprocessing using the improved dark light enhancement algorithm in step S2 is: S211, first use the image light intensity histogram to calculate the global and local light intensity distribution, and set the image global brightness gain coefficient G L and the local brightness gain coefficient G R (x,y), the formula is as follows: Among them I max is the maximum brightness value of the image, I avg is the global average brightness, I max (x, y) is the maximum brightness of the local area, I local (x,y) is the average brightness of the local area; S212, using an adaptive color balance algorithm to restore the color of the image and eliminate the color cast caused by insufficient light. Using the proportional relationship of the RGB channels, adjust the brightness of each channel to make it balanced. The formula is as follows: Where R avg , G avg , B avg is the average value of the RGB channels of the image; S213. Finally, the Laplace operator is used to enhance the edge of the image in combination with the local contrast enhancement method. The formula is as follows: in is the Laplace operator, α is the coefficient for adjusting contrast enhancement, and I′(x, y) is the enhanced image.

3. According to the artificial intelligence-based flood control monitoring image processing method of the river and lake chief system of claim 1, it is characterized in that: The implementation method of using the improved noise reduction algorithm to perform image preprocessing in step S2 is: S221, first perform wavelet transform on the image to decompose the image into a low-frequency sub-band L(x, y) and multiple high-frequency sub-bands H i (x,y), the formula is: Where i represents the direction of different high-frequency sub-bands; S222, for high frequency sub-band H i (x, y) performs adaptive threshold denoising and sets the denoising threshold T i , the formula is as follows: Where T i Adaptive adjustment based on high-frequency noise intensity; S223, the processed high frequency sub-band H i ′(x,y) and the low-frequency subband L(x,y) are inversely transformed to reconstruct the denoised image. The formula is as follows: S224. Finally, in order to further improve the denoising effect, the non-local means algorithm is used to refine the denoised image and maintain the texture details of the image. The formula is as follows: Where C(x,y) is the normalization constant, h is the smoothing parameter, Ω(x) is the search window, and j is the displacement of the pixel in the search window relative to the center pixel (x,y), which is used to calculate the similarity with the center pixel and determine the final denoised pixel value based on it.

4. According to the artificial intelligence-based flood control monitoring image processing method of the river and lake chief system of claim 1, it is characterized in that: The decision-making suggestions provided in step S4 include formulating corresponding emergency plans for different risk levels, including personnel evacuation, material deployment, and equipment preparation, and providing resource deployment suggestions, such as mobilizing rescue personnel, materials, mechanical equipment, etc., in order to quickly respond to possible flood events. , Propose public warning suggestions, guide river and lake chiefs on how to issue alerts to the public, notify evacuation or other necessary protective measures, and recommend follow-up evaluation and summary after the flood situation ends, so as to improve future flood control strategies and emergency management measures.

Images