Land resource intelligent patrol system

Abstract

The invention discloses an intelligent touring system for territorial resources, and relates to the technical field of intelligent touring. The land resource intelligent patrol system comprises an image acquisition preprocessing module, an image quality detection module, a target result prediction adjustment module and a target result verification optimization module. Wherein the image acquisition and preprocessing module is used for collecting land resource image data and preprocessing the land resource image data; the image quality detection module is used for detecting the image quality of the to-be-detected land resource image data; the target result prediction adjustment module is used for performing target prediction on the to-be-predicted territorial resource image data; and the target result verification optimization module is used for verifying the accuracy of target prediction and making a target verification adjustment measure according to a verification result. The problems of wrong report and missing report due to the fact that a complex natural environment background cannot be recognized in an existing territorial resource intelligent patrol system are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent inspection, and in particular to an intelligent inspection system for land resources. Background Art

[0002] In modern land and resources management, traditional monitoring methods have gradually shown their limitations. These methods usually include manual inspections and basic video monitoring systems, which are not only labor-intensive but also inefficient, especially when large areas or difficult-to-access areas need to be covered. In addition, these traditional methods have obvious deficiencies in data processing and real-time response, making it difficult to meet the rapidly changing needs of environmental protection and urban planning.

[0003] With the rapid development of remote sensing technology and geographic information systems, the concept of intelligent monitoring system has been introduced into the field of land and resources management. These technologies provide the possibility of real-time data collection, efficient data processing and intelligent decision support, greatly improving the scope, accuracy and efficiency of monitoring.

[0004] In the existing technology, automatic image analysis technology is often interfered by complex natural environment background when processing remote sensing data and ground monitoring images, resulting in a high false alarm rate. These false alarms are usually caused by the algorithm's inability to accurately distinguish between the target and the background, such as mistakenly identifying shadows, trees or other natural features as target objects. Such recognition errors not only waste resources, but may also cause unnecessary alarms, affecting the credibility and efficiency of the system.

[0005] At the same time, the problem of false negatives is equally serious, that is, the system fails to identify or mark the actual target object. This often occurs when the target has a high similarity with the background, resulting in the omission of key monitoring objects, and thus may miss important monitoring moments. This is particularly critical in the fields of environmental monitoring and homeland security, because ignoring these problems may lead to serious consequences.

[0006] Therefore, in response to the above problems, an intelligent land and resources inspection system is urgently needed. Summary of the invention

[0007] Technical issues solved

[0008] In view of the deficiencies in the prior art, the present invention provides an intelligent land and resources inspection system, which solves the problem of misreporting and missed reporting caused by the inability of the existing land and resources intelligent inspection system to identify complex natural environment backgrounds.

[0009] Technical Solution

[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions: a land and resources intelligent patrol system, including an image acquisition preprocessing module, an image quality detection module, a target result prediction and adjustment module and a target result verification and optimization module; wherein the image acquisition preprocessing module is used to collect land and resources image data, preprocess the land and resources image data, and mark the pre-processed land and resources image data as land and resources image data to be detected; the image quality detection module is used to detect the image quality of the land and resources image data to be detected, optimize and adjust the image of the land and resources image data to be detected according to the image quality detection result, and mark the land and resources image data to be detected after the image optimization and adjustment as land and resources image data to be predicted; the target result prediction and adjustment module is used to perform target prediction on the land and resources image data to be predicted, and make target prediction adjustment measures according to the target prediction results; the target result verification and optimization module is used to verify the accuracy of the target prediction, and make target verification adjustment measures according to the verification results.

[0011] Furthermore, the land and resources image data are preprocessed, and the preprocessed land and resources image data are marked as land and resources image data to be detected. The specific process is: acquiring land and resources image data, preprocessing the land and resources image data, the preprocessing process includes noise removal, image enhancement, color correction and size adjustment, and marking the preprocessed land and resources image data as land and resources image data to be detected.

[0012] Furthermore, the specific process of detecting the image quality of the land and resources image data to be detected is: obtaining the land and resources image data to be detected, extracting characteristic parameters in the land and resources image data to be detected, and obtaining the image quality detection value through comprehensive analysis of the characteristic parameters.

[0013] Furthermore, the specific process of extracting characteristic parameters from the land and resources image data to be detected and obtaining the image quality detection value by comprehensive analysis of the characteristic parameters is as follows: extracting the characteristic parameters of the land and resources image data to be detected, wherein the characteristic parameters include the image pixel intensity standard deviation, the image red channel standard deviation, the image green channel standard deviation, the image blue channel standard deviation, the total number of image gradients, the total number of image pixels, the image brightness average value and the image brightness standard deviation; and obtaining the image quality detection value by comprehensive calculation of the image pixel intensity standard deviation, the image red channel standard deviation, the image green channel standard deviation, the image blue channel standard deviation, the image gradient total number, the image pixels total number, the image brightness average value and the image brightness standard deviation.

[0014] Furthermore, the specific calculation formula of the image quality detection value is: Where IQI represents the image quality detection value, σ(I) represents the standard deviation of image pixel intensity, and ωc Weight factor of the standard deviation of image pixel intensity, σ(R) represents the standard deviation of the red channel of the image, σ(G) represents the standard deviation of the green channel of the image, σ(B) represents the standard deviation of the blue channel of the image, ω s The weight factor representing the image saturation, represents the total number of image gradients, N represents the total number of image pixels, ω d represents the weight factor of image sharpness, μ(I) represents the average image brightness, σ(I) represents the standard deviation of image brightness, ω d A weighting factor representing brightness.

[0015] Furthermore, the image optimization and adjustment of the land and resources image data to be detected is performed according to the image quality detection result. The specific process is: real-time comparison of the image quality detection value and the first threshold value, when the image quality detection value is higher than the first threshold value, entering the next stage; when the image quality detection value is less than or equal to the first threshold value, repeating the preprocessing process until the image quality detection value is increased to above the first threshold value.

[0016] Furthermore, the target result prediction and adjustment module is used to perform target prediction on the land and resources image data to be predicted, and to make target prediction adjustment measures according to the target prediction results. The specific process is: obtain the land and resources image data to be predicted, and predict the buildings in the land and resources image data to be predicted through a convolutional neural network to obtain a building prediction value; monitor the vegetation coverage changes and water and soil changes in the land and resources image data to be predicted through the normalized vegetation index, and use a neural network to predict the environmental changes in the land and resources image data to be predicted to obtain an environmental prediction value; use motion sensor data and video tracking technology to identify and predict specific behaviors in the land and resources image data to be predicted to obtain a behavior prediction value; obtain a target result prediction value through comprehensive calculation of the building prediction value, the environmental prediction value and the behavior prediction value; compare the target result prediction value with the second threshold in real time, when the target result prediction value is higher than the second threshold, no processing is performed, when the target result prediction value is less than or equal to the second threshold, increase the frequency of collecting land and resources image data, increase the preprocessing frequency, until the collected land and resources image data is increased to above the second threshold, and generate a target result prediction report.

[0017] Furthermore, the specific calculation formula of the target result prediction value is: TPS = ω b BDA+ω e ·EDA+ω bu ·BuDA; where TPS represents the target result prediction value, BDA represents the building prediction value, ω b represents the weight factor of the building detection value, EDA represents the environmental prediction value, ω e represents the weight factor of the environmental prediction value, BuDA represents the behavioral prediction value, ωbu A weighting factor representing the predicted value of a behavior.

[0018] Furthermore, the specific process of verifying the accuracy of the target prediction and making target verification adjustment measures based on the verification results is: obtaining a target result prediction report, calculating the prediction accuracy, prediction recall rate and prediction precision of the result prediction, and obtaining the target result verification value by comprehensive calculation of the prediction accuracy, prediction recall rate and prediction precision; comparing the target result verification value and a third threshold in real time, and when the target result verification value is greater than the third threshold, no processing is performed, and when the target result verification value is less than or equal to the third threshold, an alarm is automatically issued.

[0019] Furthermore, the specific calculation formula of the target result verification value is: TRVS=ω a ACC+ω r REC+ω p ·PRE; where TRVS represents the target result verification value, ACC represents the prediction accuracy, ω a represents the weight factor of prediction accuracy, REC represents the prediction recall, ω r represents the weight factor of the prediction recall rate, PRE represents the prediction precision, ω p The weight factor representing the prediction accuracy.

[0020] Beneficial Effects

[0021] The present invention has the following beneficial effects:

[0022] (1) The present invention introduces an image quality detection module so that the system can automatically optimize image quality, reducing the need for manual intervention. By automatically optimizing image quality, the accuracy of target recognition is further improved, effectively solving the problem of misrecognition caused by poor image quality in the prior art.

[0023] (2) The present invention can adjust the prediction parameters according to the real-time image analysis results through the target result prediction adjustment module to optimize the prediction model. This dynamic adjustment mechanism can significantly reduce false positives and false negatives, and effectively solve the problem of insufficient adaptability caused by fixed algorithm parameters in the prior art.

[0024] (3) The present invention enables the system to verify the accuracy of the prediction results in real time through the target result verification optimization module, and adjust the system settings accordingly. This instant feedback mechanism further realizes efficient real-time monitoring, effectively solving the problems of delayed response and long processing time in the prior art.

[0025] (4) The present invention integrates advanced image preprocessing techniques, including noise removal, image enhancement, and color correction, so that the system can process input images of various qualities to ensure the accuracy of analysis. This comprehensive preprocessing process enables more accurate image interpretation and effectively solves the problem in the prior art that the accuracy of analysis results is affected by image quality issues.

[0026] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a structural diagram of a land resources intelligent inspection system of the present invention. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] See also Figure 1 The embodiment of the present invention provides a technical solution: a land and resources intelligent patrol system, including an image acquisition preprocessing module, an image quality detection module, a target result prediction adjustment module and a target result verification optimization module; wherein the image acquisition preprocessing module is used to collect land and resources image data, preprocess the land and resources image data, and mark the pre-processed land and resources image data as land and resources image data to be detected; the image quality detection module is used to detect the image quality of the land and resources image data to be detected, optimize and adjust the image of the land and resources image data to be detected according to the image quality detection result, and mark the land and resources image data to be detected after the image optimization adjustment as land and resources image data to be predicted; the target result prediction adjustment module is used to perform target prediction on the land and resources image data to be predicted, and make target prediction adjustment measures according to the target prediction results; the target result verification optimization module is used to verify the accuracy of the target prediction, and make target verification adjustment measures according to the verification results.

[0030] Specifically, the land and resources image data is preprocessed, and the preprocessed land and resources image data is marked as the land and resources image data to be detected. The specific process is as follows: first, the land and resources image data is obtained. These data are usually from satellites, drones or ground monitoring equipment, covering a variety of terrains from urban buildings to natural landscapes. After collecting the land and resources image data, the land and resources image data is preprocessed to ensure the accuracy and efficiency of image analysis. The preprocessing process includes the key steps of noise removal, image enhancement, color correction and size adjustment. Noise removal aims to remove random interference introduced in the image due to sensor defects, environmental conditions or transmission errors; image enhancement makes the important features in the image more prominent by adjusting the contrast and brightness of the image; color correction ensures that the color of the image reflects the color of the real world, which is particularly important for the subsequent natural resource classification; size adjustment is to unify the resolution and size of the image for subsequent automated processing. After completing these steps, the preprocessed land and resources image data is marked as the land and resources image data to be detected for further detailed analysis and detection. This entire preprocessing process not only improves the efficiency of data processing, but also greatly enhances the availability of image data and the accuracy of analysis results.

[0031] In this implementation, a thorough preprocessing step for land and resources image data is extremely critical, and this process significantly improves the accuracy and efficiency of subsequent image analysis. By removing noise, enhancing images, correcting colors, and adjusting the size of the original image data, not only is the visual quality of the image optimized, but it also ensures that the image features are more obvious and the colors are closer to reality, thereby providing higher reliability for automatic recognition and classification. In addition, the unified image size and format simplifies the subsequent processing steps and increases the speed at which the system processes data. This preprocessing process not only reduces errors in subsequent analysis, but also effectively improves the response speed of the entire monitoring system and the quality of decision support, making land and resources management more efficient and accurate.

[0032] Specifically, the specific process of detecting the image quality of the land and resources image data to be detected is as follows: the land and resources image data to be detected is obtained from the real-time data stream. These data come from a variety of sensors, including satellites, drones, and ground surveillance cameras, covering a wide range of geographical areas and diverse environmental conditions. After obtaining the land and resources image data to be detected, the key feature parameters in the land and resources image data to be detected are extracted, and the image quality detection value is calculated through a comprehensive analysis of these feature parameters.

[0033] In this embodiment, through these complex analyses, the system can determine whether the image meets the quality standards for further processing and analysis, and then guide whether the image needs to be further optimized to ensure that the image data can provide the highest information value and accuracy in subsequent resource monitoring and management applications.

[0034] Specifically, characteristic parameters in the land and resources image data to be detected are extracted, and the image quality detection value is obtained through comprehensive analysis of the characteristic parameters. The specific process is: extract the characteristic parameters of the land and resources image data to be detected, wherein the characteristic parameters include the standard deviation of image pixel intensity, the standard deviation of image red channel, the standard deviation of image green channel, the standard deviation of image blue channel, the total number of image gradients, the total number of image pixels, the average image brightness and the standard deviation of image brightness; these parameters each reflect the characteristics of the image in color distribution, texture details and brightness changes, and are important bases for evaluating image quality. The image quality detection value is obtained by comprehensive calculation of the standard deviation of image pixel intensity, the standard deviation of image red channel, the standard deviation of image green channel, the standard deviation of image blue channel, the total number of image gradients, the total number of image pixels, the average image brightness and the standard deviation of image brightness.

[0035] In this implementation, a comprehensive image quality detection value is obtained in this way, which reflects whether the image meets the quality requirements for further analysis and processing. Through such a comprehensive analysis, the land and resources intelligent inspection system can ensure that the quality of the data meets the standards before the data enters a deeper level of analysis, thereby improving the monitoring and analysis efficiency of the entire system.

[0036] Specifically, the specific calculation formula of the image quality detection value is: Where IQI represents the image quality detection value, which is a comprehensive indicator used to evaluate the overall quality of the image, σ(I) represents the standard deviation of the image pixel intensity, which measures the variability of the overall brightness of the image, and ω c The weighting factor for the standard deviation of the image pixel intensity usually ranges from 0 to 1. Its exact value is adjusted according to the specific application requirements to ensure that the relative importance of the variability of image brightness in the overall quality score is appropriately reflected. σ(R) represents the standard deviation of the image red channel, σ(G) represents the standard deviation of the image green channel, σ(B) represents the standard deviation of the image blue channel, and ω s The weight factor representing the image saturation is also adjusted in the range of 0 to 1 and is used to measure the impact of color saturation on image quality. represents the total number of image gradients, which is an indicator of image sharpness and detail. N represents the total number of image pixels, which is used to normalize the gradient count so that the sharpness of the image can be fairly evaluated regardless of the image size. dIt represents the weight factor of image sharpness, which determines the proportion of image details in the image quality score. μ(I) represents the average image brightness. These two parameters together reflect the consistency and variation range of image brightness. σ(I) represents the standard deviation of image brightness. d A weighting factor for brightness, adjusted to reflect the importance of brightness consistency in image quality scoring.

[0037] In this implementation scheme, through a carefully designed image quality detection formula, the land and resources intelligent patrol system can comprehensively consider multiple key image parameters to evaluate its quality, thereby significantly improving the accuracy and reliability of subsequent image analysis. This step utilizes multiple metrics such as pixel intensity standard deviation, color channel standard deviation, image gradient and brightness variation, which not only ensures the visual consistency and clarity of the image, but also reflects the relative importance of different image features in practical applications through the precise adjustment of weight factors. In addition, by automatically calculating the image quality detection value, the system can effectively screen out images of poor quality to prevent these images from further affecting data analysis and result judgment, thereby enhancing the operational efficiency and decision support capabilities of the entire monitoring system. The implementation of this method not only improves the speed of data processing, but also ensures that high-quality images can be obtained under various environmental conditions, making land and resources management more accurate and efficient.

[0038] Specifically, the image of the land and resources image data to be detected is optimized and adjusted according to the image quality detection results. The specific process is: real-time monitoring and comparison of the image quality detection value and the set first threshold. This first threshold is preset according to system requirements and actual application experience, and is used to determine whether the image data has reached sufficient quality standards for subsequent analysis. When the image quality detection value is greater than the first threshold, the system determines that the image quality meets the requirements, and the image data immediately enters the next analysis stage. This means that the clarity, color accuracy and detail capture of the image have met the basic requirements of system processing, which can ensure the accuracy and reliability of subsequent analysis. On the contrary, when the image quality detection value is less than or equal to the first threshold, the system automatically triggers the image reprocessing mechanism. This includes re-executing the preprocessing process, enhancing the image contrast, adjusting the brightness and color balance, and applying various image sharpening technologies to improve the overall quality of the image. This process will be repeated until the image quality detection value is raised above the first threshold.

[0039] In this implementation scheme, the process of comparing the image quality detection value with the preset first threshold in real time and adjusting the image data according to the comparison results significantly improves the accuracy and efficiency of the system in processing data. Through this mechanism, the system can ensure that all images that enter deeper analysis meet the minimum quality standards, thereby greatly reducing analysis errors caused by insufficient image quality. When the image quality does not meet the standard, the system automatically triggers the repetition of the preprocessing steps, enhancement and color correction until the image quality exceeds the threshold. This not only ensures the high quality standards of the data, but also optimizes the efficiency of resource use and avoids wasting computing resources on low-quality data. Through this real-time quality control and automatic adjustment mechanism, the land and resources intelligent patrol system can perform its monitoring and analysis tasks more reliably, ensuring the accuracy of the results and the overall performance of the system.

[0040] Specifically, the target result prediction and adjustment module is used to perform target prediction on the land and resources image data to be predicted, and to make target prediction adjustment measures according to the target prediction results. The specific process is as follows: First, the buildings in the land and resources image data to be predicted are predicted using a convolutional neural network to obtain the building prediction value. The convolutional neural network is an efficient image analysis tool that can identify the building structure and layout in the image, which is very important for planning and supervision. Next, the vegetation coverage changes and water and soil changes in the land and resources image data to be predicted are monitored by the normalized vegetation index, and these changes are analyzed by another neural network to obtain the environmental prediction value. The normalized vegetation index is an important indicator for judging the health of vegetation, which is particularly critical for environmental protection and agricultural management. In addition, motion sensor data and video tracking technology are used to identify and predict specific behaviors in the land and resources image data to be predicted, such as illegal logging or illegal construction, and then generate behavior prediction values. This step is of great significance for maintaining law and order and protecting natural resources. The target result prediction value is obtained by comprehensive calculation of the building prediction value, the environmental prediction value and the behavior prediction value. The target result prediction value is compared with the second threshold in real time. When the target result prediction value is greater than the second threshold, the system considers that the current prediction is accurate enough and no further processing is required. When the target result prediction value is less than or equal to the second threshold, the collection frequency and preprocessing frequency of land and resources image data are automatically increased until the target result prediction value is raised above the second threshold. This process ensures that all key indicators are fully monitored and updated in a timely manner. The target result prediction report generated in the end will provide decision makers with accurate data support to make more effective management decisions.

[0041] In this implementation scheme, the application of the target result prediction and adjustment module is extremely critical in the land and resources intelligent patrol system, which effectively improves the prediction accuracy and response efficiency of monitoring data. By integrating convolutional neural networks, normalized vegetation index and motion sensing technology, the module can not only accurately predict building locations, environmental changes and specific behaviors, but also evaluate their compliance with preset thresholds through real-time analysis of these data. When the target result prediction value does not reach the threshold, the system automatically increases the frequency of data collection and processing to ensure that all analyses are based on the latest and most accurate data, thereby greatly improving the timeliness and effectiveness of decision-making. This dynamic adjustment mechanism ensures that the system can maintain high performance in various monitoring scenarios, which is particularly valuable for rapidly changing environmental conditions and emergencies, and greatly enhances the ability and flexibility of land and resources management.

[0042] Specifically, the target result prediction value is calculated as follows: TPS = ω b BDA+ω e ·EDA+ω bu BuDA; where TPS represents the target result prediction value, which is used to evaluate the overall prediction of the land and resources status, BDA represents the building prediction value, which predicts the possibility of the existence of a building based on the image data, and ω b It represents the weight factor of the building detection value, which is usually selected in the range of 0 to 1. The specific value is adjusted according to the importance of the building prediction to the overall analysis. EDA represents the environmental prediction value, which reflects the environmental status obtained by analyzing environmental monitoring data such as vegetation index. e represents the weight factor of the environmental prediction value, which is also adjusted between 0 and 1 to reflect the impact of environmental changes on land and resources monitoring. BuDA represents the behavior prediction value, which is a specific behavior predicted based on motion sensor data and video tracking technology. bu The weight factor that represents the behavior prediction value is adjusted according to the priority of behavior monitoring in the entire system. These weight factors are obtained from the database. The setting of the weight factor allows the system administrator to adjust the influence of each prediction value according to specific monitoring needs and policy priorities, thereby ensuring that the prediction results output by the system reflect the specific conditions of each aspect and comply with the overall monitoring and response strategy.

[0043] In this embodiment, by using the calculation formula of the target result prediction value, it is possible to integrate and analyze the prediction data from different sources, including building existence, environmental changes, and the dynamics of specific behaviors. This calculation model realizes a comprehensive evaluation index by multiplying and accumulating different prediction values ​​with their respective weight factors, which not only improves the accuracy of the prediction, but also enhances the effectiveness of decision support. The setting of weight factors allows the system to flexibly adjust the importance of each prediction value, so that the system output can be optimized according to the actual monitoring needs and predetermined strategies. The effect of this step is that the system can process complex data sets more effectively, provide more accurate prediction results, and formulate corresponding management and response measures based on these results, thereby achieving efficient and high-performance operations in real-time monitoring and resource management.

[0044] Specifically, verify the accuracy of the target prediction and make target verification adjustment measures based on the verification results. The specific process is: obtain the target result prediction report, calculate the prediction accuracy, prediction recall and prediction precision of the result prediction, and these three statistical indicators are standard methods for evaluating prediction performance: accuracy measures the proportion of correct predictions, recall evaluates the proportion of relevant instances identified by the system to all relevant instances, and precision is the proportion of correctly predicted positive examples to all predicted positive examples. The target result verification value is obtained by comprehensive calculation of prediction accuracy, prediction recall and prediction precision. Target result verification value. Real-time comparison of the target result verification value and the third threshold. The third threshold is pre-set based on the performance requirements of the system to determine the acceptability of the prediction result. When the target result verification value is greater than the third threshold, the system determines that the prediction result is reliable, so no additional processing is performed. However, when the target result verification value is less than or equal to the third threshold, the system will automatically issue an alarm. This alarm mechanism is to immediately notify system administrators or relevant personnel to pay attention to potential problems, such as poor prediction performance, and further inspection or adjustment of system parameters is required. This automated response process ensures that the system can react quickly in the face of predicted performance degradation, thereby maintaining the efficiency and accuracy of the overall monitoring system.

[0045] In this implementation scheme, the reliability and utility of the system are greatly enhanced by the process of verifying the accuracy of the target prediction and adjusting it according to the results. This process uses the three key performance indicators of prediction accuracy, recall rate and precision to comprehensively calculate a target result verification value, which effectively measures the overall performance of the system prediction. By comparing the target result verification value with the set third threshold in real time, the system can automatically determine the credibility of the prediction result. When the target result verification value is greater than the third threshold, the prediction result is considered reliable and no additional intervention is required; when the target result verification value is less than or equal to the third threshold, the system will automatically trigger an alarm to remind the operator or automatically adjust the system configuration to deal with potential prediction errors. This mechanism not only ensures the efficient execution of monitoring activities, but also ensures a rapid response in critical situations, thereby playing a key role in protecting land resources and optimizing resource management.

[0046] Specifically, the target result verification value is calculated as follows: TRVS = ω a ACC+ω r REC+ω p ·PRE; where TRVS represents the target result verification value, which is a quantitative indicator used to evaluate the comprehensive quality of the prediction results, ACC represents the prediction accuracy, which measures the proportion of correctly predicted instances to the total number of predictions, and ω a Represents the weight factor of prediction accuracy, usually set between 0 and 1. Its specific value depends on the importance of accuracy in specific applications. REC represents the prediction recall rate, which measures the proportion of correctly predicted positive examples to all actual positive examples. ω r It represents the weight factor of the prediction recall rate, which is also adjusted between 0 and 1 to reflect the importance of the recall rate in the overall evaluation system. PRE represents the prediction precision, which indicates the proportion of correctly predicted positive examples in all predicted positive examples. p Represents the weight factor of the prediction accuracy. This parameter is also adjusted according to the priority of accuracy in a specific monitoring task. These weight factors are obtained from the database.

[0047] In this implementation scheme, by accurately calculating the target result verification value, the system can comprehensively consider the prediction accuracy, recall rate and precision rate to obtain a comprehensive performance indicator. This calculation method uses weight factors to adjust the influence of each evaluation indicator to ensure that the evaluation process meets the priority and sensitivity of specific monitoring needs. For example, by increasing the weight of accuracy, the system can place special emphasis on the importance of correct prediction; while increasing the weight of recall rate helps to ensure that no key events are missed. Such an evaluation mechanism not only improves the reliability of the prediction results, but also optimizes the system's allocation and response strategy for resources, ensuring that decisions and adjustments can be made quickly and effectively in actual operations. Through this method, the overall performance of the system is significantly improved, and it can respond and handle various land and resources monitoring scenarios more accurately, thereby effectively improving management efficiency and resource protection effects.

[0048] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0049] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A land resources intelligent inspection system, characterized in that: It includes an image acquisition preprocessing module, an image quality detection module, a target result prediction adjustment module and a target result verification optimization module; The image acquisition and preprocessing module is used to collect land and resources image data, preprocess the land and resources image data, and mark the preprocessed land and resources image data as land and resources image data to be detected; The image quality detection module is used to detect the image quality of the land and resources image data to be detected, optimize and adjust the image of the land and resources image data to be detected according to the image quality detection result, and mark the land and resources image data to be detected after image optimization and adjustment as the land and resources image data to be predicted; The target result prediction adjustment module is used to perform target prediction on the land and resources image data to be predicted, and make target prediction adjustment measures according to the target prediction results; The target result verification optimization module is used to verify the accuracy of the target prediction and make target verification adjustment measures according to the verification results.

2. According to claim 1, the intelligent land resources inspection system is characterized by: The specific process of preprocessing the land and resources image data and marking the preprocessed land and resources image data as the land and resources image data to be detected is as follows: Acquire land and resources image data, and preprocess the land and resources image data. The preprocessing process includes noise removal, image enhancement, color correction and size adjustment. Mark the preprocessed land and resources image data as land and resources image data to be detected.

3. The intelligent land resources inspection system according to claim 1 is characterized in that: The specific process of detecting the image quality of the land and resources image data to be detected is as follows: The image data of the land and resources to be detected are obtained, the characteristic parameters in the image data of the land and resources to be detected are extracted, and the image quality detection value is obtained through comprehensive analysis of the characteristic parameters.

4. The intelligent land resources inspection system according to claim 3 is characterized in that: The specific process of extracting characteristic parameters from the land and resources image data to be detected and obtaining the image quality detection value through comprehensive analysis of the characteristic parameters is as follows: Extract characteristic parameters of the land resources image data to be detected, wherein the characteristic parameters include the standard deviation of image pixel intensity, the standard deviation of the image red channel, the standard deviation of the image green channel, the standard deviation of the image blue channel, the total number of image gradients, the total number of image pixels, the average image brightness, and the standard deviation of image brightness; The image quality detection value is obtained by comprehensively calculating the image pixel intensity standard deviation, image red channel standard deviation, image green channel standard deviation, image blue channel standard deviation, total image gradients, total image pixels, image brightness average and image brightness standard deviation.

5. The intelligent land resources inspection system according to claim 3 is characterized in that: The specific calculation formula of the image quality detection value is: Where IQI represents the image quality detection value, σ(I) represents the standard deviation of image pixel intensity, and ω c Weight factor of the standard deviation of image pixel intensity, σ(R) represents the standard deviation of the red channel of the image, σ(G) represents the standard deviation of the green channel of the image, σ(B) represents the standard deviation of the blue channel of the image, ω s represents the weight factor of the image saturation, represents the total number of image gradients, N represents the total number of image pixels, ω d represents the weight factor of image sharpness, μ(I) represents the average image brightness, σ(I) represents the standard deviation of image brightness, ω d A weighting factor representing brightness.

6. The intelligent land resources inspection system according to claim 1 is characterized in that: The image quality detection result is used to optimize and adjust the land and resources image data to be detected, and the specific process is as follows: Compare the image quality detection value with the first threshold in real time, and when the image quality detection value is higher than the first threshold, enter the next stage; When the image quality detection value is less than or equal to the first threshold, the preprocessing process is repeatedly performed until the image quality detection value is increased to above the first threshold.

7. The intelligent land resources inspection system according to claim 1 is characterized in that: The target result prediction and adjustment module is used to perform target prediction on the land and resources image data to be predicted, and to make target prediction adjustment measures according to the target prediction results. The specific process is as follows: Obtain the image data of land resources to be predicted, predict the buildings in the image data of land resources to be predicted through a convolutional neural network, and obtain the predicted value of the building; The vegetation coverage change and water and soil change in the land and resources image data to be predicted are monitored by the normalized vegetation index, and the environmental change in the land and resources image data to be predicted is predicted by using a neural network to obtain the environmental prediction value; Use motion sensor data and video tracking technology to identify and predict specific behaviors in the land and resources image data to be predicted, and obtain behavior prediction values; The target result prediction value is obtained by comprehensive calculation of the building prediction value, the environment prediction value and the behavior prediction value; The target result prediction value is compared with the second threshold in real time. When the target result prediction value is higher than the second threshold, no processing is performed. When the target result prediction value is less than or equal to the second threshold, the frequency of collecting land and resources image data is increased, and the preprocessing frequency is increased until the collected land and resources image data is increased to above the second threshold, and a target result prediction report is generated.

8. The intelligent land resources inspection system according to claim 7 is characterized in that: The specific calculation formula of the target result prediction value is: TPS=ω b ·BDA+ω e ·EDA+ω bu ·BuDA; In the formula, TPS represents the target result prediction value, BDA represents the building prediction value, ω b represents the weight factor of the building detection value, EDA represents the environmental prediction value, ω e represents the weight factor of the environmental prediction value, BuDA represents the behavioral prediction value, ω bu A weighting factor representing the predicted value of a behavior.

9. The intelligent land resources inspection system according to claim 1 is characterized by: The specific process of verifying the accuracy of the target prediction and making target verification adjustment measures based on the verification results is as follows: Obtain the target result prediction report, calculate the prediction accuracy, prediction recall rate and prediction precision of the result prediction, and obtain the target result verification value through comprehensive calculation of the prediction accuracy, prediction recall rate and prediction precision; The target result verification value and the third threshold are compared in real time. When the target result verification value is greater than the third threshold, no processing is performed. When the target result verification value is less than or equal to the third threshold, an alarm is automatically issued.

10. The intelligent land resources inspection system according to claim 9, characterized in that: The specific calculation formula of the target result verification value is: TRVS=ω a ·ACC+ω r ·REC+ω p ·PRE; In the formula, TRVS represents the target result verification value, ACC represents the prediction accuracy, ω a represents the weight factor of prediction accuracy, REC represents the prediction recall, ω r represents the weight factor of the prediction recall rate, PRE represents the prediction precision, ω p The weight factor representing the prediction accuracy.

Images