Airborne Surface Mine and Unexploded Ordnance Detection System Based on Image Recognition and Its Method

Abstract

The present invention discloses an airborne surface mine and unexploded ordnance detection system and method based on image recognition, specifically relating to the technical field of image recognition. By monitoring the wireless signal strength in real time, analyzing the signal fluctuations and image transmission delay, evaluating the impact of the delay on the image resolution and detail retention, and then analyzing the compression quality of the image, according to the degree of image quality degradation, the images are classified into two categories: slightly degraded and severely degraded, and corresponding processing measures are taken: for slightly degraded images, the image quality is improved through local enhancement; for severely degraded images, by predicting and warning the accuracy of the image recognition algorithm, the algorithm is automatically adjusted to improve the detection accuracy, which can effectively improve the accuracy of surface mine and unexploded ordnance detection and ensure real-time and stable image recognition can still be achieved in complex environments.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and particularly to an airborne surface mine and unexploded ordnance detection system and method based on image recognition. Background Art

[0002] In military training and infrastructure construction, surface mines and unexploded ordnance (UXO) are a common safety hazard. Traditional methods for removing surface mines and unexploded ordnance rely on manual approach and detection, which not only poses a significant safety risk but also limits the efficiency and accuracy of the removal operation. With the development of unmanned aerial vehicle (UAV) technology, combined with remote control and aerial image acquisition capabilities, UAVs have become an important tool for removing and destroying surface mines and unexploded ordnance. UAVs can perform global or local image acquisition in dangerous areas and use advanced image recognition technology to identify the types and status of surface mines and unexploded ordnance in real time, providing an accurate basis for the removal and destruction plan. In addition, UAVs can perform efficient risk assessment and surveillance under unmanned operation conditions, greatly improving the safety of explosive ordnance disposal personnel.

[0003] The introduction of image recognition technology has further improved the intelligence level of the UAV explosive ordnance disposal system. Through high-resolution image transmission and database comparison, the system can quickly identify and classify different types of surface mines and unexploded ordnance, determine their hazards and evaluate the difficulty of disposal. This UAV-borne detection system based on image recognition can not only reduce manual intervention but also accurately obtain target information in the shortest time. When performing the removal task, the UAV can, according to the real-time image data, dispatch appropriate shaped charges for induced detonation and destruction, effectively avoiding the danger of personnel directly contacting surface mines and unexploded ordnance, and providing precise and safe technical support for the disposal of surface mines and unexploded ordnance.

[0004] The existing technology has the following deficiencies:

[0005] During the wireless transmission of high-resolution images, the occupied bandwidth is very large. Especially during long-distance operation, it may not be possible to ensure real-time and stable data transmission. Since the UAV may be in a high flight state, the signal between the ground receiving device and the UAV may be affected by terrain, weather, and other wireless signal interferences. Delays may occur during the transmission process, and even data loss or compression may occur, resulting in a decrease in the quality of the received images. In addition, the decrease in image quality will directly affect the accuracy of the image recognition algorithm, which may cause surface mines and unexploded ordnance to be misidentified as other objects or the recognition to fail. Summary of the Invention

[0006] The purpose of the present invention is to provide an airborne surface mine and unexploded ordnance detection system and method based on image recognition to solve the deficiencies in the background art.

[0007] To achieve the above object, the present invention provides the following technical solution: An airborne surface mine and unexploded ordnance detection method based on image recognition, comprising the following steps:

[0008] S1: The unmanned aerial vehicle (UAV) flies over the target area through a remote command system, and the UAV acquires the image information of surface mines, unexploded ordnance and their surrounding environment in real time through a high-resolution camera carried thereon, and measures and records the wireless communication signal strength between the UAV and the ground receiver in real time;

[0009] S2: Compare and analyze the signal strength fluctuation and the time delay of image transmission, judge the interference degree of the delay on the image resolution and detail preservation, and evaluate the influence of the delay on the image real-time performance;

[0010] S3: If the delay has a serious impact, analyze the storage compression ratio information of each frame of image during the image compression process, compare the differences between the compressed image and the original image, and evaluate the compression quality of the image;

[0011] S4: Evaluate the degree of image quality degradation according to the influence of the delay on the image real-time performance and the image compression quality, and divide the degree of image quality degradation into serious degradation and slight degradation according to the evaluation results;

[0012] S5: For slight degradation, enhance the image locally to improve the image quality, making the features of surface mines and unexploded ordnance clearer; for serious degradation, predict and warn the accuracy of the image recognition algorithm within a fixed time period, and automatically adjust the image recognition algorithm according to the prediction results to improve the accuracy of surface mine and unexploded ordnance detection.

[0013] Preferably, in S2, after comparing the degree of detail loss of the compressed image and the original image due to the image transmission time delay, a detail loss anomaly index is generated. The method for obtaining the detail loss anomaly index is as follows:

[0014] Obtain the original image and the compressed image, match the sizes of the two images, perform gray-scale processing on the two images. The structural similarity index SSIM is used to measure the similarity between the two images, considering three aspects: brightness, contrast and structure. The formula is: In the formula, x and y are the corresponding pixel values of the original image and the compressed image, μ x and μ y are the local means of the original image and the compressed image, that is, the average value of the image block, and are the local variances of the original image and the compressed image, representing the degree of brightness change in the local area, σ xyis the local covariance of the original image and the compressed image, representing the degree of similarity between the two images; C1 and C2 are constants used to avoid a zero denominator, where L is the dynamic range of the image, and K1 and K2 are small constants; according to the SSIM value, the Detail Loss Anomaly Index (DLI) is calculated to measure the degree of detail loss caused by image transmission delay and compression. The expression is: SSIM threshold is a preset SSIM threshold.

[0015] Preferably, in S2, after analyzing the fluctuation of the compression ratio of each frame of the image, an Image Compression Ratio Fluctuation Index is generated. The method for obtaining the Image Compression Ratio Fluctuation Index is as follows:

[0016] For each frame of the image, calculate its compression ratio CR i , and record it. Assuming the image sequence has N frames, the compression ratio data is {CR1, CR2, …, CR N}; calculate the frequency distribution of all compression ratios to obtain the occurrence probability p(CR i ) of each compression ratio CR i in the entire dataset; the calculation formula for the frequency distribution is: where the frequency (CR i ) is the number of times the compression ratio CR i appears, N is the total number of frames of the compression ratio data, and the entropy is calculated through the probability distribution of the compression ratio. The expression is: where p(CR i ) is the probability of the compression ratio CR i , n is the number of different compression ratio values, H(CR) is the entropy of the compression ratio, and the Image Compression Ratio Fluctuation Index is calculated. The expression is: CRFI = H(CR); in the formula, CRFI is the Image Compression Ratio Fluctuation Index.

[0017] Preferably, in S4, according to the impact of the delay on the real-time performance of the image and the image compression quality, the degree of image quality degradation is evaluated. According to the evaluation results, the degree of image quality degradation is divided into severe degradation and slight degradation, specifically:

[0018] Convert the Detail Loss Anomaly Index and the Image Compression Ratio Fluctuation Index into a comprehensive feature vector, and use the comprehensive feature vector as the input of the machine learning model. The machine learning model takes predicting the value label of the degree of image quality degradation for each group of comprehensive feature vectors as the prediction target, and takes minimizing the sum of the prediction errors of the value labels of the degree of image quality degradation for all images as the training target. Train the machine learning model until the sum of the prediction errors reaches convergence and then stop the model training. Determine the value of the degree of image quality degradation according to the model output result, where the machine learning model is a polynomial regression model.

[0019] Preferably, the degree of image quality degradation obtained is compared with the reference threshold of the degree of image quality degradation set according to historical data. If the degree of image quality degradation is greater than or equal to the reference threshold of the set degree of image quality degradation, it indicates that the degree of image quality degradation is severe. At this time, a severe image quality degradation signal is generated, and the degree of image quality degradation is classified as severe degradation. If the degree of image quality degradation is less than the reference threshold of the set degree of image quality degradation, it indicates that the degree of image quality degradation is slight. At this time, a slight image quality degradation signal is generated, and the degree of image quality degradation is classified as slight degradation.

[0020] Preferably, in S5, for severe degradation, that is, the degree of image quality degradation generated within a fixed time period is greater than or equal to the reference threshold of the set degree of image quality degradation, the degree of quality degradation greater than or equal to the reference threshold of the degree of image quality degradation generated within the subsequent fixed time period is collected, and a corresponding data set is established. The mean and standard deviation of the data set are calculated to predict and warn the accuracy of the image recognition algorithm within the fixed time period.

[0021] Preferably, if the mean value of the degree of quality degradation in the data set is greater than or equal to the reference threshold of the mean value of the degree of quality degradation, and the standard deviation of the degree of quality degradation is less than the reference threshold of the standard deviation of the degree of quality degradation, a first-level warning signal is generated at this time, and it is necessary to immediately adjust the image recognition algorithm to increase the fault tolerance of image processing;

[0022] If the mean value of the degree of quality degradation is greater than or equal to the reference threshold of the mean value of the degree of quality degradation, and the standard deviation of the degree of quality degradation is greater than or equal to the reference threshold of the standard deviation of the degree of quality degradation, a second-level warning signal is generated at this time, and partial algorithm optimization and parameter adjustment are required;

[0023] If the mean value of the degree of quality degradation is less than the reference threshold of the mean value of the degree of quality degradation, and the standard deviation of the degree of quality degradation is greater than or equal to the reference threshold of the standard deviation of the degree of quality degradation, a third-level warning signal is generated at this time, and it is necessary to optimize the image transmission path;

[0024] If the mean value of the degree of quality degradation is less than the reference threshold of the mean value of the degree of quality degradation, and the standard deviation of the degree of quality degradation is less than the reference threshold of the standard deviation of the degree of quality degradation, no warning signal is generated at this time, and there is no need to adjust the image recognition algorithm, and the current operation is continued.

[0025] The present invention also provides an airborne surface mine and unexploded ordnance detection system based on image recognition, including an image acquisition and signal monitoring module, a signal delay evaluation module, an image compression quality analysis module, an image quality classification module, and an image enhancement and recognition adjustment module:

[0026] Image Acquisition and Signal Monitoring Module: The drone flies over the target area through the remote control command system, and uses the equipped high-resolution camera to collect real-time image information of surface mines, unexploded ordnance and their surrounding environment, and measures and records the real-time strength of the wireless communication signal between the drone and the ground receiver;

[0027] Signal Delay Evaluation Module: Compare and analyze the signal strength fluctuations and the time delay of image transmission, judge the degree of interference of the delay on image resolution and detail preservation, and evaluate the impact of the delay on image real-time performance;

[0028] Image Compression Quality Analysis Module: If the delay has a serious impact, analyze the storage compression ratio information of each frame of the image during the image compression process, compare the differences between the compressed image and the original image, and evaluate the compression quality of the image;

[0029] Image Quality Classification Module: Evaluate the degree of image quality degradation according to the impact of the delay on image real-time performance and the image compression quality. According to the evaluation results, divide the degree of image quality degradation into serious degradation and slight degradation;

[0030] Image Enhancement and Recognition Adjustment Module: For slight degradation, enhance the image locally to improve the image quality, making the features of surface mines and unexploded ordnance clearer; for serious degradation, predict and warn the accuracy of the image recognition algorithm within a fixed time period, and automatically adjust the image recognition algorithm according to the prediction results to improve the accuracy of detecting surface mines and unexploded ordnance.

[0031] In the above technical solutions, the technical effects and advantages provided by the present invention are:

[0032] 1. By measuring the signal strength between the drone and the ground receiver in real time, combining the analysis of the image transmission time delay, the present invention accurately evaluates the impact on image quality, and quantifies the degree of image quality degradation through the detail loss anomaly index and the image compression ratio fluctuation index. Combining the prediction of the machine learning model, it can dynamically judge the slight or serious degradation of the image quality, and take local enhancement or adjust the image recognition algorithm according to different situations, thereby improving the accuracy and stability of detecting surface mines and unexploded ordnance.

[0033] 2. By precisely controlling the image quality and algorithm adjustment, the present invention can effectively cope with the uncertainty of the wireless communication environment, and ensure that high-quality image information can still be provided under complex terrain and weather conditions. By comprehensively evaluating the impact of image compression quality and transmission delay on image real-time performance, and making multi-level adjustments through the warning mechanism, not only the reliability of the surface mine and unexploded ordnance detection system is improved, but also the image recognition error is effectively reduced, ensuring the efficient completion of the detection task. This innovative method significantly improves the adaptability and accuracy of the drone-borne image recognition system in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0035] Figure 1 It is a flowchart of the method of the present invention.

[0036] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0037] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0038] Embodiment 1. Please refer to Figure 1 and Figure 2 As shown, the method for detecting surface mines and unexploded ordnance on an airborne platform based on image recognition in this embodiment includes the following steps:

[0039] S1: The unmanned aerial vehicle (UAV) flies over the target area through a remote control command system, and uses the high-resolution camera carried thereon to collect image information of surface mines, unexploded ordnance, and their surrounding environments in real time, and measures and records the strength of the wireless communication signal between the UAV and the ground receiver in real time;

[0040] S2: Compare and analyze the signal strength fluctuations and the time delay of image transmission to judge the degree of interference of the delay on the image resolution and detail retention, and evaluate the impact of the delay on the real-time performance of the image;

[0041] S3: If the delay has a serious impact, analyze the storage compression ratio information of each frame of the image during the image compression process, compare the differences between the compressed image and the original image, and evaluate the compression quality of the image;

[0042] S4: Evaluate the degree of image quality degradation based on the impact of the delay on the real-time performance of the image and the image compression quality. According to the evaluation results, divide the degree of image quality degradation into serious degradation and slight degradation;

[0043] S5: For slight degradation, enhance the image locally to improve the image quality, making the features of surface mines and unexploded ordnance clearer; for severe degradation, predict and give early warning of the accuracy of the image recognition algorithm within a fixed time period, and automatically adjust the image recognition algorithm according to the prediction results to improve the accuracy of detecting surface mines and unexploded ordnance.

[0044] In S1, the drone flies over the target area through the remote control command system, and uses the equipped high-resolution camera to collect real-time image information of surface mines, unexploded ordnance and their surrounding environment, and measures and records the wireless communication signal strength between the drone and the ground receiver in real time. Specifically:

[0045] The drone is controlled to fly through the remote control command system, which usually includes automatic navigation and manual control functions to achieve precise positioning and flight route planning. Over the target area, the drone will fly according to a predetermined trajectory. The drone ensures precise positioning during flight through technologies such as GPS, inertial navigation system (INS), and visual inertial odometry (VIO). During flight, it can provide real-time feedback on the position, speed, and altitude of the drone to ensure the accuracy of flight and the coverage of the image acquisition area. Through the high-resolution camera, images of the target area are collected in real time to identify potential areas of surface mines and unexploded ordnance. The positioning information and image data of these areas will be recorded for subsequent processing.

[0046] The equipped high-resolution camera can provide high-quality images, ensuring that details of surface mines and unexploded ordnance can still be clearly captured at a long distance. The camera automatically adjusts exposure, focus, and other image parameters according to the flight altitude and speed to obtain the optimal image quality. The collected images are usually stored in lossless formats such as RAW or TIFF to ensure the integrity of the image data and provide high-quality raw data for subsequent image processing and analysis. The camera collects images in real time and transmits the images to the ground receiving end through the wireless communication system. This process is highly dependent on transmission bandwidth, latency, and image compression algorithms.

[0047] The wireless communication device (such as Wi-Fi, LTE, 5G, etc.) carried by the drone continuously monitors the signal strength between the drone and the ground receiver. The built-in signal strength detection module of the drone collects and records the signal strength data in real time. As the drone flies, the signal strength may fluctuate due to terrain, weather, and obstacles (such as buildings, mountains, etc.). By measuring the signal strength, the system can identify areas with weakening signals in real time and provide a basis for subsequent image transmission optimization. Correlate the signal strength data with the collected image information and record the signal strength change at each moment. This provides data support for subsequent analysis of image transmission quality and its impact on signal strength fluctuations.

[0048] To accurately analyze the relationship between signal strength and image quality, the system needs to add timestamps to the image data and signal strength data. This can ensure that in subsequent steps, the corresponding relationship between signal strength fluctuations and image quality degradation can be accurately identified. The signal strength data and image acquisition data need to be synchronized in real time to facilitate the monitoring of any interference or delay during image transmission. Through real-time feedback, the operator can adjust the drone flight route or communication strategy according to the actual situation.

[0049] S2: Conduct a comparative analysis of the signal strength fluctuations and the time delay of image transmission, judge the degree of interference of the delay on image resolution and detail retention, and evaluate the impact of the delay on image real-time performance.

[0050] The drone continuously monitors the signal strength during flight through the onboard wireless communication devices (such as Wi-Fi, 4G, 5G, or a dedicated drone communication protocol). These signal strength data are recorded in real time and stored synchronously with the image data. By analyzing the changes in the strength of the wireless signal, it is possible to understand whether the signal strength fluctuates violently or is within a relatively stable range. If the fluctuations are too large, it may lead to a decline in the quality of image transmission and even packet loss.

[0051] Image transmission time delay refers to the time delay from when the drone's camera captures an image to when the image reaches the ground receiving end. This delay is related to various factors, such as signal strength, flight distance, wireless bandwidth, image compression method, etc. The system can record the transmission time of each frame of the image and calculate the average time delay and the maximum time delay. High-resolution images require a larger bandwidth. Especially in scenarios of long-distance flight and high data volume transmission, insufficient bandwidth may lead to an increase in the time delay of image transmission. By comparing the relationship between bandwidth and time delay, the key factors affecting time delay can be identified.

[0052] When there is a time delay during image transmission, especially in the case of unstable signals or insufficient bandwidth, the image may need to be compressed or the resolution reduced to meet the bandwidth requirements of transmission. The decrease in resolution directly affects the detail performance of the image and may result in the loss of details of surface mines and unexploded ordnance, thus affecting subsequent identification and classification. Analysis method: The image compression ratio and resolution change under different time delay conditions can be calculated, and the detail differences between the original image and the compressed image can be compared. For example, techniques such as edge detection and contrast analysis are used to quantify the degree of resolution loss. If the delay is too long, the image may be distorted or blurred due to network congestion or signal attenuation, especially the contours, details, and color changes of surface mines and unexploded ordnance, which may lead to misjudgment or ineffective identification.

[0053] Image transmission delay not only affects the resolution, but may also cause the loss of image details during the image compression process. Especially for targets such as surface mines and unexploded ordnance that require fine recognition, the loss of any details will affect the recognition accuracy of the algorithm. By comparing the degree of detail loss between the compressed image and the original image due to image transmission delay, for example, using an image similarity metric (such as the structural similarity index SSIM) to measure the degree of detail retention, especially for details such as the contour, color, and texture of the target object. If the delay causes excessive detail loss, it may affect the correct recognition of surface mines and unexploded ordnance, increasing the risk of misjudgment and missed judgment.

[0054] During the mission execution of the UAV, the real-time transmission of images is crucial. Especially in a dynamically changing environment, the timeliness of the images is directly related to subsequent decision-making and disposal. For example, real-time image recognition can quickly determine the location, type, and danger level of surface mines and unexploded ordnance, helping to judge whether further processing is required. Analysis of real-time impact: By comparing the update frequency and timeliness of images under different delay conditions, it is possible to evaluate whether the delay exceeds the tolerance range of the system. If the image update lags too much, it may lead to outdated information and fail to reflect the changes in the current environment, thus affecting the mission execution.

[0055] After comparing the degree of detail loss between the compressed image and the original image due to image transmission delay, a detail loss anomaly index is generated to evaluate the impact of the delay on image real-time. The method for obtaining the detail loss anomaly index is as follows:

[0056] Obtain the original image and the compressed image: Original image: The image collected in real-time from the UAV, an image that has not undergone transmission or compression processing. Compressed image: The image that has undergone transmission or compression (possibly due to network delay and compression algorithm) processing. Match the sizes of the two images to ensure they have the same resolution (if the image sizes are different, interpolation adjustment can be performed). Perform grayscale processing on the two images (if it is a color image, it can be converted to a grayscale image for calculation).

[0057] The structural similarity index SSIM is used to measure the similarity between two images, considering three aspects: brightness, contrast, and structure. The formula is: In the formula, x and y are the corresponding pixel values of the original image and the compressed image, μ x and μ y are the local means (brightness information) of the original image and the compressed image, that is, the average value of the image block, and are the local variances (contrast information) of the original image and the compressed image, representing the degree of brightness change within the local area. σ xyis the local covariance (structural information) of the original image and the compressed image, representing the degree of similarity between the two images. C1 and C2 are constants used to avoid a zero denominator. Usually, C1 = K1L) 2 and C2 = (K2L) 2 , where L is the dynamic range of the image (e.g., 255 for an 8-bit image), and K1 and K2 are small constants. Usually, K1 = 0.01 and K2 = 0.03 are chosen.

[0058] According to the SSIM value, the Detail Loss Anomaly Index (DLI) is calculated to measure the degree of detail loss caused by image transmission delay and compression. The expression is: SSIM threshold is a preset SSIM threshold for judging image quality. Usually, it can be set to 0.7. If SSIM is lower than this value, it indicates that the detail loss of the image is relatively serious.

[0059] The larger the Detail Loss Anomaly Index (DLI), the more serious the impact of the delay on the real-time performance of the image. This is because in the calculation method of DLI, when significant detail loss occurs during image compression or transmission, the SSIM value will be low. The larger the DLI value, the more obvious the information loss during the compression process, the poorer the degree of detail retention, and the significant decline in image quality. This is usually related to factors such as long transmission delay, signal packet loss, or excessive compression, resulting in unclear images and blurred details, thus affecting the accuracy of image recognition algorithms. Therefore, when the DLI value is large, it indicates that the image quality has been significantly affected, and both the real-time performance and accuracy of the system may be severely interfered with.

[0060] The smaller the Detail Loss Anomaly Index (DLI), the less serious the impact of the delay on the real-time performance of the image. When the DLI value is small, it indicates that more details are retained during image transmission, and the quality loss caused by compression or delay is less. Usually, when the SSIM value is close to 1, it means that there is almost no loss of image details and structural information. At this time, even if there is a certain signal delay or compression, the image still maintains a high quality and can perform target detection and recognition more accurately. Therefore, a smaller DLI value usually means that the real-time performance and accuracy of the image are less affected, and the system can perform image recognition tasks more reliably.

[0061] S3: If the delay impact is severe, then analyze the storage compression ratio information of each frame of the image during the image compression process, compare the differences between the compressed image and the original image, and evaluate the compression quality of the image.

[0062] The compression ratio (Compression Ratio, CR) is the ratio of the size of the image before compression to the size of the image after compression, reflecting the degree of data reduction during the compression process. Its formula is: The Original Image Size is the size of the original image, usually measured in bytes. The Compressed Image Size is the size of the compressed image, also measured in bytes. The higher the compression ratio, the smaller the compressed image file, indicating better compression efficiency, but it may be accompanied by more detail loss. Conversely, the lower the compression ratio, the less the image is compressed, better retaining details, but the image file size remains larger.

[0063] For each frame of images collected by an airborne drone, first, it is necessary to measure the image sizes before and after compression and calculate the compression ratio. The specific steps are as follows:

[0064] Obtain the original size of each frame of image: Measure the file size of each frame of the original image and record the number of bytes of the original image.

[0065] Compress the image: Compress each frame of the image, usually using a certain image compression algorithm (such as JPEG, JPEG2000, HEVC, etc.).

[0066] Obtain the size of the compressed image: After compression, record the size of the image file (in bytes).

[0067] Calculate the compression ratio: Use the above formula to calculate the compression ratio of each frame of the image.

[0068] The level of the image compression ratio directly affects the quality of the compressed image. The higher the compression ratio, the more the image is compressed, and the more details of the image may be lost. Usually:

[0069] A high compression ratio (for example, the compression ratio is greater than 10:1) usually leads to significant quality loss, especially in details and textures, and the image may appear blurred or distorted.

[0070] A low compression ratio (for example, the compression ratio is close to 1:1) retains more details, but has poor compression efficiency and a larger image file size.

[0071] By analyzing the compression ratio of each frame of the image, the quality of the image can be inferred. For example, when the compression ratio is too high, it may lead to problems such as image detail loss, texture blurring, and color distortion, affecting the subsequent image recognition process.

[0072] After analyzing the fluctuation of the compression ratio of each frame of the image, an image compression ratio fluctuation index is generated to evaluate the compression quality of the image. The method for obtaining the image compression ratio fluctuation index is as follows:

[0073] For each frame of the image, calculate its compression ratio CR i, and record it. Suppose the image sequence has a total of N frames, then the compression ratio data is {CR1, CR2, …, CR N}; Calculate the frequency distribution of all compression ratio values, and obtain the occurrence probability p(CR i ) of each compression ratio CR in the entire dataset; The calculation formula for the frequency distribution is: i where the frequency (CR i ) is the number of times the compression ratio CR i appears, N is the total number of frames of the compression ratio data, and the entropy is calculated through the probability distribution of the compression ratio. The expression is: where p(CR i ) is the probability of the compression ratio CR i , n is the number of different compression ratio values (i.e., the number of discrete values of the compression ratio), and H(CR) is the entropy of the compression ratio, indicating the uncertainty or volatility of the compression ratio values. Calculate the image compression ratio fluctuation index, and the expression is: CRFI = H(CR); In the formula, CRFI is the image compression ratio fluctuation index.

[0074] The larger the image compression ratio fluctuation index, the more unstable the compression quality of the image, and there may be significant differences in image quality. When the compression ratio fluctuation index is large, it means that the change range of the compression ratio between image frames is large, which is usually caused by poor adaptability to image content or unbalanced dynamic adjustment strategies during the compression process. A high fluctuation index may lead to too low compression quality (severe loss of details) for some frames, while some other frames may be over-retained (occupying more storage space). This instability will interfere with subsequent image recognition or analysis tasks, especially in scenarios where consistent image quality is required.

[0075] The smaller the image compression ratio fluctuation index, the more stable the compression quality of the image, and the quality between frames is more consistent. When the fluctuation index is small, the change range of the compression ratio is small, which means that the compression algorithm has a high adaptability when processing different frames and can balance the quality and storage requirements of the image. This stability helps to maintain consistency in the image sequence, reduce recognition errors or visual discomfort caused by differences in quality between frames, and at the same time improve the compression efficiency and reliability of the system.

[0076] S4: Evaluate the degree of image quality degradation based on the impact of latency on image real-time performance and the image compression quality. According to the evaluation results, divide the degree of image quality degradation into severe degradation and slight degradation.

[0077] ​Convert the detail loss anomaly index and the image compression ratio fluctuation index into a comprehensive feature vector, and use the comprehensive feature vector as the input of a machine learning model. The machine learning model takes predicting the label of the image quality degradation degree value for each set of comprehensive feature vectors as the prediction target, and takes minimizing the sum of the prediction errors of the image quality degradation degree value labels for all images as the training target. Train the machine learning model until the sum of the prediction errors converges, and then stop the model training. Determine the image quality degradation degree value according to the model output result, where the machine learning model is a polynomial regression model.

[0078] The method for obtaining the image quality degradation degree value is as follows: Obtain the corresponding function expression from the comprehensive feature vector training data of the trained machine learning model: R = F(DLI, CRFI); where F is the output function of the model, DLI is the detail loss anomaly index, CRFI is the image compression ratio fluctuation index, and R is the image quality degradation degree value.

[0079] Compare the obtained image quality degradation degree value with the reference threshold of the image quality degradation degree value set according to historical data. If the image quality degradation degree value is greater than or equal to the set reference threshold of the image quality degradation degree value, it indicates that the image quality degradation is serious. At this time, generate a serious image quality degradation signal and classify the image quality degradation degree as serious degradation; if the image quality degradation degree value is less than the set reference threshold of the image quality degradation degree value, it indicates that the image quality degradation is slight. At this time, generate a slight image quality degradation signal and classify the image quality degradation degree as slight degradation.

[0080] S5: For slight degradation, enhance the image locally to improve the image quality, making the features of surface mines and unexploded ordnance clearer; for serious degradation, predict and warn the accuracy of the image recognition algorithm within a fixed time period, and automatically adjust the image recognition algorithm according to the prediction result to improve the detection accuracy of surface mines and unexploded ordnance.

[0081] The goal of local enhancement is to optimize the key areas in the image, making the details of surface mines and unexploded ordnance clearer and improving their detectability and accuracy in subsequent image recognition. For images with slight degradation, details can be restored by enhancing contrast, brightness, sharpening, etc., especially in the detail areas lost during compression or transmission.

[0082] The local enhancement method first needs to determine the important areas in the image, that is, the areas where surface mines and unexploded ordnance may exist. This can be achieved through the following methods:

[0083] Selection based on Region of Interest (ROI): Generally, regions that may contain surface mines and unexploded ordnance are roughly located first through object detection or image segmentation techniques (such as circular or rectangular candidate boxes).

[0084] Edge detection: Edge detection algorithms (such as Canny edge detection, Sobel operator) can be used to determine the edges of objects in the image, thereby selecting regions that may contain surface mines and unexploded ordnance.

[0085] Deep learning prediction: Convolutional neural networks (CNNs) or other object detection algorithms (such as YOLO or Faster R-CNN) are used to automatically identify target objects in the image and determine key regions.

[0086] According to the type of image quality degradation, select an appropriate local enhancement algorithm to improve image clarity. Common local enhancement algorithms include:

[0087] Histogram equalization can enhance the contrast of the image, especially in low-contrast image regions. By enhancing the contrast of local regions, the details of surface mines and unexploded ordnance can be made more prominent.

[0088] Adaptive Histogram Equalization, this method can perform histogram equalization according to the image features of local regions, avoiding excessive enhancement of the overall image contrast, and is suitable for local enhancement.

[0089] Local sharpening mainly helps to restore the detail loss caused by compression and transmission by enhancing the edge details of the image.

[0090] Use edge detection algorithms such as the Laplacian or Sobel operator, combined with weighted enhancement of local regions, to enhance the contours and details of surface mines and unexploded ordnance.

[0091] CLAHE (Contrast Limited Adaptive Histogram Equalization) is a local enhancement algorithm that can effectively enhance the contrast of local regions while avoiding the phenomenon of over-enhancement, and is suitable for restoring image features in the case of detail loss. The CLAHE algorithm equalizes the histogram by dividing the image into blocks and performing histogram equalization on each block, while limiting the contrast enhancement of each block to avoid the enhancement of noise.

[0092] Local brightness enhancement can perform brightness compensation for darker areas of illumination, making the details in the image more visible. Through adaptive brightness adjustment, the visible features of surface mines and unexploded ordnance can be better restored. A local filter is used to adjust the brightness, making the areas of surface mines and unexploded ordnance more distinguishable. After completing the local enhancement, it is necessary to evaluate the enhancement effect. The image enhancement effect can be verified in the following ways:

[0093] By comparing the contrast and clarity of the image before and after enhancement, verify whether the details of the image have been enhanced. Use an image recognition model to detect the enhanced image and evaluate the recognition rate of surface mines and unexploded ordnance.

[0094] For a significant decline, predict and give an early warning of the accuracy of the image recognition algorithm within a fixed time period, and automatically adjust the image recognition algorithm according to the prediction results to improve the detection accuracy of surface mines and unexploded ordnance. Specifically:

[0095] For a significant decline, that is, the degree of image quality decline generated within a fixed time period is greater than or equal to the reference threshold of the set degree of image quality decline, collect the degree of quality decline values greater than or equal to the reference threshold of the degree of image quality decline generated in the subsequent fixed time period, and establish a corresponding data set, calculate the mean and standard deviation of the data set, and predict and give an early warning of the accuracy of the image recognition algorithm within a fixed time period.

[0096] If the mean value of the degree of quality decline in the data set is greater than or equal to the reference threshold of the mean value of the degree of quality decline, and the standard deviation of the degree of quality decline is less than the reference threshold of the standard deviation of the degree of quality decline, a first-level warning signal is generated at this time. The first-level warning signal is the most serious, indicating that the change in the degree of quality decline is small but the persistence is strong. It is necessary to immediately adjust the image recognition algorithm, and it may be necessary to enable redundant detection or increase the fault tolerance ability of image processing.

[0097] If the mean value of the degree of quality decline is greater than or equal to the reference threshold of the mean value of the degree of quality decline, and the standard deviation of the degree of quality decline is greater than or equal to the reference threshold of the standard deviation of the degree of quality decline, a second-level warning signal is generated at this time. The second-level warning signal indicates that the quality decline fluctuates within a certain range, and partial algorithm optimization is required, such as adjusting parameters or enhancing the diversity of the data set.

[0098] If the mean value of the degree of quality decline is less than the reference threshold of the mean value of the degree of quality decline, and the standard deviation of the degree of quality decline is greater than or equal to the reference threshold of the standard deviation of the degree of quality decline, a third-level warning signal is generated at this time. The third-level warning signal indicates that the fluctuation of the image quality decline is large, but the overall decline is not serious. It may indicate network interference or short-term performance fluctuations, and it is necessary to optimize the image transmission path or perform image enhancement.

[0099] If the average value of the quality degradation degree is less than the reference threshold of the average value of the quality degradation degree, and the standard deviation of the quality degradation degree is less than the reference threshold of the standard deviation of the quality degradation degree, no warning signal is generated at this time. The absence of a warning signal indicates that the image quality is good, and there is no need to adjust the image recognition algorithm, and the current operation can be continued.

[0100] Here, it should be noted that the importance of the first-level warning signal is greater than that of the second-level warning signal, and the importance of the second-level warning signal is greater than that of the third-level warning signal. Relevant personnel can take corresponding treatment measures according to different warning signal levels.

[0101] According to different warning signal levels, different measures are taken to adjust the image recognition algorithm to improve the detection accuracy of surface mines and unexploded ordnance:

[0102] First-level warning: Conduct global adjustment of the image recognition algorithm, such as adding redundant detection, adding image preprocessing steps, and improving the resolution of the image. More complex deep learning models can be enabled to deal with poor image quality.

[0103] Second-level warning: Conduct local optimization and adjustment, such as adjusting model parameters or enhancing the diversity of the dataset, to adapt to fluctuations in different image qualities.

[0104] Third-level warning: Improve the recognition accuracy by improving the transmission path, enhancing the signal, or performing local image enhancement, while keeping the image quality within an acceptable range.

[0105] In this embodiment, a high-resolution camera is carried by a drone to collect images of surface mines, unexploded ordnance, and the surrounding environment in real time, and combined with the real-time measurement of the wireless communication signal strength, the delay and compression quality of image transmission are analyzed. When the signal delay has a serious impact on the real-time performance and detail preservation of the image, the system evaluates the image compression ratio and compression quality, determines the degree of image quality degradation, and classifies it into two categories: serious degradation and slight degradation. In the case of slight degradation, the image quality is improved by local enhancement to make the features of surface mines and unexploded ordnance clearer; while in the case of serious degradation, the system predicts the accuracy of the image recognition algorithm, gives a timely warning, and automatically adjusts the recognition algorithm to improve the detection accuracy of surface mines and unexploded ordnance.

[0106] Embodiment 2. The airborne surface mine and unexploded ordnance detection system based on image recognition described in this embodiment includes an image acquisition and signal monitoring module, a signal delay evaluation module, an image compression quality analysis module, an image quality classification module, and an image enhancement and recognition adjustment module:

[0107] Image Acquisition and Signal Monitoring Module: The drone flies over the target area through the remote control command system, and uses the equipped high-resolution camera to collect real-time image information of surface mines, unexploded ordnance and their surrounding environment, and measures and records the real-time wireless communication signal strength between the drone and the ground receiver;

[0108] Signal Delay Evaluation Module: Compare and analyze the signal strength fluctuation and the time delay of image transmission, judge the interference degree of the delay on the image resolution and detail preservation, and evaluate the impact of the delay on the real-time performance of the image;

[0109] Image Compression Quality Analysis Module: If the delay has a serious impact, analyze the storage compression ratio information of each frame of the image during the image compression process, compare the differences between the compressed image and the original image, and evaluate the compression quality of the image;

[0110] Image Quality Classification Module: Evaluate the degree of image quality degradation according to the impact of the delay on the real-time performance of the image and the image compression quality. According to the evaluation results, divide the degree of image quality degradation into severe degradation and slight degradation;

[0111] Image Enhancement and Recognition Adjustment Module: For slight degradation, enhance the image locally to improve the image quality, making the features of surface mines and unexploded ordnance clearer; for severe degradation, predict and warn the accuracy of the image recognition algorithm within a fixed time period, and automatically adjust the image recognition algorithm according to the prediction results to improve the detection accuracy of surface mines and unexploded ordnance.

[0112] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data and performing software simulation to get a formula closest to the real situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0113] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Among them, A and B can be singular or plural. In addition, the character " / " in this article generally represents an "or" relationship between the front and rear associated objects, but it may also represent an "and / or" relationship, which can be specifically understood by referring to the context.

[0114] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or by a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professionals can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of this application.

[0115] As described above, it is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application.

Claims

1. An airborne surface mine and unexploded ordnance detection method based on image recognition, characterized in that: It includes the following steps: S1: The drone flies over the target area through the remote control command system, and uses the onboard high-resolution camera to collect real-time image information of surface mines, unexploded ordnance and their surrounding environments, and measures and records the real-time wireless communication signal strength between the drone and the ground receiver; S2: Compare and analyze the signal strength fluctuation and the time delay of image transmission. By comparing the degree of detail loss of the compressed image and the original image due to the image transmission time delay, generate a detail loss anomaly index, judge the interference degree of the delay on the image resolution and detail preservation, and evaluate the impact of the delay on the real-time performance of the image; S3: If the delay has a serious impact, analyze the storage compression ratio information of each frame of the image during the image compression process, generate an image compression ratio fluctuation index after analyzing the fluctuation of the compression ratio of each frame of the image, compare the difference between the compressed image and the original image, and evaluate the compression quality of the image; S4: Evaluate the degree of image quality degradation based on the impact of the delay on the real-time performance of the image and the image compression quality. According to the evaluation results, divide the degree of image quality degradation into serious degradation and slight degradation. Specifically: convert the detail loss anomaly index and the image compression ratio fluctuation index into a comprehensive feature vector, use the comprehensive feature vector as the input of the machine learning model, and use the machine learning model to predict the image quality degradation degree value label for each group of comprehensive feature vectors as the prediction target, and use minimizing the sum of the prediction errors of the image quality degradation degree value labels for all images as the training target to train the machine learning model until the sum of the prediction errors reaches convergence and stop the model training. Determine the image quality degradation degree value according to the model output result. Among them, the machine learning model is a polynomial regression model; S5: For slight degradation, enhance the image locally to improve the image quality and make the features of surface mines and unexploded ordnance clearer; for serious degradation, predict and warn the accuracy of the image recognition algorithm within a fixed time period, and automatically adjust the image recognition algorithm according to the prediction result to improve the accuracy of detecting surface mines and unexploded ordnance.

2. The method for detecting surface mines and unexploded ordnance on board based on image recognition according to claim 1, characterized in that: In S2, after comparing the degree of detail loss of the compressed image and the original image due to the image transmission time delay, generate a detail loss anomaly index. The method for obtaining the detail loss anomaly index is: Obtain the original image and the compressed image, match the sizes of the two images, perform grayscale processing on the two images. The structural similarity index SSIM is used to measure the similarity between the two images, considering three aspects: brightness, contrast, and structure. The formula is as follows: ; where x and y are the corresponding pixel values of the original image and the compressed image, and are the local means of the original image and the compressed image, that is, the average value of the image block, and are the local variances of the original image and the compressed image, representing the degree of brightness change within the local area, is the local covariance of the original image and the compressed image, representing the degree of similarity between the two images; and are constants used to avoid the denominator being zero, where L is the dynamic range of the image, and are small constants; according to the SSIM value, calculate the detail loss anomaly index DLI, which is used to measure the degree of detail loss caused by image transmission delay and compression. The expression is: ; is the preset SSIM threshold.

3. The method for detecting surface mines and unexploded ordnance on board based on image recognition according to claim 2, wherein: In S3, after analyzing the fluctuation of the compression ratio of each frame of the image, generate an image compression ratio fluctuation index. The method for obtaining the image compression ratio fluctuation index is: For each frame of the image, calculate its compression ratio , and record it. Assuming that there are N frames in the image sequence, the compression ratio data is ; Calculate the frequency distribution of all compression ratio values to obtain the occurrence probability of each compression ratio value in the entire dataset ; The calculation formula for the frequency distribution is: ; Among them, is the number of times the compression ratio appears, N is the total number of frames of the compression ratio data, and the entropy is calculated through the probability distribution of the compression ratio. The expression is: ; Among them, is the probability of the compression ratio , n is the number of different compression ratio values, H(CR) is the entropy of the compression ratio. Calculate the image compression ratio fluctuation index, and the expression is: ; In the formula, CRFI is the image compression ratio fluctuation index.

4. The airborne surface mine and unexploded ordnance detection method based on image recognition according to claim 1, characterized in that: Compare the obtained image quality degradation degree value with the reference threshold of the image quality degradation degree value set according to historical data. If the image quality degradation degree value is greater than or equal to the set reference threshold of the image quality degradation degree value, it indicates that the image quality degradation is serious. At this time, generate a serious image quality degradation signal and divide the image quality degradation degree into serious degradation; if the image quality degradation degree value is less than the set reference threshold of the image quality degradation degree value, it indicates that the image quality degradation is slight. At this time, generate a slight image quality degradation signal and divide the image quality degradation degree into slight degradation.

5. The method for detecting surface mines and unexploded ordnance on board based on image recognition according to claim 1, characterized in that: In S5, for a severe decline, that is, the degree of decline in the quality of the images generated within a fixed time period is greater than or equal to the reference threshold of the set degree of decline in image quality, the degree of decline in quality greater than or equal to the reference threshold of the degree of decline in image quality generated in the subsequent fixed time period is collected, and a corresponding data set is established. The mean and standard deviation of the data set are calculated to predict and give an early warning of the accuracy of the image recognition algorithm within the fixed time period.

6. The airborne surface mine and unexploded ordnance detection method based on image recognition according to claim 5, wherein: If the mean value of the degree of decline in quality within the data set is greater than or equal to the reference threshold of the mean value of the degree of decline in quality, and the standard deviation of the degree of decline in quality is less than the reference threshold of the standard deviation of the degree of decline in quality, a first-level warning signal is generated at this time, and the image recognition algorithm needs to be adjusted immediately to increase the fault tolerance of image processing. If the mean value of the degree of decline in quality is greater than or equal to the reference threshold of the mean value of the degree of decline in quality, and the standard deviation of the degree of decline in quality is greater than or equal to the reference threshold of the standard deviation of the degree of decline in quality, a second-level warning signal is generated at this time, and partial algorithm optimization and parameter adjustment are required. If the mean value of the degree of decline in quality is less than the reference threshold of the mean value of the degree of decline in quality, and the standard deviation of the degree of decline in quality is greater than or equal to the reference threshold of the standard deviation of the degree of decline in quality, a third-level warning signal is generated at this time, and the image transmission path needs to be optimized. If the mean value of the degree of decline in quality is less than the reference threshold of the mean value of the degree of decline in quality, and the standard deviation of the degree of decline in quality is less than the reference threshold of the standard deviation of the degree of decline in quality, no warning signal is generated at this time, there is no need to adjust the image recognition algorithm, and the current operation continues.

7. An airborne surface mine and unexploded ordnance detection system based on image recognition, which is used to implement the airborne surface mine and unexploded ordnance detection method based on image recognition according to any one of claims 1-6, and is characterized in that: It includes an image acquisition and signal monitoring module, a signal delay evaluation module, an image compression quality analysis module, an image quality classification module, and an image enhancement and recognition adjustment module: Image acquisition and signal monitoring module: The unmanned aerial vehicle (UAV) flies over the target area through the remote control command system, and uses the equipped high-resolution camera to collect the image information of surface mines, unexploded ordnance and their surrounding environment in real time, and measures and records the strength of the wireless communication signal between the UAV and the ground receiver in real time. Signal delay evaluation module: Compare and analyze the signal strength fluctuation and the time delay of image transmission, judge the interference degree of the delay on the image resolution and detail preservation, and evaluate the impact of the delay on the real-time performance of the image. Image compression quality analysis module: If the delay has a serious impact, analyze the storage compression ratio information of each frame of the image during the image compression process, compare the difference between the compressed image and the original image, and evaluate the compression quality of the image. Image quality classification module: Evaluate the degree of decline in the quality of the image according to the impact of the delay on the real-time performance of the image and the image compression quality. According to the evaluation results, divide the degree of decline in the quality of the image into severe decline and slight decline. Image enhancement and recognition adjustment module: For slight decline, enhance the image locally to improve the image quality, making the features of surface mines and unexploded ordnance clearer. For severe decline, predict and give an early warning of the accuracy of the image recognition algorithm within a fixed time period, and automatically adjust the image recognition algorithm according to the prediction results to improve the accuracy of detecting surface mines and unexploded ordnance.

Images