A method for detecting insufficient distance for personnel to ride on a detachable monkey car

Abstract

The present application belongs to the technical field of coal mine safety engineering and artificial intelligence, and discloses a detection method for insufficient distance of personnel riding on a detachable monkey car, and the specific technical scheme is as follows: different regions and conditions in the underground transportation system are selected for data collection; noise removal processing and data enhancement are performed; image normalization and standardization processing are performed; image labeling is performed using a labeling tool labelimg; the labeled data set is split into a training set, a verification set and a test set according to a ratio of 7:2:1; a network structure is determined and training hyperparameters are selected, a YOLOv5 network structure is selected, and the optimal value is selected through a grid search method; detection result post-processing and alarm triggering are performed; the present application fully combines artificial intelligence, deep learning and advanced image processing technology, and constructs an efficient and intelligent monitoring system, which can realize accurate personnel monitoring in a complex underground environment.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of coal mine safety engineering and artificial intelligence, and specifically relates to a method for detecting insufficient distance of personnel riding on a detachable monkey car. BACKGROUND

[0002] The detachable monkey car is a tool widely used in underground coal mine transportation, mainly used for transporting personnel and materials. In the underground coal mine working environment, the shaft space is narrow, and it is crucial to maintain a safe distance between the personnel riding the monkey car. In recent years, with the expansion of coal mine production scale and the improvement of mechanization degree, the safety of underground transportation has attracted more and more attention.

[0003] According to the requirements of relevant policy documents, coal mine enterprises must take effective measures to ensure the safety of underground personnel transportation. Coal mine enterprises should establish and improve the safety production responsibility system, strengthen the monitoring and management of underground transportation system, and prevent safety accidents caused by insufficient distance of personnel riding.

[0004] In order to solve the problem of personnel distance monitoring in underground transportation, an automatic and accurate technology is urgently needed. In recent years, the existing personnel riding distance monitoring method based on artificial intelligence and deep learning technology mainly relies on manual observation and management, which not only is inefficient, but also is prone to omissions. Artificial monitoring not only requires high concentration of personnel, but also has the problem of inaccurate subjective judgment, which is difficult to cope with complex underground environment. In addition, with the development of automation and intelligentization of underground coal mine transportation system, traditional monitoring methods cannot meet the needs of modern coal mine safety production. SUMMARY

[0005] To solve the technical problems existing in the prior art, the present application provides an efficient and accurate riding personnel distance monitoring method based on the mainstream YOLOv5 detection model and combining image preprocessing, post-processing algorithm and improved target detection method. This method optimizes the model structure, enhances image processing, introduces joint training and real-time analysis technology, realizes accurate detection and monitoring of riding personnel in complex underground environment, and effectively guarantees the safety and reliability of underground transportation.

[0006] To achieve the above purpose, the technical solution adopted by the present application is as follows: a method for detecting insufficient distance of personnel riding on a detachable monkey car, the specific steps are as follows:

[0007] Step S1, Data Set Collection: In the data collection phase, first, high-quality and stable cameras need to be deployed at key locations of the underground transportation system, and it is necessary to ensure that the cameras can cover the areas that need to be monitored. The collection process should cover different time periods to ensure that effective static image data and dynamic video data can be collected under various lighting and environmental conditions. For the special environment of coal mine underground, data collection should cover different transportation system scenarios as much as possible, such as different running speeds, different tunnels, and different time periods of the monkey car. During data labeling, the passengers in each frame of image need to be accurately labeled, recording their position, size, and other information. The labeling tool can choose a rectangular box labeling form, and at the same time, the specific state of the labeling personnel (such as different monkey car passenger gestures) is labeled. In order to improve the quality of the data, it is recommended to check the images in real time during the collection process to ensure that each frame has sufficient clarity and usability, avoiding labeling errors caused by image blur or obstruction.

[0008] Step S2, Data Preprocessing: Data preprocessing includes two key steps of denoising and data augmentation to ensure the optimization of model training effect. First, denoising processing (such as Gaussian filtering, median filtering, and bilateral filtering, etc.) significantly improves the clarity and detail performance of the image by reducing the noise in the image, especially avoiding the strong light interference caused by the direct light of the headlight to the camera. This step helps to eliminate the visual interference caused by strong light, ensuring that the passengers and monkey car features are clearer. Second, data augmentation techniques increase the diversity of the data set by rotating, flipping, scaling, and color transforming the images. This technique not only simulates various actual operation scenarios, but also enhances the model's ability to adapt to different environmental changes. Through these preprocessing measures, the quality and diversity of the data are significantly improved, providing more accurate and rich input data for model training, thereby effectively improving the accuracy and robustness of target detection.

[0009] Step S3, Data Labeling and Splitting:

[0010] Data labeling and splitting is an important step to ensure the quality of model training and testing. The specific steps include:

[0011] Labeling tool selection: Use professional labeling tools such as labelimg for image labeling. These tools support various labeling methods such as rectangular boxes and polygons, and provide an intuitive user interface to facilitate efficient labeling of targets in images.

[0012] Labeling quality control: Conduct strict labeling quality checks to ensure accuracy and consistency. Check whether each labeled box accurately covers the target area and correct any labeling errors or omissions to ensure high-quality data sets.

[0013] Data splitting: The annotated dataset is split into training, validation, and test sets in the ratio of 7:2:1. The training set is used for model training, the validation set is used for model tuning and parameter selection, and the test set is used for final performance evaluation to ensure that each dataset can fully reflect various situations in actual applications.

[0014] Step S4, in improving the YOLOv5 network structure, by introducing Optimal Transport Assignment (OTA) algorithm and advanced loss function (EIOU, SIOU and AlphaIOU), the further improvement of target detection is realized. OTA algorithm converts the distance between target and detection frame into probability distribution, and minimizes the Wasserstein distance between two probability distributions, which significantly improves the detection speed and accuracy. At the same time, EIOU, SIOU and AlphaIOU loss function optimizes the matching between detection frame and real frame, considering the rotation relationship, shape difference and overlap degree, further enhancing the robustness and accuracy of the model. Through these improvements, YOLOv5 has significantly improved its detection performance in complex underground environments, providing reliable technical support for detecting insufficient distance between passengers.

[0015] Step S5, model training and tuning includes determining network structure and selecting appropriate training hyperparameters, such as learning rate, batch size, iteration number, optimizer and loss function, etc. In this process, first, select the appropriate YOLOv5 network structure, and adjust its layer number, convolution kernel size, etc. according to task requirements. Then, set the training hyperparameters, select the optimal value through grid search and other methods to ensure the efficiency and stability of the training process. During training, use standard loss functions such as confidence loss and boundary box regression loss, and can be customized and improved according to actual needs. During training, continuously monitor the performance of the model on the validation set, fine-tune the hyperparameters to avoid overfitting, and use early stopping mechanism to improve the generalization ability of the model. Finally, through careful training and tuning, the accuracy and stability of the model in the target detection task are improved, ensuring its application effect in the coal mine underground transportation environment.

[0016] Step S6, detection result post-processing, by recognizing, screening and analyzing each frame of image detected personnel, realizes the calculation of the interval time and actual distance of the passengers riding the monkey car. First, use the target detection algorithm to mark out the personnel meeting the fixed area and size range, and continuously calculate the time interval of adjacent passengers. Then, according to the speed of the monkey car, estimate the actual distance between the personnel, and compare it with the preset safety threshold, if the distance is less than the threshold, the system will trigger an alarm to ensure the accuracy and safety of the monitoring.

[0017] During the data collection process in step S1, to ensure the comprehensiveness and adaptability of the data, personnel image and video data under different lighting conditions, posture changes, distances, and monkey car driving speeds are collected. The collection scene covers various coal mine tunnel environments, and multi-angle, high-frame-rate cameras are used for synchronous collection to ensure the capture of dynamic changes of personnel under high-speed and low-speed driving conditions. In addition, special attention is paid to details such as personnel safety clothing and safety helmets in the data set, and images containing different personnel clothing and behaviors (such as standing, walking, sitting, etc.) are collected. To deal with complex mine environments, high dynamic range (HDR) cameras are also used to capture images under extreme lighting conditions such as strong light and shadows to ensure that the model can adapt to different lighting and environmental conditions. The diversity and complexity of these data provide rich support for subsequent model training, improving the robustness and accuracy of the model.

[0018] In step S2, bilateral filtering is used for denoising, and the specific denoising process is as follows:

[0019]

[0020] where I(x) is the output image pixel value, I(x i ) is the input image pixel value, W p is the normalized coefficient, σ d controls the spatial distance weight, and σ r controls the pixel value difference weight. The core of bilateral filtering is to consider both the spatial distance and the pixel value difference of the image, so that the edge information is preserved while effectively removing noise.

[0021] In the image enhancement process, histogram equalization is used to enhance the contrast of the image, making the gray scale distribution of the image more uniform, as follows:

[0022]

[0023] where S(i) is the equalized gray level, L is the number of gray levels, MN is the total number of pixels in the image, and h(j) is the number of pixels at gray level j. This formula redistributes the gray values of the image, improving the contrast of the image, thereby improving the visual effect of the image, especially for low-contrast images.

[0024] In step S4, for the matching of the target frame and the predicted frame, the OTA (Optimal Transport Algorithm) algorithm is used, which converts the distance between the target frame and the predicted frame into a probability distribution and uses the Wasserstein distance for matching. The specific process is described by the following formula:

[0025]

[0026] where C ij represents the distance between the ith target box and the jth predicted box, T ij is an element of the matching matrix, Ψ is the set of all legal matching matrices, the distance matrix C is transformed into probability distributions P and Q, where:

[0027]

[0028] Here τ is a temperature parameter that controls the smoothness of the probability distributions. The algorithm computes the minimum transportation cost between the target boxes and the predicted boxes to obtain the matching result, thereby improving the matching accuracy.

[0029] Optimizing the loss function: considering the angle difference between the target box and the predicted box, the detection accuracy is improved by optimizing the angle term, the formula is:

[0030]

[0031] where w t ,h t are the width and height of the target box, w p ,h p are the width and height of the predicted box. The angle difference term helps to optimize the alignment of the direction of the box, thereby improving the accuracy of detection.

[0032] Introducing a smoothing term to alleviate shape differences: by introducing a smoothing term to alleviate the shape difference between the target box and the real box, the formula is as follows:

[0033]

[0034] This smoothing term helps to reduce the shape difference between the target box and the predicted box, so that the model can better adapt to the small changes between the target and the predicted box.

[0035] Adjusting parameter α to balance the overlap: by adjusting the parameter α to balance the overlap between the predicted box and the real box, the formula is:

[0036]

[0037] where α is the adjustment parameter, ∈ is a very small constant to prevent division by zero error. This formula allows the model to adjust the importance of the overlapping part according to the characteristics of the actual task and the data set, thereby optimizing the detection result and further improving the detection accuracy.

[0038] This allows the model to adjust the importance of the overlapping part according to the characteristics of the actual task and the data set, thereby optimizing the detection result.

[0039] In step S6, the following steps are included:

[0040] Confidence screening: During the detection process, a confidence threshold is first applied to all detection boxes to ensure that only high-confidence detection boxes are retained. This step uses a set threshold (e.g., 0.5 or 0.7) to remove boxes below the threshold, reducing false positives and redundant information interference. Through this screening process, the system can focus more on high-confidence targets, improving detection accuracy.

[0041] Up-down classification: To achieve accurate personnel classification, the position, size, and preset area boundaries of the detection box are further combined with the actual up-down monitoring area to classify personnel as up or down. This process is achieved by dividing the monitoring area into multiple sub-areas and setting size standards for each sub-area. For example, for a person detection box in a specific area, if it is located in the designated up area and the detection box size meets the specified up personnel size standard, it is determined to be an up person, otherwise it is a down person. The innovation of this classification step is to adjust the area and size standards in real time to adapt to the monitoring needs in different environments, ensuring the accuracy of personnel classification.

[0042] Interval time calculation: When calculating the time interval between adjacent frames, if the video stream can directly record time information, the timestamp is directly read and used; otherwise, the time interval is calculated based on the interval frame number and frame rate. This method uses frame rate (e.g., 30fps) and frame spacing to accurately estimate the time interval of each frame, ensuring that the system can accurately calculate the personnel interval in real time, providing reliable data input for subsequent distance estimation.

[0043] Distance estimation: Based on the calculated interval time and the travel speed of the monkey car, the actual distance between the personnel riding the monkey car is estimated using a physical formula;

[0044] D = V * T

[0045] where D is the actual distance between personnel, V is the travel speed of the monkey car, and T is the time interval between two personnel detection boxes; through this formula, the relative distance between personnel can be calculated, providing a basis for safety judgment.

[0046] Safety threshold judgment: After calculating the actual distance between the personnel riding the monkey car, the system compares this distance with the preset safety threshold. If the distance is less than the safety threshold, the alarm system is triggered. This threshold is set not only according to process requirements and safety standards, but also can be dynamically adjusted by the model according to actual conditions to adapt to the safety distance requirements between personnel in different operating environments. When the distance is insufficient, the system will issue a real-time alarm to remind the operator to take timely measures to avoid potential safety risks.

[0047] The present application fully combines artificial intelligence, deep learning and advanced image processing technology, and constructs an efficient and intelligent monitoring system, which can realize accurate personnel monitoring in complex underground environment. By optimizing the YOLOv5 target detection model and combining advanced image preprocessing (such as denoising and enhancement technology) and post-processing algorithm, the system can effectively identify and measure the distance of the personnel, ensuring the accuracy and real-time of the data. The combination of real-time video stream processing and intelligent analysis technology enables the system to quickly capture and respond to potential dangers, greatly improving the early warning ability of safety hazards.

[0048] In the underground environment, due to insufficient light, many obstructions and complex personnel posture, traditional safety monitoring methods often face high false positive rate and false negative rate. The present application can significantly improve the adaptability and robustness of the system in these complex environments through image enhancement, target frame screening, up-down classification and other technical optimization. In addition, the joint training and adaptive safety threshold adjustment method introduced makes the monitoring system able to dynamically adjust the monitoring parameters according to real-time data, ensuring efficient operation and accurate detection in different working conditions.

[0049] Through continuous optimization and innovation of the system, the present application realizes the whole process automation from image acquisition to safety warning, significantly reduces manual intervention, improves monitoring efficiency and response speed. Most importantly, the present application not only can effectively identify and predict the insufficient distance of the personnel, but also can issue an early warning when a safety hazard is found, helping the workers to take timely preventive measures, ensuring the overall safety and reliability of the coal mine underground transportation system. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 is the overall flowchart of the present application.

[0051] Figure 2 is the overall flowchart of the OTA algorithm.

[0052] Figure 3 is the post-processing logic diagram.

[0053] Figure 4 is the actual scene schematic diagram.

[0054] Figure 5 For the actual scene downlink personnel in the same area of the contrast chart. DETAILED DESCRIPTION

[0055] In order to make the technical problems to be solved by the present application, technical solutions and beneficial effects clearer, the present application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.

[0056] As Figures 1-5 shown, a method for detecting insufficient distance of personnel riding on a detachable monkey car, the specific steps are as follows:

[0057] Step S1, data set collection:

[0058] During the data set collection process, a large amount of image and video data containing monkey car personnel and personnel in different safety clothes and safety helmets were collected. These data include personnel images under various lighting conditions, different postures and different distances. In order to ensure the diversity and comprehensiveness of the data, the collection process covers different coal mine tunnels, different monitoring angles, and monkey car scenes of various speeds. At the same time, special consideration is given to the collection effect of the camera at different installation positions to ensure that the data set can truly reflect various situations in actual application. Through multiple experiments, the data collection process is optimized to minimize the influence of light reflection, headlight direct projection and other factors on image quality, thereby laying a solid foundation for subsequent data processing and model training.

[0059] Step S2, data preprocessing:

[0060] In image processing and computer vision tasks, the quality of data directly affects the training effect and prediction accuracy of the model. The original data usually contains various noise and undesirable conditions, such as excessive camera photosensitivity, low light environment, or uneven brightness in the image. These factors can interfere with the learning process of the model, making the model sensitive to noise during training, thereby affecting the final detection and classification performance. Therefore, data preprocessing before model training to remove noise and enhance image quality is a key step to improve the robustness and accuracy of the model.

[0061] The purpose of denoising is to reduce the interference information in the image, especially the light changes or other noise effects in complex environments. The present application uses bilateral filtering (Bilateral Filter), which has the advantage of not only smoothing the image but also better preserving the edge details compared to Gaussian filtering. Bilateral filtering combines spatial information and pixel value information of the image to perform filtering, which can effectively reduce the influence of strong light while avoiding blurring important edge details.

[0062]

[0063] where I(x) is the output image pixel value, I(x i ) is the input image pixel value, W p is the normalized coefficient, σ d controls the spatial distance weight, and σ r controls the pixel value difference weight.

[0064] The purpose of image enhancement is to improve the contrast and clarity of the image, so that the key features are more obvious, which is beneficial to subsequent model training and target detection. Histogram equalization can improve the contrast of the image by redistributing the gray values of the image, and is especially suitable for images with uneven lighting.

[0065]

[0066] where S(i) is the equalized gray level, L is the number of gray levels, MN is the total number of pixels in the image, and h(j) is the number of pixels of gray level j.

[0067] Step S3, data labeling and splitting:

[0068] In the data labeling and splitting stage, first, a professional labeling tool such as LabelImg is selected to accurately label the target area in the image. LabelImg supports multiple labeling methods such as rectangular frame and polygon, and the operation interface is intuitive, which can efficiently complete the labeling task.

[0069] The main rules of data labeling include: not labeling too small targets and half-body targets (targets with only heads or feet), and exporting the labeling results in txt format. These rules help to ensure the high quality of the data set and avoid introducing unnecessary noise data in model training. The labeled data set also needs to be randomly divided into training set, validation set and test set according to the proportion of 7:2:1. This random division method can effectively avoid the bias that may be caused by sequential division, ensure that the data in the training set, validation set and test set are representative, and thus improve the training effect and generalization ability of the model.

[0070] Step S4, improving the YOLOv5 network structure:

[0071] In the scene of target detection for the coal mine underground transportation system, in order to improve the detection accuracy and efficiency of the system for the personnel riding the monkey car, the present application improves the YOLOv5 network structure as follows:

[0072] Step S41. Introducing Optimal Transport Assignment (OTA) algorithm

[0073] In the underground transportation system, target detection needs to handle complex environmental factors such as insufficient lighting and motion blur. Traditional Hungarian algorithm may cause slow detection speed when dealing with these complex scenes. OTA algorithm introduces a new matching method between target and detection box, which can significantly improve the detection speed while maintaining the detection accuracy.

[0074] OTA algorithm converts the distance between target box and prediction box into probability distribution and uses Wasserstein distance for matching, avoiding the high computational problem of traditional algorithm, the specific formula is:

[0075]

[0076] Where C ij represents the distance between the ith target box and the jth prediction box, T ij is the element of the matching matrix, and Ψ is the set of all legal matching matrices. The distance matrix C is converted into probability distribution P and Q, where:

[0077]

[0078] Here τ is the temperature parameter, used to control the smoothness of the probability distribution.

[0079] Step S42, optimize the loss function:

[0080] In the underground transportation system, the angle of the target box may vary due to changes in camera position or lighting conditions. EIOU considers the angular difference between the target box and the prediction box, and improves the detection accuracy by optimizing the angle term, the formula is:

[0081]

[0082] This helps to handle the detection problem of targets at different angles and improve the accuracy of the system in complex environments.

[0083] The target in the coal mine environment may present irregular shape due to occlusion or perspective problem, SIOU introduces a smoothing term to alleviate the shape difference between the target box and the real box, the formula is:

[0084]

[0085] This processing method helps to improve the detection ability of complex target shape and ensures that the system can accurately identify personnel in actual application.

[0086] In the underground transportation system, the degree of overlap of the target may vary due to distance or perspective, and AlphaIOU adjusts the parameter α to balance the degree of overlap between the prediction box and the real box, the formula is:

[0087]

[0088] This allows the model to adjust the emphasis on overlapping parts according to the characteristics of the actual task and data set, thereby optimizing the detection results.

[0089] Through these improvements, YOLOv5 can provide higher detection accuracy and faster processing speed in complex coal mine underground environments, ensuring real-time monitoring and safety of personnel riding the monkey car. These optimizations not only improve the performance of the model in practical applications, but also enhance the system's adaptability to various challenges.

[0090] Step S5, model training and model optimization:

[0091] In the personnel detection task of coal mine underground transportation system, model training and optimization are the key steps to ensure the efficient operation of the system. During the training process, various hyperparameters are first set, including learning rate, learning rate decay parameter, iteration number, batch size, and parameters to prevent overfitting. Reasonable setting of these hyperparameters is crucial for improving the training effect and performance stability of the model.

[0092] To improve detection accuracy and model robustness, the following improved loss functions are introduced during training:

[0093] EIOU (Efficient Intersection over Union): Based on traditional IoU, it adds an angle term, considering the rotational relationship between the detection box and the real box, thereby improving the detection ability of rotating objects.

[0094] SIOU (Smooth Intersection over Union): Based on IoU, it adds a smoothing term to alleviate the shape difference between the detection box and the real box, improving the detection accuracy of objects of various shapes.

[0095] AlphaIOU: Introduces a adjustable parameter α, which is used to optimize the overlap between the real box and the predicted box, so that the model can adapt to different detection requirements and enhance the processing ability of the relationship between the detection box and the real box.

[0096] During model training, the improved loss functions mentioned above are used, and the training is carried out in combination with the set hyperparameters. According to the training and verification indicators, the hyperparameters are fine-tuned to achieve the best effect, which includes adjusting the learning rate, batch size and other hyperparameters, and continuously optimizing the model parameters according to the training curve and verification results. Through these steps, the model can accurately identify and locate personnel riding the monkey car in the coal mine underground transportation environment, thereby enhancing the safety and reliability of the system.

[0097] Step S6, detection result post-processing and alarm triggering;

[0098] In the post-processing stage of the detection result, first of all, the detection boxes with high confidence are screened out by setting a confidence threshold, so as to ensure that only accurate detection results are retained. Then, the detection boxes are subjected to region screening, focusing on the detection boxes within the fixed region; according to the size and position of the detection boxes, the personnel are classified into two categories of uplink and downlink; this process is distinguished by setting specific region boundaries and size ranges, so as to effectively classify personnel in different directions.

[0099] The specific operation steps are as follows:

[0100] Confidence screening: In the results output by the target detection algorithm, each detection box has a corresponding confidence score, which represents the possibility that the detection box contains a target object (here, personnel riding the monkey car). A confidence threshold is set, for example, 0.7. If the confidence score of the detection box is higher than the threshold, it is considered to be reliable and retained; if it is lower than the threshold, it is considered to be a false detection and is removed. Figure 3 In the , after the model detects the target box of the personnel riding the monkey car, confidence judgment is first performed. If the confidence is greater than the threshold, it enters the next step of processing, otherwise the target box is discarded.

[0101] Uplink and downlink classification: According to the actual situation of the underground monkey car operation, the uplink and downlink monitoring regions are pre-set. The position and size of the detection box in the image are judged to determine whether the personnel are in the uplink or downlink state. The uplink and downlink region boundaries and size standards are defined. For example, in the Figure 4 and Figure 5 , it can be seen that the underground scene is divided into different regions, and the direction of the personnel is determined by the position of the target box. If the detection box is located in the upper region of the image (the uplink region is marked by a green box), the personnel are classified as uplink; if it is located in the lower region (the downlink region is marked by a red box), the personnel are classified as downlink. In the Figure 2 , after the target box that passes the confidence screening appears, it is further judged whether it is within the uplink and downlink regions, so as to determine the direction of the personnel and classify the personnel into uplink target boxes and downlink target boxes.

[0102] Interval time calculation: In order to calculate the distance between the personnel riding the monkey car, it is necessary to know the time interval between adjacent personnel in different frames. If the camera itself can record the accurate time of each frame, the time difference between adjacent frames is directly obtained as the interval time. If the camera does not have this function, it is calculated through the known frame rate (for example, 30 frames / second) and the frame number difference. Assuming that the frame rate is f and the frame number difference between adjacent frames is n, then the interval time t = n / f. In the Figure 3In the above, for both the uplink target frame and the downlink target frame, the interval time between adjacent target frames needs to be calculated respectively in order to subsequently estimate the distance.

[0103] Distance estimation: Given the travel speed v of the monkey car and the interval time t between adjacent personnel, the actual distance between personnel is estimated according to the physical formula distance d = v x t. First, determine the travel speed of the monkey car (which can be a fixed value measured and set in advance), then multiply the calculated interval time to get the distance between personnel. In Figure 3 In the above, by calculating the interval time of adjacent target frames and combining the speed of the monkey car, the actual distance between personnel riding the monkey car can be estimated.

[0104] Safety threshold judgment: Set a safety threshold, when the estimated distance between personnel is less than this threshold, it is considered that there is a safety hazard. Compare the calculated distance with the safety threshold (for example, 1 meter). If the distance is less than the safety threshold, the alarm system is triggered. In Figure 3 In the above, when the calculated distance between personnel riding the monkey car is less than the safety threshold, an alarm signal is triggered to inform relevant personnel to take measures to ensure the safety of the underground transportation environment.

[0105] Through such a comprehensive post-processing process, combined with image analysis and physical calculation, the personnel dynamics in the underground transportation environment can be effectively and accurately monitored, ensuring the safe and reliable operation of the system.

[0106] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present application shall be included in the scope of the present application.

Claims

1. A method for detecting insufficient distance for a passenger to be seated in a detachable chairlift passenger car, characterized in that The specific steps are as follows: Step S1, data set collection: by deploying stable camera and recording in different time periods, data collection is carried out by selecting different areas and conditions in the underground transportation system, and data annotation records the position and size of the passengers; Step S2, including denoising and data enhancement, denoising to reduce noise in the image, data enhancement by rotating, flipping, scaling and color transformation operation, normalization and standardization processing; Step S3, data annotation and splitting: Annotation tool selection: use the annotation tool labelimg for image annotation; Annotation quality control: check whether each annotation box accurately covers the target area and correct the annotation error; Data splitting: the annotated data set is split into training set, validation set and test set according to the proportion of 7:2:1, the training set is used for model training, the validation set is used for model tuning and parameter selection, and the test set is used for final performance evaluation; Step S4, improve YOLOv5 network structure: introduce OTA algorithm and advanced loss function, OTA algorithm converts the distance between target and detection box into probability distribution, and minimizes the Wasserstein distance between two probability distributions, advanced loss function optimizes the matching between detection box and real box; Step S5, determine network structure and select training hyperparameters, select YOLOv5 network structure, adjust its layer number, convolution kernel size according to task requirements, set training hyperparameters, select optimal value through grid search, and use standard loss function during training; Step S6, detection result post-processing and alarm triggering: detection result post-processing, by identifying, screening and analyzing each frame of image detected personnel, realizing the calculation of interval time and actual distance of passengers riding monkey car, using target detection algorithm to mark out personnel meeting fixed area and size range, and continuously calculating the time interval of adjacent passengers, according to the speed of monkey car, the actual distance between personnel is estimated, and compared with the preset safety threshold, if the distance is less than the threshold, the system will trigger alarm.

2. A method for detecting insufficient distance for a passenger to be seated in a monkey seat according to claim 1, wherein In the data set collection process of step S1, image and video data containing monkey car personnel and personnel with different safety clothes and safety hats are collected, which includes personnel images in various light conditions, different postures and different distances, and the collection process covers different coal mine tunnels, different monitoring angles and various speed monkey car scenes.

3. A method for detecting insufficient distance for a passenger to be seated in a monkey seat according to claim 2, wherein In step S2, bilateral filtering is used for denoising, and the specific denoising process is as follows: where I(x) is the output image pixel value, I(x i ) is the input image pixel value, W p is the normalization coefficient, σ d controls the spatial distance weight, and σ r controls the pixel value difference weight. In the image enhancement process, histogram equalization is used to enhance the contrast of the image, and the specific process is as follows: Where S(i) is the equalized gray level, L is the number of gray levels, MN is the total number of pixels of the image, and h(j) is the number of pixels of gray level j.

4. A method for detecting insufficient distance for a passenger to be seated in a monkey seat according to claim 3, wherein In step S4, for the matching of target box and prediction box, OTA algorithm is used to convert the distance between target box and prediction box into probability distribution, and Wasserstein distance is used for matching, and the specific process is described by the following formula: where C ij represents the distance between the ith target box and the jth predicted box, T ij is an element of the matching matrix, Ψ is the set of all legal matching matrices, the distance matrix C is transformed into probability distributions P and Q, where: Tau is a temperature parameter used to control the smoothness of the probability distribution. The algorithm calculates the minimum transportation cost between the target frame and the predicted frame to obtain the matching result. Optimize the loss function: Introduce the angle difference term: consider the angle difference between the target frame and the predicted frame, and optimize the angle term to improve the detection accuracy. The specific formula is: where w t is the width of the target box, h t is the height of the target box, w p is the width of the predicted box, and h p is the height of the predicted box. Introduce the smoothing term to reduce the shape difference between the target frame and the real frame. The specific formula is: Adjust the parameter a to balance the overlap: adjust the parameter a to balance the overlap between the predicted frame and the real frame. The formula is: Where a is the adjustment parameter, and ∈ is a very small constant.

5. A method for detecting insufficient distance for a passenger to be seated in a monkey seat according to claim 4, wherein In step S6, the following steps are included: Confidence screening: in the detection process, first apply the confidence threshold to all detection frames, and use the set threshold to remove frames below the threshold; Up and down classification: according to the position, size and preset area boundary of the detection frame, further combined with the actual up and down monitoring area, the personnel are classified as up or down, by dividing the monitoring area into multiple sub-areas and setting size standards for each sub-area to realize; Interval time calculation: when calculating the time interval between adjacent frames, if the video stream can directly record time information, directly read and use the timestamp; otherwise, based on the interval frame number and frame rate to calculate the time interval, through the frame rate and frame interval to estimate the time interval of each frame, to provide data input for subsequent distance estimation; Distance estimation: according to the calculated interval time and the travel speed of the monkey car, the actual distance between the personnel riding the monkey car is estimated using the physical formula; Distance estimation: using the calculated time interval and the known travel speed of the monkey car, the actual distance between the personnel is estimated through the physical formula, the specific formula is: D = V * T Where D is the actual distance between the personnel, V is the travel speed of the monkey car, and T is the time interval of the two personnel detection frames. The relative distance between the personnel is calculated through this formula; Safety threshold judgment: after calculating the actual distance between the personnel riding the monkey car, the system compares the distance with the preset safety threshold. If the distance is less than the safety threshold, the alarm system is triggered.

Images