Radar monitoring and early warning method and system based on mine slope instability trend

Abstract

The present invention relates to the field of slope monitoring technology, and more specifically to a radar monitoring and early warning method and system based on the instability trend of mine slopes. The method comprises the following steps: using a pre-trained Fast R-CNN model to identify slope areas at risk of instability on a geological image as instability risk areas; using a slope radar to scan the instability risk areas in real time to obtain a radar image sequence of the instability risk areas; obtaining a geological image sequence of the instability risk areas that is co-sequential with the radar image sequence, and determining the instability trend of the instability risk areas based on the radar image sequence and the geological image sequence, and then assigning an early warning level based on the instability trend. The present invention uses the Fast R-CNN model to achieve early locking of instability factors, and based on the LSTM network, quickly and in real time analyzes the sliding trajectory trends of falling rocks and loose soil based on the actual trajectory of falling rocks and loose soil obtained by solving multimodal data, thereby achieving integrated display of risk points in the mining area, focused monitoring, and graded early warning.

Description

Technical Field

[0001] The present invention relates to the technical field of slope monitoring, and in particular to a radar monitoring and early warning method and system based on the instability trend of a mine slope. Background Art

[0002] Radar technology is widely used in mine slope monitoring. By monitoring slope displacement, deformation, and cracks, it can promptly detect potential landslide hazards. The mine slope radar early warning mechanism process can take different response measures based on different warning levels, effectively reducing the probability of accidents.

[0003] The rise and development of deep learning-based computer vision algorithms and the advancement of computing power have ushered in a new era of research in slope hazard monitoring technology. Convolutional Neural Networks (CNNs), a classic deep learning algorithm, are highly capable of extracting texture, color, shape, and other feature information from two-dimensional images. Therefore, for slope hazard identification, CNN-based algorithms such as object recognition and semantic segmentation are primarily used to process slope hazard image data, primarily from high-resolution satellite imagery and drone footage. Xu et al. combined the cloud platform Google Earth Engine (GEE) with the U-Net semantic segmentation network to segment and identify post-earthquake landslides in Landsat imagery, enabling earthquake landslide interpretation and playing a key role in post-disaster reconstruction efforts.

[0004] While existing technologies using deep learning can achieve some success in slope hazard identification, they are unable to identify instability factors (i.e., rockfall and loose soil areas prone to slope hazard conditions) in advance, nor can they quickly and in real time analyze instability trends (i.e., the trajectory of rockfall and loose soil) after the instability factors become unstable, thus enabling a combined dynamic and static slope monitoring process for graded early warnings. Therefore, current slope monitoring and early warning systems rely on indiscriminate or unfocused regional monitoring, which has a wide monitoring range and requires a large amount of data processing. Deep learning algorithms take a long time to detect hazard areas, or even make it impossible to detect hazard areas. This means that deep learning technology's disaster identification accuracy and efficiency are insufficient. Furthermore, current disaster identification relies solely on landslide interpretation based on established disaster facts, and it is impossible to predict the potential impact of a disaster shortly after it occurs, and then provide graded early warnings based on the impact. Summary of the Invention

[0005] The purpose of the present invention is to provide a radar monitoring and early warning method and system based on the instability trend of mine slopes, so as to solve the technical problems in the existing technology that the slope disaster monitoring range is wide and the data processing volume is large, resulting in insufficient disaster identification accuracy and efficiency, and it is difficult to predict the possible impact of disasters and provide graded early warnings, and the early warning flexibility is limited.

[0006] In order to solve the above technical problems, the present invention specifically provides the following technical solutions:

[0007] The radar monitoring and early warning method based on the mine slope instability trend includes the following steps:

[0008] Collect geological images of the open-pit mine slope and use the pre-trained Fast R-CNN model to identify slope areas with instability risks on the geological images as instability risk areas;

[0009] The instability risk area is set as the monitoring area of ​​the slope radar, and the instability risk area is scanned in real time by the slope radar to obtain a radar image sequence of the instability risk area;

[0010] Acquire a geological image sequence of the instability risk area that is contemporaneous with the radar image sequence, determine an instability trend of the instability risk area based on the radar image sequence and the geological image sequence, and assign an early warning level based on the instability trend;

[0011] The method for determining the instability trend includes:

[0012] An image conversion model for converting geological images into radar images is constructed using the CycleGAN network structure;

[0013] The geological images in the geological image sequence of the instability risk area are converted into radar images through an image conversion model to generate a radar conversion image sequence of the instability risk area;

[0014] Fusing the radar conversion image sequence of the instability risk area with the radar image sequence of the instability risk area to generate a radar multi-mode image sequence of the instability risk area;

[0015] Solving the local instability trajectory of falling rocks and loose soil that characterizes the instability trend in the radar multi-mode image sequence of the instability risk area, and predicting the global instability trajectory of falling rocks and loose soil based on the local instability trajectory through an LSTM network structure;

[0016] Graded warnings are issued for instability risk areas in open-pit mines based on the warning levels.

[0017] As a preferred solution of the present invention, the pre-training method of the Fast R CNN model includes:

[0018] Obtain multiple geological images and mark areas containing fallen rocks and loose soil in each geological image as instability risk areas;

[0019] Divide the dataset consisting of multiple geological images into a test set and a training set;

[0020] On the training set, the Fast R-CNN model is trained with geological images as input and the instability risk areas of the geological images as output;

[0021] On the test set, the Fast R-CNN model is evaluated for its performance in identifying unstable risk areas.

[0022] As a preferred solution of the present invention, the method for constructing the image conversion model includes:

[0023] Acquiring a plurality of geological images of the instability risk area and acquiring a plurality of radar images of the instability risk area at the same time as the geological images;

[0024] A data set consisting of multiple geological images of instability risk areas and multiple radar images of instability risk areas is divided into a training set and a test set;

[0025] On the training set, the first-layer GAN network structure in the CycleGAN network structure is used to convert geological images of instability risk areas into radar images of instability risk areas, and the second-layer GAN network structure in the CycleGAN network structure is used to convert radar images of instability risk areas into geological images of instability risk areas. The CycleGAN network structure is trained to obtain an image conversion model.

[0026] On the test set, the image conversion model is evaluated for its image modality conversion performance;

[0027] in,

[0028] The loss function of training the CycleGAN network structure is the cycle consistency loss Lcycle and the adversarial generation loss L GAN On this basis, the coordinate consistency loss L of falling rocks and loose soil is added P ;

[0029] ;

[0030] Where, is the radar image of the instability risk area converted from the geological image of the instability risk area, is the geological image of the instability risk area converted from the radar image of the instability risk area, x is the geological image of the instability risk area, y is the radar image of the instability risk area, Because the Fast RCNN model The coordinates of the falling rocks and loose soil are obtained by identification. Because the Fast RCNN model The coordinates of the falling rocks and loose soil are obtained by identification. In order to obtain the coordinates of rockfall and loose soil positions identified by the Fast R CNN model on x, In order to obtain the coordinates of rockfall and loose soil positions identified by the Fast R-CNN model on y, is the L1 norm, and Both are mathematical expectations.

[0031] As a preferred solution of the present invention, the method for generating the radar multi-mode image sequence of the instability risk area includes:

[0032] By using a cross-attention mechanism, the radar image sequence is mapped as a query matrix, and the radar conversion image sequence is mapped as a key and value matrix to calculate the attention weight of the radar image sequence;

[0033] By using a cross-attention mechanism, the radar conversion image sequence is mapped as a query matrix, and the radar image sequence is mapped as a key and value matrix, and the attention weight of the radar conversion image sequence is calculated;

[0034] The radar image sequence weighted by the attention weight of the radar image sequence and the radar converted image sequence weighted by the attention weight of the radar converted image sequence are added to obtain the radar multimode image sequence.

[0035] As a preferred solution of the present invention, the calculation and solution method of the local instability trajectory includes:

[0036] In the radar multimode image sequence, phase difference processing is performed on two adjacent radar multimode images in sequence to generate an interference pattern sequence. , where is the interference pattern at the tth time sequence, is the phase information of the radar multimode image at the t+1th time sequence, is the phase information of the radar multimode image at the tth time sequence, and m is the total number of time sequences of the radar multimode image sequence;

[0037] Based on the phase difference information in the interference pattern sequence, the motion trajectory of falling rocks and loose soil in the instability risk area is calculated as the local instability trajectory. , where is the displacement of falling rocks and loose soil at the tth time sequence, is the radar wavelength.

[0038] As a preferred solution of the present invention, the method for predicting the global instability trajectory includes:

[0039] The LSTM network structure is used to predict the local instability trajectory and obtain the global instability trajectory. , where is the motion displacement of fallen rocks and loose soil at the tth time series, m is the total number of time series of radar multi-mode image sequences, and n is the total number of subsequent predicted time series.

[0040] As a preferred solution of the present invention, the method for allocating warning levels includes:

[0041] calculate The distance from the mine protection area is If the distance to the mine protection zone is less than the preset safety distance, the instability risk area is assigned a high risk level;

[0042] when If the distance to the mine protection zone is greater than or equal to the preset safety distance, the instability risk area is assigned a medium or low risk level.

[0043] The present invention provides a radar monitoring and early warning system based on the instability trend of mine slopes. According to the above-mentioned radar monitoring and early warning method based on the instability trend of mine slopes, the system includes:

[0044] Visual equipment for collecting geological images;

[0045] Slope radar, used to collect radar images;

[0046] a data processor configured to identify slope areas with instability risks on the geological image using a pre-trained Fast R-CNN model as instability risk areas; determine an instability trend of the instability risk areas based on the radar image sequence and the geological image sequence, and assign a warning level based on the instability trend;

[0047] The display device is equipped with software for graded early warning of open-pit mine slope monitoring, which is used to display radar images of open-pit mines, mark early warning areas on the radar images in a graded manner, display instability trends, and display early warning pop-ups.

[0048] As a preferred solution of the present invention, the visual equipment includes remote sensing image acquisition equipment and drone camera equipment.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The present invention uses the Fast R-CNN model to early identify instability factors (i.e., rockfall and loose soil areas prone to slope disasters). Furthermore, based on the LSTM network, the present invention quickly and in real time analyzes the sliding trajectory trends of rockfall and loose soil based on the actual trajectories of rockfall and loose soil obtained by solving multimodal data, thereby providing corresponding graded early warnings. This realizes the integrated display of risk points in the mining area, focused monitoring, and graded early warning, thus improving the effectiveness of slope monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other implementation drawings based on the provided drawings without inventive effort.

[0052] Figure 1 A flow chart of a radar monitoring and early warning method based on mine slope instability trends provided by an embodiment of the present invention;

[0053] Figure 2 This is a structural block diagram of a radar monitoring and early warning system based on mine slope instability trends provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0055] like Figure 1 As shown, the present invention provides a radar monitoring and early warning method based on the instability trend of mine slopes, comprising the following steps:

[0056] Collect geological images of the open-pit mine slope and use the pre-trained Fast R-CNN model to identify slope areas at risk of instability on the geological images as instability risk areas;

[0057] The instability risk area is set as the monitoring area of ​​the slope radar, and the instability risk area is scanned in real time by the slope radar to obtain a radar image sequence of the instability risk area;

[0058] Acquire a geological image sequence of the instability risk area that is synchronous with the radar image sequence, determine the instability trend of the instability risk area based on the radar image sequence and the geological image sequence, and then assign an early warning level based on the instability trend;

[0059] Graded warnings are issued for instability risk areas in open-pit mines based on the warning levels.

[0060] The present invention collects geological images (such as satellite remote sensing images and drone-photographed images) of the open-pit mine slope during the silent period. The Fast R-CNN model with image segmentation function is used to divide the geological images into areas containing disaster elements such as fallen rocks and loose soil. The fallen rocks and loose soil elements in this area are more likely to cause instability and have a hazard impact. For example, they may roll into the mining protection area and injure people or damage houses and construction equipment. In other words, there is an instability risk in this area. Therefore, the present invention identifies such areas of the slope with instability risk during the silent period, locks the instability risk areas in advance, and then performs radar scanning and real-time monitoring on them to achieve zoned monitoring for slope early warning.

[0061] The present invention can obtain radar image data of the instability outbreak period (rockfall, loose soil sliding) in the instability risk area in real time through real-time scanning of the slope radar. At the same time, geological image data synchronized with the radar image data of the outbreak period is obtained, and the image data of these two modes are integrated to solve the actual trajectory of the rockfall and loose soil sliding during the instability outbreak period. The LSTM network with time series prediction function is used to predict the trajectory of subsequent rockfall and loose soil sliding based on the actual trajectory of the rockfall and loose soil sliding during the instability outbreak period. The instability trend can be grasped in advance a short period of time after the outbreak period occurs, and the instability trend, that is, the trajectory of rockfall and loose soil sliding, can be quickly analyzed in real time after the instability factor becomes unstable. Whether a disaster impact will occur is judged according to the instability trend, and a graded warning is given based on the degree of disaster impact.

[0062] The present invention solves the actual trajectory of rockfall and loose soil sliding during the instability outbreak period by fusing image data of two modes, radar image data and geological image data. It can obtain the accuracy advantage of solving by multi-modal data fusion, improve the accuracy of solving the actual trajectory, and thus accurately grasp the actual trajectory of rockfall and loose soil sliding during the outbreak period. Further improvement in the accuracy of the actual trajectory can ensure the improvement in the accuracy of the predicted trajectory of subsequent rockfall and loose soil sliding after the actual trajectory, thereby ensuring the accuracy of slope disaster warning.

[0063] The image conversion network constructed through the CycleGAN network structure in the present invention can convert geological images into radar image form, thereby converting geological image data synchronized with the radar image data during the eruption period into radar image form (i.e., radar converted images). The radar converted images and radar images are fused to achieve multimodal data fusion. The fused radar multimodal images are used to calculate the actual trajectories of falling rocks and loose soil. Compared with calculations using only radar images, the multimodal data provides richer information features, reduces the random errors of single-modal data, and more accurately locates falling rocks and loose soil, thereby making the calculation of the actual trajectories of falling rocks and loose soil more accurate.

[0064] The present invention combines the attention mechanism in multimodal data fusion, and utilizes the key information focusing characteristics of the attention mechanism to give high weight to important information in the radar conversion image and the radar image, highlighting the key performance in the two modal images, and realizing information interaction between the two images of different modalities, thereby improving the cross-modal understanding ability of the model, helping the model to better analyze character features, and improving the quality of image multimodal fusion. Therefore, through the multimodal fusion of the attention mechanism, the radar conversion image and the radar image can be better improved, retaining the key information of each modality, and at the same time interacting with the location information of fallen rocks and loose soil sliding, to obtain accurate location information of fallen rocks and loose soil sliding.

[0065] The present invention adopts phase interferometry to calculate the actual trajectory of falling rocks and loose soil, and calculates the actual trajectory of falling rocks and loose soil through a radar multi-mode image sequence of a short period of time during the burst period.

[0066] After calculating the actual trajectories of falling rocks and loose soil sliding, the present invention predicts the trajectories of subsequent falling rocks and loose soil sliding based on the actual trajectories of falling rocks and loose soil sliding during the instability outbreak period through an LSTM network with a time series prediction function. The instability trend can be grasped in advance a short period of time after the outbreak period occurs, and the instability trend can be quickly analyzed in real time after the instability factor becomes unstable.

[0067] After predicting the trajectory of subsequent rockfalls and loose soil slides, the present invention determines whether the trajectory has invaded the mine protection zone, that is, whether substantial damage will be caused. If it has occurred, it will be an emergency event and a high-level warning will be issued. If it has not occurred, a medium or low-level warning will be issued. The graded warning of the present invention is consistent with the actual situation, and the accuracy of the graded warning is improved.

[0068] The present invention collects geological images (such as satellite remote sensing images and drone images) of the open-pit mine slope during its quiet period. The Fast R-CNN model with image segmentation function is used to segment the geological images to identify areas containing hazard elements such as fallen rocks and loose soil. The specific details are as follows:

[0069] Fast R-CNN (Fast Region-based Convolutional Neural Network, an improved object detection model designed to improve the speed and efficiency of object detection while maintaining or improving detection accuracy) pre-training methods include:

[0070] Obtain multiple geological images and mark areas containing fallen rocks and loose soil in each geological image as instability risk areas;

[0071] Divide the dataset consisting of multiple geological images into a test set and a training set;

[0072] On the training set, the Fast R-CNN model is trained with geological images as input and the instability risk areas of the geological images as output;

[0073] On the test set, the Fast R-CNN model is evaluated for its performance in identifying unstable risk areas.

[0074] The present invention uses an image conversion network constructed through the CycleGAN network structure to convert geological images into radar image form, thereby converting geological image data synchronized with the eruption period radar image data into radar image form (i.e., radar-converted images). The radar-converted images are fused with the radar images to achieve multimodal data fusion. The fused radar multimodal images are used to calculate the actual trajectories of falling rocks and loose soil. Compared with calculations using only radar images, multimodal data provides richer information features, reduces the random errors of single-modal data, and more accurately locates falling rocks and loose soil, thereby making the calculation of the actual trajectories of falling rocks and loose soil more accurate, as follows:

[0075] Methods for determining instability trends include:

[0076] An image conversion model for converting geological images into radar images was constructed using the CycleGAN network structure (Cycle-Consistent Generative Adversarial Networks, a deep learning model for unsupervised image-to-image conversion. Its main goal is to achieve image conversion between different domains without the need for paired training data. CycleGAN uses two generators and two discriminators, and utilizes cycle consistency loss and adversarial loss to ensure that the image maintains the consistency of its main content and structure during the conversion process).

[0077] The geological images in the geological image sequence of the instability risk area are converted into radar images through an image conversion model to generate a radar conversion image sequence of the instability risk area;

[0078] Fusing the radar conversion image sequence of the instability risk area with the radar image sequence of the instability risk area to generate a radar multi-mode image sequence of the instability risk area;

[0079] The local instability trajectories of falling rocks and loose soil that represent the instability trend are solved in the radar multi-mode image sequence of the instability risk area, and the global instability trajectories of falling rocks and loose soil are predicted based on the local instability trajectories through the LSTM network structure.

[0080] The construction method of the image conversion model includes:

[0081] Acquire multiple geological images of the instability risk area and acquire multiple radar images of the instability risk area simultaneously with the geological images;

[0082] A data set consisting of multiple geological images of instability risk areas and multiple radar images of instability risk areas is divided into a training set and a test set;

[0083] On the training set, the first-layer GAN network structure in the CycleGAN network structure is used to convert geological images of instability risk areas into radar images of instability risk areas, and the second-layer GAN network structure in the CycleGAN network structure is used to convert radar images of instability risk areas into geological images of instability risk areas. The CycleGAN network structure is trained to obtain an image conversion model.

[0084] On the test set, the image conversion model is evaluated for its image modality conversion performance;

[0085] Among them, the loss function of training the CycleGAN network structure is the cycle consistency loss Lcycle and the adversarial generation loss L GAN On this basis, the coordinate consistency loss L of falling rocks and loose soil is added P ;

[0086] ;

[0087] Where, is the radar image of the instability risk area converted from the geological image of the instability risk area, is the geological image of the instability risk area converted from the radar image of the instability risk area, x is the geological image of the instability risk area, y is the radar image of the instability risk area, Because the Fast RCNN model The coordinates of the falling rocks and loose soil are obtained by identification. Because the Fast RCNN model The coordinates of the falling rocks and loose soil are obtained by identification. In order to obtain the coordinates of rockfall and loose soil positions identified by the Fast R CNN model on x, In order to obtain the coordinates of rockfall and loose soil positions identified by the Fast R-CNN model on y, is the L1 norm, and Both are mathematical expectations.

[0088] in:

[0089] ;

[0090] ;

[0091] Where G is the generator that converts geological images into radar images, F is the generator that converts radar images into geological images, and D x is the discriminator, which determines whether the input is in the form of radar image, D y is the discriminator, which determines whether the input belongs to the form of geological image, pdata(x): the real data distribution of geological image, pdata(y): the real data distribution of radar image, E: expected value operator, which means taking the average of all samples in the data distribution, : Discriminator D y right The discriminant result is taken as logarithm, and the goal is to maximize this value (to encourage the discriminator to correctly identify the real sample). : Discriminator Pair Generation The discriminant result of is logarithmically complemented, and the goal is to maximize this value (encourage the generator to deceive the discriminator). Similarly, : Discriminator right The discriminant result is taken as logarithm, and the goal is to maximize this value (to encourage the discriminator to correctly identify the real sample). : Discriminator Pair Generation The goal is to maximize the logarithmic complement of the discriminant result (encourage the generator to deceive the discriminator).

[0092] The present invention additionally supplements the coordinate consistency loss of falling rocks and loose soil It is expected that the position coordinates of fallen rocks and loose soil in the radar image generated by geological image conversion can be accurately mapped, so as to ensure that the position coordinates of fallen rocks and loose soil are still accurate after the image form conversion, and provide accurate position coordinate information of fallen rocks and loose soil for subsequent multimodal image fusion, so as to integrate the position of fallen rocks and loose soil in geological images into radar images, so that the multimodal data can provide richer position information, reduce the random error of the position data of single modal data, and locate fallen rocks and loose soil more accurately, so as to make the calculation of the actual trajectory of fallen rocks and loose soil more accurate.

[0093] The present invention combines the attention mechanism in multimodal data fusion, and utilizes the key information focusing characteristics of the attention mechanism to give high weight to important information in the radar conversion image and radar image, highlighting the key performance of the two modal images, and realizing information interaction between the two images of different modalities, thereby improving the cross-modal understanding ability of the model, helping the model to better analyze character features, and improving the quality of image multimodal fusion. Therefore, through the multimodal fusion of the attention mechanism, the radar conversion image and radar image can be better improved, retaining the key information of each modality, while exchanging the location information of fallen rocks and loose soil sliding, and obtaining accurate location information of fallen rocks and loose soil sliding, as follows:

[0094] The method for generating a radar multi-mode image sequence of an instability risk area includes:

[0095] Through the cross-attention mechanism, the radar image sequence is mapped as the query matrix, and the radar conversion image sequence is mapped as the key and value matrix to calculate the attention weight of the radar image sequence;

[0096] Through the cross-attention mechanism, the radar conversion image sequence is mapped as the query matrix, and the radar image sequence is mapped as the key and value matrix to calculate the attention weight of the radar conversion image sequence;

[0097] , where Q, K, and V are Query, Key, and Value respectively. It is a scaling factor to prevent the dot product result from being too large and causing the gradient to disappear.

[0098] The radar image sequence weighted by the attention weight of the radar image sequence and the radar converted image sequence weighted by the attention weight of the radar conversion image sequence are added to obtain a radar multi-mode image sequence.

[0099] The present invention uses phase interferometry to calculate the actual trajectory of falling rocks and loose soil. The actual trajectory of falling rocks and loose soil is calculated by using a radar multi-mode image sequence of a short period of time during the burst period. The specific details are as follows:

[0100] The calculation and solution methods for local instability trajectories include:

[0101] In the radar multimode image sequence, phase difference processing is performed on two adjacent radar multimode images in sequence to generate an interference pattern sequence. , where is the interference pattern at the tth time sequence, is the phase information of the radar multimode image at the t+1th time sequence, is the phase information of the radar multimode image at the tth time sequence, and m is the total number of time sequences of the radar multimode image sequence;

[0102] Based on the phase difference information in the interference pattern sequence, the motion trajectory of falling rocks and loose soil in the instability risk area is calculated as the local instability trajectory. , where is the displacement of falling rocks and loose soil at the tth time sequence, is the radar wavelength.

[0103] After calculating the actual trajectory of rockfall and loose soil sliding, the present invention uses an LSTM network with a time series prediction function to predict the trajectory of subsequent rockfall and loose soil sliding based on the actual trajectory of rockfall and loose soil sliding during the instability outbreak period. This allows the instability trend to be grasped in advance a short period of time after the outbreak period occurs, and the instability trend can be quickly analyzed in real time after the instability factor occurs. The details are as follows:

[0104] The prediction methods of global instability trajectory include:

[0105] The LSTM network structure (Long Short-Term Memory, a special recurrent neural network designed to solve the long-term dependency problem faced by traditional RNN when processing long sequence data) is used to predict the local instability trajectory and obtain the global instability trajectory. , where is the displacement of the rockfall or loose soil at the tth time sequence, m is the total number of time sequences of the radar multi-mode image sequence, and n is the total number of subsequent predicted time sequences. After the present invention predicts the trajectory of the subsequent rockfall or loose soil sliding, it determines whether its trajectory has invaded the mine protection zone, that is, whether it will cause substantial damage. If it has, it will be an emergency event and a high-level warning should be issued. If it has not, a medium or low-level warning should be issued. The graded warning of the present invention is consistent with the actual situation and the accuracy of the graded warning is improved, as follows:

[0106] The methods used to assign alert levels include:

[0107] calculate The distance from the mine protection area is If the distance to the mine protection zone is less than the preset safety distance, the instability risk area is assigned a high risk level;

[0108] when If the distance to the mine protection zone is greater than or equal to the preset safety distance, the instability risk area is assigned a medium or low risk level.

[0109] like Figure 2 As shown, the present invention provides a radar monitoring and early warning system based on the instability trend of mine slopes. According to the above-mentioned radar monitoring and early warning method based on the instability trend of mine slopes, the system includes:

[0110] Visual equipment for collecting geological images;

[0111] Slope radar, used to collect radar images;

[0112] A data processor is used to identify slope areas with instability risks on geological images using a pre-trained Fast R-CNN model as instability risk areas; determine the instability trend of the instability risk areas based on the radar image sequence and the geological image sequence, and then assign a warning level based on the instability trend;

[0113] The display device is equipped with software for graded early warning of open-pit mine slope monitoring. It is used to display radar images of open-pit mines, integrate and display instability risk areas, mark early warning areas in grades on the radar images, display instability trends, and issue early warning pop-ups.

[0114] Visual equipment includes remote sensing image acquisition equipment, drone camera equipment, etc., so as to obtain high-resolution satellite remote sensing images, drone-photographed images, etc.

[0115] The present invention uses the Fast R-CNN model to early identify instability factors (i.e., rockfall and loose soil areas prone to slope disasters). Furthermore, based on the LSTM network, the present invention quickly and in real time analyzes the sliding trajectory trends of rockfall and loose soil based on the actual trajectories of rockfall and loose soil obtained by solving multimodal data, thereby providing corresponding graded early warnings. This realizes the integrated display of risk points in the mining area, focused monitoring, and graded early warning, thus improving the effectiveness of slope monitoring.

[0116] The above embodiments are merely exemplary embodiments of the present application and are not intended to limit the scope of the present application. The scope of protection of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and scope of protection of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the scope of protection of the present application.

Claims

1. A radar monitoring and early warning method based on the instability trend of mine slopes, characterized by: The following steps are involved: Collect geological images of the open-pit mine slope and use the pre-trained Fast R-CNN model to identify slope areas at risk of instability on the geological images of the open-pit mine slope as instability risk areas; The instability risk area is set as the monitoring area of ​​the slope radar, and the instability risk area is scanned in real time by the slope radar to obtain a radar image sequence of the instability risk area; Acquire a geological image of the instability risk area at the same time as the radar image sequence as a geological image sequence of the instability risk area; determining an instability trend of the instability risk area based on the radar image sequence and the geological image sequence of the instability risk area, and then assigning an early warning level according to the instability trend; The method for determining the instability trend includes: An image conversion model for converting geological images into radar images is constructed using the CycleGAN network structure; The geological images in the geological image sequence of the instability risk area are converted into radar images through an image conversion model to generate a radar conversion image sequence of the instability risk area; Fusing the radar conversion image sequence of the instability risk area with the radar image sequence of the instability risk area to generate a radar multi-mode image sequence of the instability risk area; Solving the local instability trajectory of falling rocks and loose soil that characterizes the instability trend in the radar multi-mode image sequence of the instability risk area, and predicting the global instability trajectory of falling rocks and loose soil based on the local instability trajectory through an LSTM network structure; Graded warnings are issued for instability risk areas in open-pit mines based on the warning levels.

2. The radar monitoring and early warning method based on the mine slope instability trend according to claim 1 is characterized in that: The pre-training method of the Fast R-CNN model includes: Obtain multiple geological images of the open pit mine slopes and mark the areas containing fallen rocks and loose soil in each geological image as instability risk areas; The dataset consisting of multiple geological images of open-pit mine slopes is divided into a test set and a training set; On the training set, the Fast R-CNN model is trained with the geological image of the open-pit mine slope as input and the instability risk area in the geological image of the open-pit mine slope as output; On the test set, the Fast R-CNN model is evaluated for its performance in identifying unstable risk areas.

3. The radar monitoring and early warning method based on the mine slope instability trend according to claim 2 is characterized in that: The method for constructing the image conversion model includes: Acquire multiple geological images of the instability risk area, and acquire multiple radar images of the instability risk area at the same time as the geological images of the instability risk area; A data set consisting of multiple geological images of instability risk areas and multiple radar images of instability risk areas is divided into a training set and a test set; On the training set, the first-layer GAN network structure in the CycleGAN network structure is used to convert geological images of instability risk areas into radar images of instability risk areas, and the second-layer GAN network structure in the CycleGAN network structure is used to convert radar images of instability risk areas into geological images of instability risk areas. The CycleGAN network structure is trained to obtain an image conversion model. On the test set, the image conversion model is evaluated for its image modality conversion performance; Among them, the loss function of training the CycleGAN network structure is the cycle consistency loss Lcycle and the adversarial generation loss L GAN On this basis, the coordinate consistency loss L of falling rocks and loose soil is added P ; Where, is the radar image of the instability risk area converted from the geological image of the instability risk area, is the geological image of the instability risk area converted from the radar image of the instability risk area, x is the geological image of the instability risk area, y is the radar image of the instability risk area, The Fast R CNN model is used in The coordinates of the falling rocks and loose soil are obtained by identification. The Fast R CNN model is used in The coordinates of the falling rocks and loose soil are obtained by identification. In order to obtain the coordinates of rockfall and loose soil positions identified by the Fast R CNN model on x, In order to obtain the coordinates of rockfall and loose soil positions identified by the Fast R-CNN model on y, is the L1 norm, and Both are mathematical expectations.

4. The radar monitoring and early warning method based on the mine slope instability trend according to claim 3 is characterized in that: The method for generating a radar multi-mode image sequence of the instability risk area includes: By using a cross-attention mechanism, the radar image sequence is mapped as a query matrix, and the radar conversion image sequence is mapped as a key and value matrix to calculate the attention weight of the radar image sequence; By using a cross-attention mechanism, the radar conversion image sequence is mapped as a query matrix, and the radar image sequence is mapped as a key and value matrix, and the attention weight of the radar conversion image sequence is calculated; The radar image sequence weighted by the attention weight of the radar image sequence and the radar converted image sequence weighted by the attention weight of the radar converted image sequence are added to obtain the radar multimode image sequence.

5. The radar monitoring and early warning method based on the mine slope instability trend according to claim 4 is characterized in that: The calculation and solution method of the local instability trajectory includes: In the radar multimode image sequence, phase difference processing is performed on two adjacent radar multimode images in sequence to generate an interference pattern sequence. , where is the interference pattern at the tth time sequence, is the phase information of the radar multimode image at the t+1th time sequence, is the phase information of the radar multimode image at the tth time sequence, and m is the total number of time sequences of the radar multimode image sequence; Based on the phase difference information in the interference pattern sequence, the motion trajectory of falling rocks and loose soil in the instability risk area is calculated as the local instability trajectory. , where is the displacement of falling rocks and loose soil at the tth time sequence, is the radar wavelength.

6. The radar monitoring and early warning method based on the mine slope instability trend according to claim 5 is characterized in that: The method for predicting the global instability trajectory includes: The LSTM network structure is used to predict the local instability trajectory and obtain the global instability trajectory. , where is the motion displacement of fallen rocks and loose soil at the tth time series, m is the total number of time series of radar multi-mode image sequences, and n is the total number of subsequent predicted time series.

7. The radar monitoring and early warning method based on the mine slope instability trend according to claim 6 is characterized in that: The method for allocating the warning level includes: calculate The distance from the mine protection area is If the distance to the mine protection zone is less than the preset safety distance, the instability risk area is assigned a high risk level; when If the distance to the mine protection zone is greater than or equal to the preset safety distance, the instability risk area is assigned a medium or low risk level.

8. A radar monitoring and early warning system based on the instability trend of mine slopes is characterized by: The radar monitoring and early warning method based on the mine slope instability trend according to any one of claims 1 to 7 comprises: Visual equipment for collecting geological images; Slope radar, used to collect radar images; a data processor configured to identify, using a pre-trained Fast R-CNN model, slope areas with instability risks on geological images of open-pit mine slopes as instability risk areas; determine instability trends of the instability risk areas based on the radar image sequence and the geological image sequence of the instability risk areas; and assign warning levels based on the instability trends; The display device is equipped with software for graded early warning of open-pit mine slope monitoring, which is used to display radar images of open-pit mines, mark early warning areas on the radar images in a graded manner, display instability trends, and display early warning pop-ups.

9. The radar monitoring and early warning system based on mine slope instability trend according to claim 8 is characterized in that: The visual equipment includes remote sensing image acquisition equipment and drone camera equipment.

Images