A polyester fiber protective net abnormality monitoring system and monitoring method

Abstract

The present invention relates to the technical field of abnormal monitoring of protective nets, specifically to a polyester fiber protective net abnormal monitoring system and monitoring method, including an abnormal map construction unit, an abnormal sorting unit, an inspection sorting unit, an abnormal inspection unit, a video detection unit and an abnormal early warning unit, and also includes: an image acquisition unit, the image acquisition unit is used to set a plurality of image acquisition nodes in the polyester fiber protective net in the target area, set an image acquisition device at the image acquisition node, and use the image acquisition device to perform image acquisition on the polyester fiber protective net segment around the image acquisition node to obtain a plurality of polyester fiber protective net segment images, and transmit the plurality of polyester fiber protective net segment images to an image enhancement unit. The present invention generates a drone inspection task sequence by sorting abnormal nodes to achieve rapid inspection of abnormal areas, and also uses an ant algorithm to optimize routes during the inspection process to ensure the efficiency and comprehensiveness of the inspection.

Description

Technical Field

[0001] The invention relates to the technical field of abnormal monitoring of protective nets, and in particular to an abnormal monitoring system and a monitoring method for a polyester fiber protective net. Background Art

[0002] Polyester fiber protective net usually refers to a protective net made of polyester fiber. Polyester fiber is a synthetic fiber with the characteristics of wear resistance, tensile resistance, and corrosion resistance. It is often used to make various protective equipment.

[0003] Traditional systems usually rely on manual inspections, which is inefficient and prone to omissions, especially in large-scale and difficult inspection tasks; and traditional systems usually lack real-time monitoring and automatic early warning capabilities. Once an abnormality occurs, manual discovery and reporting may be required, resulting in a slow response; and traditional systems may only be able to monitor at limited monitoring points and cannot cover the entire polyester fiber protection net, resulting in blind spots and missed inspections; and traditional systems require a lot of manpower and material resources for inspection and maintenance, with high long-term costs, and human factors may lead to inconsistency and reliability issues in inspection results. Summary of the invention

[0004] The purpose of the present invention is to solve the problems existing in the background technology and to provide a polyester fiber protective net abnormality monitoring system and monitoring method.

[0005] The technical solution of the present invention is: a polyester fiber protection net abnormality monitoring system, including an abnormality map construction unit, an abnormality sorting unit, an inspection sorting unit, an abnormality inspection unit, a video detection unit and an abnormality early warning unit, and also includes:

[0006] An image acquisition unit, wherein the image acquisition unit is used to set a plurality of image acquisition nodes on the polyester fiber protection net in the target area, set an image acquisition device at the image acquisition node, and use the image acquisition device to acquire images of the polyester fiber protection net segments around the image acquisition nodes to obtain a plurality of polyester fiber protection net segment images, and transmit the plurality of polyester fiber protection net segment images to an image enhancement unit;

[0007] An image enhancement unit, which receives the multiple polyester fiber protection network segment images transmitted by the image acquisition unit, and performs image enhancement on the multiple polyester fiber protection network segment images by an image enhancement method to obtain multiple polyester fiber protection network segment enhanced images, and transmits the multiple polyester fiber protection network segment enhanced images to the image detection unit;

[0008] An image detection unit receives the multiple polyester fiber protection network segment enhanced images transmitted by the image enhancement unit, and performs image detection on the multiple polyester fiber protection network segment enhanced images through a trained image detection model to obtain detection results corresponding to the polyester fiber protection network segment enhanced images, that is, the detection results corresponding to the polyester fiber protection network segments. If the detection results are abnormal, the image acquisition nodes corresponding to the abnormal polyester fiber protection network segments are transmitted to the abnormal graph construction unit.

[0009] Preferably, the abnormal graph construction unit receives the image acquisition node corresponding to the abnormal polyester fiber protection network segment transmitted by the image detection unit, and constructs an abnormal weighted undirected graph based on the image acquisition node corresponding to the abnormal polyester fiber protection network segment, the abnormal weighted undirected graph includes abnormal nodes, connecting edges and connecting point weights, the abnormal nodes correspond to the image acquisition nodes, the connecting edges are used to connect two image acquisition nodes based on the polyester fiber protection network segment between the two image acquisition nodes, the connecting point weight corresponds to the length of the polyester fiber protection network segment between the two image acquisition nodes, and the abnormal weighted undirected graph is transmitted to the abnormal sorting unit.

[0010] Preferably, the abnormal sorting unit receives the abnormal weighted undirected graph transmitted by the abnormal graph construction unit, and calculates the total length of the connection lines between the abnormal nodes based on the abnormal weighted undirected graph, constructs an abnormal sorting objective function based on the total length of the connection lines between the abnormal nodes, and uses a penalty function to solve the abnormal sorting objective function to minimize the total length of the connection lines between the abnormal nodes, thereby obtaining an abnormal node sorting sequence, and transmitting the abnormal node sorting sequence to the inspection sorting unit.

[0011] Preferably, the inspection sorting unit receives the abnormal node sorting sequence transmitted by the abnormal sorting unit, and generates a UAV inspection task sorting sequence based on the abnormal node sorting sequence, the UAV inspection task includes inspecting and photographing the image acquisition node corresponding to the abnormal node and inspecting and photographing the adjacent image acquisition nodes, and the UAV inspection task sorting sequence is transmitted to the abnormal inspection unit.

[0012] Preferably, the abnormal inspection unit receives the UAV inspection task sorting sequence transmitted by the inspection sorting unit, and obtains a first UAV inspection route based on the UAV inspection task sorting sequence, optimizes the first UAV inspection route through an ant algorithm to obtain a second UAV inspection route, performs a UAV inspection task based on the second UAV inspection route to obtain a polyester fiber protection network segment inspection video, and transmits the polyester fiber protection network segment inspection video to the video detection unit.

[0013] Preferably, the video detection unit receives the polyester fiber protection net segment inspection video transmitted by the abnormal inspection unit, and performs abnormality detection on the polyester fiber protection net segment inspection video through a trained video abnormality detection model to obtain the abnormality type corresponding to the polyester fiber protection net segment, and transmits the abnormality type corresponding to the polyester fiber protection net segment to the abnormal warning unit.

[0014] Preferably, the abnormal warning unit receives the abnormal type corresponding to the polyester fiber protection net segment transmitted by the video detection unit, and matches the abnormal type corresponding to the polyester fiber protection net segment with a preset abnormal type table to obtain the abnormal level corresponding to the polyester fiber protection net segment, and sends corresponding warning information to the abnormal monitoring terminal based on the abnormal level corresponding to the polyester fiber protection net segment.

[0015] Preferably, the image enhancement method comprises the following steps:

[0016] A1. Process the polyester fiber protection network segment image using a gamma unsharp mask algorithm to obtain a first polyester fiber protection network segment enhanced image, and calculate a first weight of the first polyester fiber protection network segment enhanced image. The first weight calculation formula is as follows:

[0017]

[0018] in, represents the first weight of the first polyester fiber protection network segment enhanced image, Indicates indexing the exported enhanced image of the first polyester fiber protection network segment. represents the average pixel value of the enhanced image of the first polyester fiber protection network segment, σ G represents the standard deviation of the pixel values ​​of the enhanced image of the first polyester fiber protection network segment;

[0019] A2. Process the polyester fiber protection network segment image by multi-scale Retinex and color restoration algorithm to obtain a second polyester fiber protection network segment enhanced image, and calculate a second weight of the second polyester fiber protection network segment enhanced image. The second weight calculation formula is as follows:

[0020]

[0021] in, represents the second weight of the second polyester fiber protection network segment enhanced image, Indicates indexing the exported enhanced image of the second polyester fiber protection network segment. represents the average pixel value of the enhanced image of the second polyester fiber protection network segment, σ MRepresents the standard deviation of the pixel values ​​of the enhanced image of the second polyester fiber protection net segment.

[0022] Preferably, the image enhancement method further comprises the following steps:

[0023] A3. Based on the first weight, the first polyester fiber protection network segment enhanced image is multiplied pixel by pixel through a Gaussian pyramid and a Laplacian pyramid to obtain a first output image. The first output image is expressed as follows:

[0024]

[0025] in, represents the first output image, G l represents the lth layer of the Gaussian pyramid, L l represents the lth level of the Laplace pyramid, represents the first weight after normalization;

[0026] A4. Based on the second weight, the second polyester fiber protection network segment enhanced image is multiplied pixel by pixel through a Gaussian pyramid and a Laplacian pyramid to obtain a second output image. The second output image is expressed as follows:

[0027]

[0028] in, represents the second output image, represents the normalized second weight;

[0029] A5. Reconstruct the first output image and the second output image to obtain a final polyester fiber protection network segment enhanced image, wherein the final polyester fiber protection network segment enhanced image is expressed as follows:

[0030]

[0031] in, Represents the second output image.

[0032] Preferably, the video anomaly detection model includes an encoder and a decoder, the encoder consists of 3 spatial channel attention modules and 1 basic block, the spatial channel attention module sequentially includes a 3×3 convolution layer, a batch normalization layer, a ReLU activation function, a spatial channel convolution layer, a batch normalization layer and a ReLU activation function, the basic block consists of 2 3×3 convolution layers, a batch normalization layer and a ReLU activation function, and the decoder consists of a deconvolution layer with a convolution kernel size of 3×3, a batch normalization layer and a ReLU activation function.

[0033] Preferably, the total length of the connection between abnormal nodes is calculated as follows:

[0034]

[0035] Where L(Q) represents the total length of the connection between abnormal nodes, W(i,j) represents the total weight, d[Q(i),Q(j)] represents the distance between abnormal nodes, and d[Q(i),Q(j)]=|x i -x j |+|y i -y j |,(x i ,y i ) represents the horizontal and vertical coordinate values ​​of abnormal node i, (x j ,y j ) represents the horizontal and vertical coordinate values ​​of the j abnormal node, and n represents the total number of abnormal nodes;

[0036] The anomaly sorting objective function is as follows:

[0037]

[0038] Among them, minL represents the abnormal sorting objective function.

[0039] Preferably, optimizing the first UAV inspection route by using an ant algorithm to obtain a second UAV inspection route comprises the following steps:

[0040] B1. Initialize the initial pheromone concentration, the maximum number of iterations, the random distribution of birth points for M ants, the number of nodes, the pheromone concentration increment, the total amount of pheromone, and substitute the coordinates of the target abnormal node of the first drone inspection route into the allowed set;

[0041] B2. Access the allowed set and search for the next abnormal node based on a probability formula, where the probability formula is as follows:

[0042]

[0043] in, represents the probability of searching for the next abnormal node, It represents the amount of pheromone accumulated from node x to node y under the pheromone influence factor. represents the expected degree of transfer from node x to y under the influence factor of the heuristic function, α represents the pheromone influence factor, and β represents the heuristic function influence factor;

[0044] B3, updating the allowed set based on the next abnormal node searched, and putting the searched abnormal node into the exploration set;

[0045] B4, determine whether all ants have traversed all abnormal nodes, that is, whether the number of ants that have visited all abnormal nodes is less than the total number M. If so, repeat step B2, otherwise, execute step B5;

[0046] B5. Update the pheromone concentration of the route of the abnormal node combination in the exploration set. The pheromone concentration update formula is as follows:

[0047]

[0048] in, represents the amount of pheromone accumulated from node x to node y, ρ represents a random number in the interval (0,1), The ant that found the shortest route releases pheromones;

[0049] B5. Determine whether the current number of iterations is less than the maximum number of iterations. If so, repeat step B2. Otherwise, output the optimal route of the abnormal node combination in the exploration set. The optimal route is the second UAV inspection route.

[0050] The technical solution of the present invention is a method for monitoring abnormalities of a polyester fiber protection net, which is applicable to the above-mentioned abnormality monitoring system of a polyester fiber protection net, and comprises the following steps:

[0051] S1. A plurality of image acquisition nodes are set on the polyester fiber protection net in the target area, and image acquisition devices are set on the image acquisition nodes. The image acquisition devices are used to acquire images of the polyester fiber protection net segments around the image acquisition nodes to obtain a plurality of polyester fiber protection net segment images.

[0052] S2. Performing image enhancement on a plurality of polyester fiber protection network segment images by an image enhancement method to obtain a plurality of polyester fiber protection network segment enhanced images, and performing image detection on the plurality of polyester fiber protection network segment enhanced images by a trained image detection model to obtain detection results corresponding to the polyester fiber protection network segment enhanced images;

[0053] S3, constructing an abnormal weighted undirected graph based on the image acquisition nodes corresponding to the polyester fiber protection network segments with abnormalities;

[0054] S4. Calculate the total length of the connection between abnormal nodes based on the abnormal weighted undirected graph, construct an abnormal sorting objective function based on the total length of the connection between abnormal nodes, and use a penalty function to solve the abnormal sorting objective function so that the total length of the connection between abnormal nodes is minimized, thereby obtaining an abnormal node sorting sequence;

[0055] S5. Generate a drone inspection route according to the inspection task sorting sequence, execute the inspection task, and obtain the inspection video of the polyester fiber protection network segment;

[0056] S6. Perform anomaly detection on the polyester fiber protection network segment inspection video using the trained video anomaly detection model to obtain the anomaly type corresponding to the polyester fiber protection network segment.

[0057] Compared with the prior art, the above technical solution of the present invention has the following beneficial technical effects:

[0058] 1. The present invention monitors various parts of the polyester fiber protective net in real time, and identifies abnormal situations through image acquisition, enhancement and detection. Once an abnormality is detected, the system can immediately issue an early warning, which helps to take timely measures to prevent further damage or accidents. It also generates a drone inspection task sequence by sorting abnormal nodes to achieve rapid inspection of abnormal areas. During the inspection process, the ant algorithm is also used to optimize the route to ensure the efficiency and comprehensiveness of the inspection.

[0059] 2. The present invention uses an image detection unit and a video detection unit, so the system can detect anomalies on the polyester fiber protection net at multiple levels. Image detection is used for anomaly recognition of static images, and video detection is used for anomaly recognition of dynamic video streams, which improves the accuracy and coverage of anomaly detection. The trained model is used for anomaly detection and task optimization, which greatly reduces the need for manual intervention and improves the automation and intelligence level of monitoring. In addition, by generating anomaly graphs and sorting anomalies, it helps managers to perform data analysis and optimize preventive maintenance strategies, thereby improving the overall management efficiency and safety of the polyester fiber protection net. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A schematic diagram of the overall system flow in an embodiment of the present invention;

[0061] Figure 2 The figure is a flow chart of an overall method in one embodiment of the present invention.

[0062] Figure numerals: 1. Image acquisition unit; 2. Image enhancement unit; 3. Image detection unit; 4. Abnormal map construction unit; 5. Abnormal sorting unit; 6. Inspection sorting unit; 7. Abnormal inspection unit; 8. Video detection unit; 9. Abnormal warning unit. DETAILED DESCRIPTION

[0063] Embodiment 1, as Figure 1 As shown, a polyester fiber protection net abnormality monitoring system proposed by the present invention includes an abnormality map construction unit 4, an abnormality sorting unit 5, an inspection sorting unit 6, an abnormality inspection unit 7, a video detection unit 8 and an abnormality early warning unit 9, and also includes:

[0064] The image acquisition unit 1 is used to set a plurality of image acquisition nodes on the polyester fiber protection net in the target area, set an image acquisition device at the image acquisition node, and use the image acquisition device to acquire images of the polyester fiber protection net segments around the image acquisition nodes to obtain a plurality of polyester fiber protection net segment images, and transmit the plurality of polyester fiber protection net segment images to the image enhancement unit 2;

[0065] An image enhancement unit 2 receives the multiple polyester fiber protection network segment images transmitted by the image acquisition unit 1, and performs image enhancement on the multiple polyester fiber protection network segment images by an image enhancement method to obtain multiple polyester fiber protection network segment enhanced images, and transmits the multiple polyester fiber protection network segment enhanced images to the image detection unit 3;

[0066] The image detection unit 3 receives the multiple polyester fiber protection network segment enhanced images transmitted by the image enhancement unit 2, and performs image detection on the multiple polyester fiber protection network segment enhanced images through the trained image detection model to obtain the detection results corresponding to the polyester fiber protection network segment enhanced images, that is, the detection results corresponding to the polyester fiber protection network segments. If there is an abnormality in the detection result, the image acquisition node corresponding to the abnormal polyester fiber protection network segment is transmitted to the abnormal map construction unit 4.

[0067] In the present invention, the image acquisition device generally refers to a specific hardware device used to capture images. For the image acquisition of polyester fiber protective nets, common image acquisition devices may be high-definition cameras or surveillance cameras for specific purposes; image detection models include but are not limited to FasterR-CNN, YOLO series models, SSD and MaskR-CNN.

[0068] In an optional embodiment, the abnormal graph construction unit 4 receives the image acquisition node corresponding to the abnormal polyester fiber protection network segment transmitted by the image detection unit 3, and constructs an abnormal weighted undirected graph based on the image acquisition node corresponding to the abnormal polyester fiber protection network segment. The abnormal weighted undirected graph includes abnormal nodes, connecting edges and connecting point weights. The abnormal nodes correspond to the image acquisition nodes. The connecting edges are used to connect the two image acquisition nodes based on the polyester fiber protection network segment between the two image acquisition nodes. The connecting point weight corresponds to the length of the polyester fiber protection network segment between the two image acquisition nodes. The abnormal weighted undirected graph is transmitted to the abnormal sorting unit 5.

[0069] In an optional embodiment, the abnormal sorting unit 5 receives the abnormal weighted undirected graph transmitted by the abnormal graph construction unit 4, and calculates the total length of the connection lines between the abnormal nodes based on the abnormal weighted undirected graph, constructs the abnormal sorting objective function based on the total length of the connection lines between the abnormal nodes, and uses the penalty function to solve the abnormal sorting objective function to minimize the total length of the connection lines between the abnormal nodes, thereby obtaining the abnormal node sorting sequence, and transmitting the abnormal node sorting sequence to the inspection sorting unit 6.

[0070] It should be noted that the penalty function usually constrains the search space of the solution by penalizing or weighting certain parts of the objective function, thereby optimizing the overall result of the objective function.

[0071] In an optional embodiment, the inspection sorting unit 6 receives the abnormal node sorting sequence transmitted by the abnormal sorting unit 5, and generates a drone inspection task sorting sequence based on the abnormal node sorting sequence. The drone inspection task includes inspecting and photographing the image acquisition nodes corresponding to the abnormal nodes and inspecting and photographing the adjacent image acquisition nodes, and transmitting the drone inspection task sorting sequence to the abnormal inspection unit 7.

[0072] In an optional embodiment, the abnormal inspection unit 7 receives the UAV inspection task sorting sequence transmitted by the inspection sorting unit 6, and obtains a first UAV inspection route based on the UAV inspection task sorting sequence, optimizes the first UAV inspection route through the ant algorithm to obtain a second UAV inspection route, performs a UAV inspection task based on the second UAV inspection route to obtain a polyester fiber protection network segment inspection video, and transmits the polyester fiber protection network segment inspection video to the video detection unit 8.

[0073] In an optional embodiment, the video detection unit 8 receives the polyester fiber protection network segment inspection video transmitted by the abnormal inspection unit 7, and performs abnormality detection on the polyester fiber protection network segment inspection video through a trained video abnormality detection model to obtain the abnormality type corresponding to the polyester fiber protection network segment, and transmits the abnormality type corresponding to the polyester fiber protection network segment to the abnormal warning unit 9.

[0074] In an optional embodiment, the abnormal warning unit 9 receives the abnormal type corresponding to the polyester fiber protection network segment transmitted by the video detection unit 8, and matches the abnormal type corresponding to the polyester fiber protection network segment with a preset abnormal type table to obtain the abnormal level corresponding to the polyester fiber protection network segment, and sends corresponding warning information to the abnormal monitoring terminal based on the abnormal level corresponding to the polyester fiber protection network segment.

[0075] It should be noted that the abnormality monitoring terminal is usually a device or system specially designed to receive, process and display abnormal information.

[0076] Embodiment 2, a polyester fiber protective net abnormality monitoring system proposed by the present invention, compared with embodiment 1, this embodiment also includes: an image enhancement method, comprising the following steps:

[0077] A1. Process the polyester fiber protection network segment image using a gamma unsharp mask algorithm to obtain a first polyester fiber protection network segment enhanced image, and calculate a first weight of the first polyester fiber protection network segment enhanced image. The first weight calculation formula is as follows:

[0078]

[0079] in, A first weight representing a first polyester fiber protection network segment enhanced image; Indicates indexing the exported enhanced image of the first polyester fiber protection network segment. represents the average pixel value of the enhanced image of the first polyester fiber protection network segment, σ G represents the standard deviation of the pixel values ​​of the enhanced image of the first polyester fiber protection network segment;

[0080] A2. The polyester fiber protection network segment image is processed by multi-scale Retinex and color restoration algorithm to obtain a second polyester fiber protection network segment enhanced image, and a second weight of the second polyester fiber protection network segment enhanced image is calculated. The second weight calculation formula is as follows:

[0081]

[0082] in, represents the second weight of the second polyester fiber protection network segment enhanced image, Indicates indexing the exported enhanced image of the second polyester fiber protection network segment. represents the average pixel value of the enhanced image of the second polyester fiber protection network segment, σ M Represents the standard deviation of the pixel values ​​of the enhanced image of the second polyester fiber protection net segment.

[0083] In this embodiment, the gamma unsharp mask algorithm is an image enhancement technology used to enhance the details of the image, especially the edge part, and enhance the overall visual effect of the image by adjusting the gamma value of the image and applying the unsharp mask; the multi-scale Retinex and color restoration algorithm aims to enhance the contrast and color balance of the image by simulating the working principle of the human visual system.

[0084] In an optional embodiment, the image enhancement method further includes the following steps:

[0085] A3. Based on the first weight, the first polyester fiber protection network segment enhanced image is multiplied pixel by pixel through a Gaussian pyramid and a Laplacian pyramid to obtain a first output image. The first output image is expressed as follows:

[0086]

[0087] in, represents the first output image, G l represents the lth layer of the Gaussian pyramid; L l represents the lth level of the Laplace pyramid, represents the first weight after normalization;

[0088] A4. Based on the second weight, the second polyester fiber protection network segment enhanced image is multiplied pixel by pixel through a Gaussian pyramid and a Laplacian pyramid to obtain a second output image. The expression of the second output image is as follows:

[0089]

[0090] in, represents the second output image, represents the normalized second weight;

[0091] A5. Reconstruct the first output image and the second output image to obtain a final polyester fiber protection network segment enhanced image. The final polyester fiber protection network segment enhanced image expression is as follows:

[0092]

[0093] in, Represents the second output image.

[0094] It should be noted that Gaussian pyramid and Laplacian pyramid are two common types of image pyramids. Gaussian pyramid is a pyramid constructed by reducing the image resolution layer by layer, while Laplacian pyramid is constructed by reconstructing a low-resolution image from a high-resolution image.

[0095] In an optional embodiment, the video anomaly detection model includes an encoder and a decoder, the encoder consists of 3 spatial channel attention modules and 1 basic block, the spatial channel attention module sequentially includes a 3×3 convolution layer, a batch normalization layer, a ReLU activation function, a spatial channel convolution layer, a batch normalization layer and a ReLU activation function, the basic block consists of 2 3×3 convolution layers, a batch normalization layer and a ReLU activation function, and the decoder consists of a deconvolution layer with a convolution kernel size of 3×3, a batch normalization layer and a ReLU activation function.

[0096] In an optional embodiment, the total length of the connection between abnormal nodes is calculated as follows:

[0097]

[0098] Where L(Q) represents the total length of the connection between abnormal nodes, W(i,j) represents the total weight, d[Q(i),Q(j)] represents the distance between abnormal nodes, and d[Q(i),Q(j)]=|x i -x j |+|y i -y j |,(x i ,y i ) represents the horizontal and vertical coordinate values ​​of abnormal node i, (x j ,y j ) represents the horizontal and vertical coordinate values ​​of the j abnormal node, and n represents the total number of abnormal nodes;

[0099] The exception sorting objective function is as follows:

[0100]

[0101] Among them, minL represents the abnormal sorting objective function.

[0102] In an optional embodiment, optimizing the first UAV inspection route by using an ant algorithm to obtain a second UAV inspection route includes the following steps:

[0103] B1. Initialize the initial pheromone concentration, the maximum number of iterations, the random distribution of birth points for M ants, the number of nodes, the pheromone concentration increment, the total amount of pheromone, and substitute the coordinates of the target abnormal node of the first drone inspection route into the allowed set;

[0104] B2. Access the allowed set and search for the next abnormal node based on the probability formula. The probability formula is as follows:

[0105]

[0106] in, represents the probability of searching for the next abnormal node, It represents the amount of pheromone accumulated from node x to node y under the pheromone influence factor. represents the expected degree of transfer from node x to y under the influence factor of the heuristic function, α represents the pheromone influence factor, and β represents the heuristic function influence factor;

[0107] B3. Update the allowed set based on the next abnormal node searched, and put the searched abnormal node into the exploration set;

[0108] B4, determine whether all ants have traversed all abnormal nodes, that is, whether the number of ants that have visited all abnormal nodes is less than the total number M. If so, repeat step B2, otherwise, execute step B5;

[0109] B5. Update the pheromone concentration of the route of the abnormal node combination in the exploration set. The pheromone concentration update formula is as follows:

[0110]

[0111] in, represents the amount of pheromone accumulated from node x to node y, ρ represents a random number in the interval 0,1, The ant that found the shortest route releases pheromones;

[0112] B5. Determine whether the current number of iterations is less than the maximum number of iterations. If so, repeat step B2. Otherwise, output the optimal route of the abnormal node combination in the exploration set. The optimal route is the second UAV inspection route.

[0113] Embodiment three, as Figure 2 As shown, a polyester fiber protection net abnormality monitoring method proposed by the present invention is applicable to the polyester fiber protection net abnormality monitoring system described above, and comprises the following steps:

[0114] S1. A plurality of image acquisition nodes are set on the polyester fiber protection net in the target area, and image acquisition devices are set on the image acquisition nodes. The image acquisition devices are used to acquire images of the polyester fiber protection net segments around the image acquisition nodes to obtain a plurality of polyester fiber protection net segment images.

[0115] S2. Performing image enhancement on a plurality of polyester fiber protection network segment images by an image enhancement method to obtain a plurality of polyester fiber protection network segment enhanced images, and performing image detection on the plurality of polyester fiber protection network segment enhanced images by a trained image detection model to obtain detection results corresponding to the polyester fiber protection network segment enhanced images;

[0116] S3, constructing an abnormal weighted undirected graph based on the image acquisition nodes corresponding to the polyester fiber protection network segments with abnormalities;

[0117] S4. Calculate the total length of the connection between abnormal nodes based on the abnormal weighted undirected graph, construct an abnormal sorting objective function based on the total length of the connection between abnormal nodes, and use a penalty function to solve the abnormal sorting objective function so that the total length of the connection between abnormal nodes is minimized, thereby obtaining an abnormal node sorting sequence;

[0118] S5. Generate a drone inspection route according to the inspection task sorting sequence, execute the inspection task, and obtain the inspection video of the polyester fiber protection network segment;

[0119] S6. Perform anomaly detection on the polyester fiber protection network segment inspection video using the trained video anomaly detection model to obtain the anomaly type corresponding to the polyester fiber protection network segment.

[0120] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto, and various changes can be made within the knowledge scope of technicians in the relevant technical field without departing from the purpose of the present invention.

Claims

1. A polyester fiber protective net abnormality monitoring system, characterized in that: include: An image acquisition unit (1), the image acquisition unit (1) being used to set a plurality of image acquisition nodes on the polyester fiber protection net in the target area, to set an image acquisition device on the image acquisition node, to perform image acquisition on the polyester fiber protection net segment around the image acquisition node by the image acquisition device, to obtain a plurality of polyester fiber protection net segment images, and to transmit the plurality of polyester fiber protection net segment images to an image enhancement unit (2); An image enhancement unit (2), the image enhancement unit (2) receiving the plurality of polyester fiber protection network segment images transmitted by the image acquisition unit (1), and performing image enhancement on the plurality of polyester fiber protection network segment images by an image enhancement method to obtain a plurality of polyester fiber protection network segment enhanced images, and transmitting the plurality of polyester fiber protection network segment enhanced images to the image detection unit (3); An image detection unit (3), the image detection unit (3) receiving the plurality of polyester fiber protection network segment enhanced images transmitted by the image enhancement unit (2), and performing image detection on the plurality of polyester fiber protection network segment enhanced images using a trained image detection model to obtain detection results corresponding to the polyester fiber protection network segment enhanced images, that is, detection results corresponding to the polyester fiber protection network segments; if the detection results are abnormal, the image acquisition node corresponding to the polyester fiber protection network segment with the abnormality is transmitted to the abnormality map construction unit (4); The abnormal graph construction unit (4) receives the image acquisition node corresponding to the abnormal polyester fiber protection network segment transmitted by the image detection unit (3), and constructs an abnormal weighted undirected graph based on the image acquisition node corresponding to the abnormal polyester fiber protection network segment, wherein the abnormal weighted undirected graph includes an abnormal node, a connection edge and a connection point weight, wherein the abnormal node corresponds to the image acquisition node, the connection edge is used to connect the two image acquisition nodes based on the polyester fiber protection network segment between the two image acquisition nodes, and the connection point weight corresponds to the length of the polyester fiber protection network segment between the two image acquisition nodes, and the abnormal weighted undirected graph is transmitted to the abnormal sorting unit (5); The abnormal sorting unit (5) receives the abnormal weighted undirected graph transmitted by the abnormal graph construction unit (4), calculates the total length of the connection lines between the abnormal nodes based on the abnormal weighted undirected graph, constructs an abnormal sorting objective function based on the total length of the connection lines between the abnormal nodes, solves the abnormal sorting objective function using a penalty function to minimize the total length of the connection lines between the abnormal nodes, thereby obtaining an abnormal node sorting sequence, and transmits the abnormal node sorting sequence to the inspection sorting unit (6); The inspection sorting unit (6) receives the abnormal node sorting sequence transmitted by the abnormal sorting unit (5), and generates a drone inspection task sorting sequence based on the abnormal node sorting sequence, wherein the drone inspection task includes inspecting and photographing the image acquisition node corresponding to the abnormal node and inspecting and photographing the adjacent image acquisition nodes, and transmits the drone inspection task sorting sequence to the abnormal inspection unit (7); The abnormal inspection unit (7) receives the drone inspection task sorting sequence transmitted by the inspection sorting unit (6), obtains a first drone inspection route based on the drone inspection task sorting sequence, optimizes the first drone inspection route by an ant algorithm to obtain a second drone inspection route, performs a drone inspection task based on the second drone inspection route to obtain a polyester fiber protection network segment inspection video, and transmits the polyester fiber protection network segment inspection video to a video detection unit (8); The video detection unit (8) receives the inspection video of the polyester fiber protection network segment transmitted by the abnormal inspection unit (7), and performs abnormality detection on the inspection video of the polyester fiber protection network segment using a trained video abnormality detection model to obtain the abnormality type corresponding to the polyester fiber protection network segment, and transmits the abnormality type corresponding to the polyester fiber protection network segment to the abnormality warning unit (9); The abnormal warning unit (9) receives the abnormal type corresponding to the polyester fiber protection net segment transmitted by the video detection unit (8), and matches the abnormal type corresponding to the polyester fiber protection net segment with a preset abnormal type table to obtain the abnormal level corresponding to the polyester fiber protection net segment, and sends corresponding warning information to the abnormal monitoring terminal based on the abnormal level corresponding to the polyester fiber protection net segment.

2. The polyester fiber protection net abnormality monitoring system according to claim 1 is characterized in that: The video anomaly detection model includes an encoder and a decoder. The encoder consists of three spatial channel attention modules and one basic block. The spatial channel attention module sequentially includes a 3×3 convolution layer, a batch normalization layer, a ReLU activation function, a spatial channel convolution layer, a batch normalization layer and a ReLU activation function. The basic block consists of two 3×3 convolution layers, a batch normalization layer and a ReLU activation function. The decoder consists of a deconvolution layer with a convolution kernel size of 3×3, a batch normalization layer and a ReLU activation function.

3. The polyester fiber protection net abnormality monitoring system according to claim 1, characterized in that: The calculation formula for the total length of the connection between abnormal nodes is as follows: Where L(Q) represents the total length of the connection between abnormal nodes, W(i,j) represents the total weight, d[Q(i),Q(j)] represents the distance between abnormal nodes, and d[Q(i),Q(j)]=|x i -x j |+|y i -y j |,(x i ,y i ) represents the horizontal and vertical coordinate values ​​of abnormal node i, (x j ,y j ) represents the horizontal and vertical coordinate values ​​of the j abnormal node, and n represents the total number of abnormal nodes; The anomaly sorting objective function is as follows: Among them, minL represents the abnormal sorting objective function.

4. The polyester fiber protection net abnormality monitoring system according to claim 1, characterized in that: Optimizing the first UAV inspection route by using an ant algorithm to obtain a second UAV inspection route includes the following steps: B1. Initialize the initial pheromone concentration, the maximum number of iterations, the random distribution of birth points for M ants, the number of nodes, the pheromone concentration increment, the total amount of pheromone, and substitute the coordinates of the target abnormal node of the first drone inspection route into the allowed set; B2. Access the allowed set and search for the next abnormal node based on a probability formula, where the probability formula is as follows: in, represents the probability of searching for the next abnormal node, It represents the amount of pheromone accumulated from node x to node y under the pheromone influence factor. represents the expected degree of transfer from node x to y under the influence factor of the heuristic function, α represents the pheromone influence factor, and β represents the heuristic function influence factor; B3, updating the allowed set based on the next abnormal node searched, and putting the searched abnormal node into the exploration set; B4, determine whether all ants have traversed all abnormal nodes, that is, whether the number of ants that have visited all abnormal nodes is less than the total number M. If so, repeat step B2, otherwise, execute step B5; B5. Update the pheromone concentration of the route of the abnormal node combination in the exploration set. The pheromone concentration update formula is as follows: in, represents the amount of pheromone accumulated from node x to node y, ρ represents a random number in the interval (0,1), The ant that found the shortest route releases pheromones; B5. Determine whether the current number of iterations is less than the maximum number of iterations. If so, repeat step B2. Otherwise, output the optimal route of the abnormal node combination in the exploration set. The optimal route is the second UAV inspection route.

5. A polyester fiber protection net abnormality monitoring method, which is applicable to a polyester fiber protection net abnormality monitoring system according to any one of claims 1 to 4, characterized in that: The following steps are involved: S1. A plurality of image acquisition nodes are set on the polyester fiber protection net in the target area, and image acquisition devices are set on the image acquisition nodes. The image acquisition devices are used to acquire images of the polyester fiber protection net segments around the image acquisition nodes to obtain a plurality of polyester fiber protection net segment images. S2. Performing image enhancement on a plurality of polyester fiber protection network segment images by an image enhancement method to obtain a plurality of polyester fiber protection network segment enhanced images, and performing image detection on the plurality of polyester fiber protection network segment enhanced images by a trained image detection model to obtain detection results corresponding to the polyester fiber protection network segment enhanced images; S3, constructing an abnormal weighted undirected graph based on the image acquisition nodes corresponding to the polyester fiber protection network segments with abnormalities; S4. Calculate the total length of the connection between abnormal nodes based on the abnormal weighted undirected graph, construct an abnormal sorting objective function based on the total length of the connection between abnormal nodes, and use a penalty function to solve the abnormal sorting objective function so that the total length of the connection between abnormal nodes is minimized, thereby obtaining an abnormal node sorting sequence; S5. Generate a drone inspection route according to the inspection task sorting sequence, execute the inspection task, and obtain the inspection video of the polyester fiber protection network segment; S6. Perform anomaly detection on the polyester fiber protection network segment inspection video using the trained video anomaly detection model to obtain the anomaly type corresponding to the polyester fiber protection network segment.

Images