CN-121982814-A - Portable monitoring and early warning method and system for underground buried cable

CN121982814ACN 121982814 ACN121982814 ACN 121982814ACN-121982814-A

Abstract

A portable monitoring and early warning method and system for underground buried cables belong to the technical field of railway construction, and mainly comprise the steps of collecting three-dimensional point cloud data of excavating equipment and surrounding environment by utilizing a laser radar, collecting two-dimensional image data of the excavating equipment and the surrounding environment by utilizing a monocular camera, extracting a two-dimensional detection frame of a bucket of the excavating equipment from the two-dimensional image, extracting a three-dimensional bounding box of a real-time space position of the bucket from the three-dimensional point cloud, associating the three-dimensional bounding box with the two-dimensional detection frame through point cloud and image fusion, obtaining real-time position information of the bucket of the excavating equipment, comparing the real-time position information with an existing model of the underground cables which is imported in advance, and sending out an alarm signal. The method and the system can realize monitoring and early warning of underground cable construction, effectively inhibit cable damage accidents caused by construction misoperation, and reduce construction accidents.

Inventors

ZHANG ZHIPENG
WU HAOTIAN
WU HUIJUN
LIU YAQIANG
LIU YUNJIE
WU HUIJIE
SHAO XUEJUN
GUO YOUWEI
PANG QING
WANG MINGLEI
LI ZHIXUAN
ZHOU BAOLONG

Assignees

中国铁道科学研究院集团有限公司标准计量研究所
中国铁道科学研究院集团有限公司

Dates

Publication Date: 20260505
Application Date: 20251125

Claims (8)

1. The portable monitoring and early warning method for the underground buried cable is characterized by comprising the following steps of: The method comprises the steps of S1, collecting three-dimensional point cloud data of excavating equipment and surrounding environment by using a laser radar, collecting two-dimensional image data of the excavating equipment and the surrounding environment by using a monocular camera, performing space synchronization on pixel points in a two-dimensional image and point clouds in the three-dimensional laser radar by joint calibration; S2, constructing a detection model based on an image convolutional neural network, extracting a two-dimensional detection frame of the bucket of the excavating equipment from a two-dimensional image, constructing an intelligent detection algorithm based on three-dimensional point cloud, extracting real-time spatial position information of the bucket and point cloud characteristic information, establishing a connection between the three-dimensional bounding box and the two-dimensional detection frame by constructing the point cloud and the image fusion frame, and outputting real-time position information of the bucket of the excavating equipment; and S3, comparing the real-time position information of the bucket of the excavating equipment with a pre-imported existing model of the underground cable, and sending out an alarm signal to realize monitoring and early warning of underground cable construction.
2. The method according to claim 1, wherein the step S1 specifically comprises: S11, the laser radar finishes scanning for a plurality of times every second, and a complete three-dimensional point cloud model covering the whole construction area is formed by accumulating and splicing point cloud data of continuous multiframes; the monocular camera converts an optical signal of a construction area into an electric signal by means of an image sensor of the monocular camera, synchronously acquires two-dimensional image information of the area, and provides a visual basis for subsequent data fusion; S12, after the position of the system equipment moves each time, carrying out joint calibration on the laser radar and the monocular camera before monitoring, wherein the joint calibration comprises camera internal parameter calibration and radar-camera external parameter calibration, and the specific steps are as follows: Collecting a plurality of images from different angles for a checkerboard calibration plate by using Zhang Zhengyou calibration method, extracting corner pixel coordinates of a checkerboard in each image, and calculating camera internal parameters by using the coordinate information, wherein the camera internal parameters comprise radial distortion, focal length and principal point coordinates; the same checkerboard calibration plate is placed in a common view field of a laser radar and a monocular camera, multiple groups of point clouds and image data are collected from different angles, angular point coordinates of corresponding checkerboards are respectively extracted, and a radar-camera rotation matrix and a translation vector external parameter are obtained through joint calculation; S13, registering, namely calibrating an internal reference matrix based on a monocular camera External parameter matrix for joint calibration of laser radar and monocular camera Converting the point cloud data into corresponding pixel data, and setting the point cloud coordinates as The pixel coordinates are The conversion of the point cloud coordinates to pixel coordinates is achieved using the following: In the formula, As a rotation matrix in the extrinsic matrix, As a translation vector in the extrinsic matrix, Is a scale factor.
3. The method according to claim 1, wherein in the step S2, the specific method for extracting the two-dimensional detection frame of the excavating equipment bucket from the two-dimensional image includes: forming a dataset from the image data of the bucket of the excavating equipment acquired in the earlier stage through external labeling, and training by adopting a Yolov single-stage target detection model from end to end based on the dataset; Analyzing image data acquired by a monocular camera in real time as input of a detection model, and identifying a boundary frame of a bucket of the excavating equipment; the image-based Yolov detection model comprises an input end, an output end, a Neck network and a backbone network, wherein: The method comprises the steps of adding adaptive zooming of pictures at an input end and enhancing by using Mosaic data, splicing input images in a random arrangement, zooming and clipping mode, enhancing the practicability of a small target data set and improving the robustness of a network; the loss function is added at the output end, so that the prediction precision is improved; The CSPNet structure and Foucs result are used for improving the backbone network, preventing gradient explosion and reducing information loss; According to the Yolov detection model, a two-dimensional bucket detection frame is obtained, the detection frame is equal to the size of the intersection ratio IOU of the real frame G and the prediction frame C, and the calculation formula is as follows: In the formula, The area of the prediction frame C is indicated, The area of the real frame G is shown.
4. The method according to claim 1, wherein in the step S2, the specific method for constructing the intelligent detection algorithm based on the three-dimensional point cloud, and extracting the real-time spatial position information and the point cloud feature information of the bucket includes: performing cyclic plane fitting on the seed points selected in the small areas according to the point cloud height by adopting an area plane fitting algorithm, and when the distance from the points to the plane is smaller than a certain threshold value, representing that the ground is formed, and dividing the complete ground after the plurality of areas are fitted at the same time; clustering the bucket by using a clustering algorithm, presetting a neighborhood radius Eps and a minimum point MinPts, and marking all points as 'unaccessed'; Traversing the point cloud, namely randomly selecting one non-access point, calculating the number of adjacent points in the radius of the Eps, marking the point as a core point and creating a new cluster if the number exceeds MinPts, and recursively incorporating all the adjacent points with reachable densities into the cluster, including other core points and the neighborhood thereof, marking the point as a boundary point of the cluster if the number of the adjacent points of the point is less than MinPts but is positioned in the neighborhood of a certain core point, otherwise, treating the point as a noise point, and continuously iterating until all the points are accessed, and finally outputting the point cloud cluster and the noise point; Selecting the minimum point in the bucket point cloud cluster And maximum point Constructing a three-dimensional bounding box, wherein the calculation formula of the length, width and height of the bounding box is as follows: 。
5. The method according to claim 1, wherein in the step S2, the specific method for establishing a relationship between the three-dimensional bounding box and the two-dimensional detection frame by establishing a point cloud and an image fusion frame includes: According to the calibrated external parameter conversion matrix, converting the bucket three-dimensional bounding box into a two-dimensional detection frame In order to increase the query speed, a KD-Tree is constructed to search the two-dimensional detection frame of the nearest neighbor after conversion, and the two-dimensional detection frame is regarded as Calculation of And (3) with Is judged by the preset superposition rate: If it is greater than the threshold value The three-dimensional bounding box and the two-dimensional detection frame are considered to point to the same target, a label of the target is added on the three-dimensional bounding box, and the barycenter coordinate of the three-dimensional bounding box is real-time position information of the bucket; If it is less than the threshold value And considering that the three-dimensional bounding box and the two-dimensional detection frame point to different targets, and adding other labels on the three-dimensional bounding box.
6. The method according to claim 1, wherein the step S3 specifically includes: S31, acquiring a three-dimensional bounding box of the bucket label, and acquiring the minimum coordinates of the three-dimensional bounding box I.e. the lowest position of the current bucket tip, with each point in the pre-introduced existing three-dimensional model of the underground cable Calculating Euclidean distance The calculation formula is as follows: ; S32, will Comparing with the set alarm threshold value, judging whether the alarm condition is reached: (1) Setting an alarm threshold according to cable attributes, construction equipment types and geological environment factors ; (2) For all calculated results d and Different levels of early warning response are set in proportion to: If it is Indicating that the excavation construction progress is in a low risk state, sending a normal signal, if Indicating that the excavation construction progress is in a warning state and sending out a warning signal if The excavation construction progress is in a dangerous state, and an alarm signal is sent out; wherein, the alarm threshold value Is defined by a base threshold value And risk factors The dynamic generation is specifically as follows: the risk factors are obtained by weighting functions under the combined action of cable attributes, construction equipment types and geological environments: Wherein, the Is a cable attribute factor; is a construction equipment type factor; Is a soil geological environment factor.
7. An underground buried cable portable monitoring and warning system applying the method of any one of claims 1-6, comprising a portable lidar, a monocular camera, a processing computing device, an alarm device, and other devices, wherein: The laser radar is used for acquiring three-dimensional point cloud data of the excavating equipment and the surrounding environment, and the monocular camera is used for acquiring two-dimensional image data of the excavating equipment and the surrounding environment; The processing computing device is used for carrying out space synchronization on pixel points in the two-dimensional image and point clouds in the three-dimensional laser radar, realizing data fusion of the two-dimensional image and the three-dimensional point clouds under the same coordinate system by utilizing a registration algorithm, extracting a two-dimensional detection frame of the bucket of the excavating equipment from the two-dimensional image by utilizing a constructed detection model based on an image convolution neural network, extracting real-time spatial position information and point cloud characteristic information of the bucket by utilizing a constructed intelligent detection algorithm based on the three-dimensional point clouds, establishing a connection between the three-dimensional bounding box and the two-dimensional detection frame by utilizing the point clouds and the image fusion frame, outputting real-time position information of the bucket of the excavating equipment, and carrying out position comparison on the real-time position information of the bucket of the excavating equipment and a pre-imported underground cable existing model to send an early warning or alarm signal to the alarm device; the alarm device is used for receiving the early warning and alarm signals and making sound and lamplight responses; other devices are used for providing power supply and supporting structures for the system and meeting the portability and stability requirements of on-site monitoring.
8. The system of claim 7, wherein the alarm device comprises a buzzer and an alarm lamp, both of which respond to an alarm signal at a set frequency, wherein green is normally on and has no beeping, is a normal signal, yellow is low-frequency flashing and low-frequency beeping is an early warning signal, and red is high-frequency flashing and high-frequency beeping is an alarm signal.

Description

Portable monitoring and early warning method and system for underground buried cable Technical Field The invention relates to the technical field of railway construction, in particular to a portable monitoring and early warning method and system for an underground buried cable. Background Underground cable systems along railways are complex in construction and cover a wide variety of cables for signals, power, communications, and the like. The method is influenced by long paving time, complex professional category and repeated construction superposition, and the deviation of the actual position information of the underground cable is large, so that great challenges are brought to subsequent maintenance, overhaul and energy expansion work. In the projects such as line extension, station reconstruction and the like, because the cable distribution is difficult to comprehensively master, effective preventive measures and real-time monitoring and protecting equipment are lacked, the fine management cannot be realized, the cables are extremely easy to damage, the economic loss is caused, the construction safety risk is increased, and particularly, when the ground construction of the existing cable area is carried out, the cables are easy to dig or destroy, the railway operation safety is endangered, and adverse social influence is caused. At present, an electromagnetic induction method is mostly adopted for underground cable detection, and a metal pipeline is positioned through electromagnetic field change, but the static detection method cannot be carried out synchronously with construction, and monitoring, protection and early warning during construction are difficult to realize. Although the conventional video monitoring can monitor the excavating equipment in real time, the conventional video monitoring only depends on a single image sensor, lacks of three-dimensional space positioning capability, cannot accurately judge the excavating depth and the distance between the bucket and the cable, is easily influenced by illumination and weather, and greatly reduces the detection performance. Disclosure of Invention The invention provides a portable underground cable monitoring and early warning method and system, which can realize monitoring and early warning of the existing underground cable area during the construction of the railway along the ground, avoid the risk of cable damage and ensure the quality safety of railway engineering. The method mainly comprises the following steps: Receiving three-dimensional point cloud data of mining equipment and surrounding environments acquired by a laser radar, mining equipment and two-dimensional image data of surrounding environments acquired by a monocular camera, performing spatial synchronization on pixel points in a two-dimensional image and point clouds in the three-dimensional laser radar through joint calibration, and realizing data fusion of the two-dimensional image and the three-dimensional point clouds under the same coordinate system by using a registration algorithm; Constructing a detection model based on an image convolution neural network to acquire two-dimensional position information and image characteristic information of a bucket, acquiring real-time spatial position information and point cloud characteristic information of the bucket of the excavating equipment by utilizing an intelligent detection algorithm based on three-dimensional point cloud, establishing a connection between a three-dimensional bounding box and a two-dimensional bounding box by constructing a point cloud and an image fusion frame, and outputting real-time position information of the bucket of the excavating equipment; And comparing the position of the underground cable with the position of the pre-imported existing model of the underground cable, if the space distance between the tip of the bucket of the excavating equipment and the cable reaches an early warning threshold value, sending out an early warning signal, and if the space distance does not reach the threshold value, sending out an alarm signal, and if the space distance does not reach the threshold value, not sending out a signal, thereby realizing monitoring and early warning of underground cable construction. The underground buried cable portable monitoring and early warning system applying the method mainly comprises: portable lidar, monocular camera, processor, alarm device, and other devices, wherein: the laser radar obtains three-dimensional point cloud data of the excavating equipment and the surrounding environment by emitting laser beams and reflecting the laser beams through the surface of the object, and the monocular camera acquires two-dimensional image data of the excavating equipment and the surrounding environment and transmits the data to the processor; The processor performs space synchronization on pixel points in the two-dimensional image and point clouds in the three-dimensional laser radar, rea