CN-115830494-B - Model training method, intelligent spraying method and system for mining sprinkler

CN115830494BCN 115830494 BCN115830494 BCN 115830494BCN-115830494-B

Abstract

The invention relates to the field of monitoring of watering carts of strip mines, and the data acquisition is carried out to obtain monitoring video data with fire coal and dust; the fire coal and dust video data are processed into picture data of fire coal and dust; the invention provides a training method of a fire coal and dust recognition model and a dispatching method of an optimal sprinkling path, wherein the training method of the dust recognition model is used for obtaining a model with the fire coal and dust recognition capability, at least solving the problem of low manual recognition efficiency, and meanwhile, the sprinkling path can be reasonably planned through the recognized fire coal and dust position information.

Inventors

ZHANG FEI
HAO BIN
ZHANG XINYI
WANG JINMING
REN XIAOYING
LIU PILIANG
LI JINZHU

Assignees

内蒙古科技大学
扎鲁特旗扎哈淖尔煤业有限公司

Dates

Publication Date: 20260505
Application Date: 20221122

Claims (8)

1. The intelligent spraying method of the mining sprinkler is characterized by comprising the following steps of: Step S1, collecting real-time video data; Step S2, transmitting the fire coal and the dust to DEEPSTREAM for processing in a wireless mode, deploying a trained fire coal and dust identification model on DEEPSTREAM, and identifying and detecting the fire coal and the dust; the fire coal and dust recognition model is trained by the following method, and specifically comprises the following steps: Step one, data acquisition, namely acquiring video data with fire coal and dust monitoring; step two, processing fire coal and dust video data into picture data of fire coal and dust; Marking picture data of fire coal and dust to form a data set; Building YoloV a model, importing the data set processed in the third step, training to obtain a model with fire coal and dust recognition capability; Step S3, positioning the final fire coal dust position through the laser holder, specifically, adopting angle and acceleration information of a high-precision gyroscope to drive a motor so that the directions of a camera and a laser transmitter carried on the laser holder and a locking target are kept unchanged in an inertial space, and realizing target locking and automatic tracking, thereby locking and tracking the fire coal and dust positions in real time; and S4, planning an optimal path for the sprinkler to travel, and spraying the sprinkler to a designated position through the optimal path.
2. The intelligent spraying method of the mining sprinkler is characterized in that labelme software is used for enabling dust-raising picture data, fire coal and dust-raising places to be formed by rectangular frames and giving labels, and marked images are converted into txt formats, and the fire coal and dust-raising picture data and the labels correspond to each other to form a data set.
3. The intelligent spraying method for the mining sprinkler according to claim 2, wherein the data set is further processed according to a training set, namely a test set, namely 7:2:1, by using a python automation script, and the processed picture and the label thereof are stored in a server.
4. The intelligent spraying method of the mining sprinkler according to claim 1, wherein the parameter adjustment is carried out for a plurality of times in the training process of the step four.
5. The intelligent spraying method of the mining sprinkler according to claim 1, wherein the step S4 is specifically that the water level, the flow rate and the position in the sprinkler need to be monitored before the sprinkler starts.
6. The intelligent spraying method of the mining sprinkler according to claim 1, wherein the optimal path planning method is as follows: abstracting a road in a mine into a network model, abstracting a path into edges in the network model, and converting the path distance into weight values of the edges; abstracting the dispatch center and each spray point into N nodes, The weighted directed graph with N nodes is represented by a weighted adjacency matrix Cost, the weight value of an arc < Vi, vj > is represented by a Cost [ i, j ], and if Vi is not communicated with V, the Cost [ i, j ] is enabled to be in an infinite state; Then introducing a vector Dist, wherein Dist [ i ] refers to the minimum path weight from the starting point to the key point Vi; setting a vector with a directed graph number of m, setting the vector value as follows: where V is the set of network nodes; The end point from which the shortest path has been found starting from the start point Vm is classified as set S, where the initial value of S is s= { V m }, then: (1) Vj is selected such that: Wherein Vj is the focal point of the shortest path that has been found from Vm, then let: ; (2) Modify the shortest path length from vertex V m to any vertex V k in set V-S, if any, dist [ j ] +Cost [ j, K ] < Dist [ K ] (3) Modifying Disk to be Dist [ K ] =Dist [ j ] +cost [ j, K ] Repeating the steps (2) and (3) for N-1 times, thereby obtaining the shortest path from the starting point scheduling center V m to each vertex in the directed graph.
7. The intelligent spraying system of the mining sprinkler disclosed by the claim 1 is characterized by comprising high-precision cameras, an image operation server, a laser holder and a sprinkler system, wherein the high-precision cameras are arranged on two sides of an open-air mine road and are used for collecting real-time video data, the high-precision cameras are in communication connection with the image operation server, the image operation server is used for receiving the real-time video data and processing the real-time video data, identifying fire coal and dust information, transmitting signals to the laser holder, the laser holder is arranged in the open-air mine and used for locking the fire coal and dust positions after receiving signals of the image operation server to obtain position information, and the sprinkler system is used for sprinkling and dust suppression after being dispatched.
8. The intelligent sprinkler system for the mining sprinkler of claim 7, wherein the sprinkler system comprises a sprinkler and measuring hardware arranged on the sprinkler, the measuring hardware comprises millimeter wave radars and a flowmeter, the millimeter wave radars are arranged in a water tank of the sprinkler and used for measuring the water level in the sprinkler, and the flowmeter is arranged in front of a water valve of the sprinkler and used for measuring the sprinkling flow of the sprinkler.

Description

Model training method, intelligent spraying method and system for mining sprinkler Technical Field The invention relates to the field of supervision of mine watering carts of strip mines, in particular to a model training method, an intelligent spraying method and an intelligent spraying system of mine watering carts. Background Along with the continuous expansion of coal mine industry, heavy dump truck transportation is adopted in large open-pit mining areas at present, the transportation road is generally an unstructured soil road, the road surface flatness is poor, road dust can not be avoided, surrounding environment is polluted, the body health of workers is damaged, the service life of the vehicle is shortened, in addition, dust can influence the sight of a vehicle driver, the driving safety is influenced, the production efficiency is reduced, a free face is formed after coal mining, the free face is fully contacted with air, the ignition point is partially self-ignited, loss of coal resources and ground collapse are caused, a large amount of harmful gases such as carbon monoxide and sulfur dioxide are discharged into the air, and great pollution is generated to the environment, so that water sprinkling dust removal treatment is required to fire coal and dust in the mining areas. The conventional dust removing method for the fire coal and the dust of the strip mine is to continuously spray water on the road surface through a water spraying vehicle, and the method can achieve a certain dust removing effect, but continuously spray water, waste water resources, and has the defects of complex road condition of the strip mine and reasonable planning of the path, so that how to quickly and accurately acquire the road surface information becomes a key place of the road maintenance work. Disclosure of Invention The invention provides a training method of a fire coal and dust emission recognition model and a scheduling method of an optimal sprinkling path, wherein the training method of the dust emission recognition model is used for obtaining a model with fire coal and dust emission recognition capability, at least solving the problem of low manual recognition efficiency, and simultaneously reasonably planning the sprinkling path through recognized fire coal and dust emission position information. The technical scheme adopted by the invention is as follows, a training method of a fire coal and dust recognition model comprises the following steps: Step one, data acquisition, namely acquiring video data with fire coal and dust monitoring; step two, processing fire coal and dust video data into picture data of fire coal and dust; Marking picture data of fire coal and dust to form a data set; And fourthly, constructing YoloV models, importing the data sets processed in the third step, and training to obtain the models with fire coal and dust recognition capability. And a third specific mode is that labelme software is used for making fire coal and dust places in the dust picture data, rectangular frames are used for giving labels, the marked images are converted into txt format, and the fire coal and dust picture data and the labels thereof correspond to form a data set. Further, the data set is further subjected to the following operation, the python automation script is used, the data set is verified to be 7:2:1 according to the training set, and the processed picture is stored in the server in the label. Specifically, the parameter adjustment is carried out for a plurality of times in the training process in the step four. The application also provides an intelligent spraying method of the mining sprinkler, which comprises the following steps of Step S1, collecting real-time video data; Step S2, transmitting the fire coal and the dust to DEEPSTREAM for processing in a wireless mode, disposing a trained fire coal and dust identification model in the embodiment I on DEEPSTREAM, and identifying and detecting the fire coal and the dust; S3, positioning a final fire coal dust position through a laser holder; and S4, enabling the sprinkler to reach a designated position for sprinkling. The step S4 specifically includes monitoring the water level, flow and position in the sprinkler before the sprinkler starts. Preferably, in the step S4, an optimal path for the sprinkler to travel is planned, and the sprinkler sprays through the optimal path Specifically, the planning method of the optimal path is as follows: abstracting a road in a mine into a network model, abstracting a path into edges in the network model, and converting the path distance into weight values of the edges; abstracting the dispatch center and each spray point into N nodes, The weighted directed graph with N nodes is represented by a weighted adjacency matrix Cost, the weights of arcs < Vi, vj > are represented by Cost [ i, j ], and if Vi to V are not connected, then Cost [ i, j ] = infinity is made. Then, vector Dist is introduced, dist [ i ] refers