A Rapid Detection System Of Rock And Soil Moisture Based On Near-Infrared Image

Gas drilling is a new type of drilling technology.Compared with traditional drilling technology,it is more efficient and can effectively protect oil and gas reservoirs.However,encountering formation water in the process of gas drilling can cause complicated accidents such as sticking and borehole collapse.Therefore,it is necessary to provide timely early warning and take corresponding measures to solve the problem of water effusion from the formation.In the gas drilling,detecting the moisture of the rock and soil returned from the tunneling process can effectively determine whether there is water in the formation.At present,at the gas drilling site,the rock and soil samples are usually collected manually at the outlet of the pipeline,and the moisture of the rock and soil is obtained by the drying and weighing method.This method is time-consuming and labor-intensive,and the detection speed is slow.In order to realize the early warning of the formation of water in the gas drilling environment,this paper designs a near-infrared image-based rock and soil moisture rapid detection system,which is composed of a data acquisition device,a hardware part and a host computer.The detection process as follows: firstly,create a near-infrared parallel light shooting environment through a data acquisition device,and then use the embedded device Raspberry Pi to control the near-infrared camera to collect near-infrared images of rock and soil under a fixed band of 850 nm,and then base on the near-infrared image of rock and soil,through Huffman compresses and send it to the host computer in real time.Among them,the remote communication function between the host computer and the hardware is realized by the TCP protocol.In the host computer,the speckle noise of the gray image of the rock and soil is removed by the filtering algorithm based on wavelet transform,and then the impurity interference in the rock and soil is removed by the OTSU algorithm.On this basis,calculate and normalize the gray-gradient co-occurrence matrix characteristics of the near-infrared rock and soil image,and the support vector machine algorithm is further used to determine the rock and soil category.At the same time,the DBSCAN clustering algorithm is used to extract the rock and soil pixel points and calculate the gray center value,and finally use the piecewise linear regression model to predict the soil moisture value.This paper compares the filtering effects of mean filtering,median filtering,and wavelet transform filtering algorithms applied in geotechnical near-infrared images.Aiming at the problem of interference in the rock and soil near-infrared image,the segmentation algorithm based on the Gaussian mixture model,the ordinary threshold segmentation algorithm and the OTSU algorithm applied to the rock and soil near-infrared image are compared.In the problem of rock and soil classification,the texture and gray characteristics of the rock and soil region in the near-infrared rock and soil image are extracted through the gray-gradient co-occurrence matrix.And inputting it into some common classifiers to predict,then calculating and comparing the classification error.Designing a piecewise linear regression algorithm to divide high humidity and low humidity to meet the measurement requirements of low humidity in rock and soil.This paper verifies the effect and repeatability of the system through the designed data acquisition device,and analyzes the error of the experimental results,especially the detection accuracy of the system under low humidity,which proves the effectiveness of the system designed in this paper.In addition,this article also conducted an anti-interference experiment through the data acquisition device,which proved that the rapid detection method of soil moisture proposed in this article can deal with the influence of interference and environmental light changes,and get more accuracy under low soil moisture.It shows a fast response time for the formation water,which can effectively achieve the early warning effect,and has a certain reference value.