Research On The Method Of Detecting Leakage Of Floor Heating Pipes Based On Infrared Imaging Technology

Due to damage to floor heating lines,mechanical aging,changes in the surrounding environment,chemical erosion and human damage,the direct buried heating lines commonly used in citizens’ homes today are at risk of leakage,which is likely to cause great waste of resources and economic loss.The current means of detecting leaks in China is often through temperature measurement or pressure,which largely relies on traditional experience,making the accuracy and efficiency of detection affected.Therefore,the research on the automation method of water leakage of floor heating pipes has both important application value.In this thesis,based on the thermal phenomenon of pipe leakage,the principle of thermal diffusion is studied,and a set of floor heating pipe leakage detection system based on infrared image technology is developed and established.The pre-processing algorithm of infrared image and the model of leakage detection on infrared image are carefully studied,which can realize the accurate,rapid and real-time positioning of pipe leakage points and have good theoretical significance and practicality.The main results and conclusions of this thesis are as follows.(1)The thermal phenomena and the principle of thermal diffusion when the pipeline leaks are fully analyzed.On this basis,the system scheme and software processing flow of pipeline leakage detection are designed,and the hardware selection and construction of the infrared image data acquisition system are completed.(2)The pre-processing of the data collected by the floor heating pipe leakage sensor is completed by using a combination of differential calculation,median filtering,Gaussian filtering,and histogram equalization pre-processing method.The thermal diffusion characteristics of the images are well preserved under this processing,which is conducive to the rapid detection of pipe leakage points and greatly reduces the computational effort of the subsequent algorithms compared with the unprocessed ones.(3)Three detection algorithms for infrared images of pipeline leaks are investigated,a machine learning SVM model based on traditional image processing,a YOLO V3 model based on deep learning,and an improved YOLO V3 model.Among them,the improved YOLO V3 model has good performance in detecting small targets by optimizing the network structure and adding convolutional layers,and finally achieves the effect of strong real-time detection and good generality of floor heating pipe leakage points.Finally,the system experiment platform was built to verify the algorithm of the floor heating pipe leak detection system,and the experimental results showed that the improved network model could detect the leaks well,with an accuracy of 96.45%,a recall of 90.23%,and an average accuracy of 0.882.The improved loss function and suitable anchor frame play a great role in fitting the data set and speeding up the convergence.