Research On Gas Leakage Location Technology Based On Image Mosaic

With rapid development of industrial technologies,requirements of gas tightness detection are becoming more stringent for people.It is difficult to realize automatic operation due to low efficiency and large error for traditional gas tightness detection methods.The leak detection method of infrared technology has gradually attracted attentions of enterprises and scientific research units with its convenient,fast,easy to detect,real-time characteristics.However,with development of application on infrared technology and improvement of requirement on gas leakage detection requirements,there are many defects in infrared technology detection.For example,infrared thermal imager is susceptible to interference from surrounding environment and produces noise.Moreover,it is difficult to get infrared images of a larger measured object with small photographing angle.In response to above problems,a set of gas leak detection system is designed in this research project.Temperature field of container is collected real-time by an infrared camera,then a panoramic image of the container is obtained by multiple images with different shooting angles for the container based on image stitching technology.A leak detection algorithm is designed to achieve purpose of detecting and locating for leak point.In this paper,an infrared gas leak detection experimental platform is built firstly,and infrared images of the workpiece at moment of charging and discharging gas are collected by FLIR A20 m infrared camera.Based on principles of existing image stitching,an improved SIFT feature point detection method is proposed to improve speed of image stitching based on FAST detection principle and PCA dimension reduction method.Feature points of extracted images to be stitched are matched by BBF search method based on KD tree.And transformation matrix parameters between two images to be stitched are obtained by removing mismatched points based on RANSAC algorithm.To solve problems on image mosaic cracks,an image fusion technique is proposed based on wavelet transform.An improved bilateral filtering algorithm is proposed to achieve locate of leakage points accurately,which combines methods of least squares support vectormachine and point source diffusion.By experimental verification,the improved SIFT feature point detection method and BBF search method significantly improve the speed of image stitching.Image fusion based on wavelet transform can remove splice cracks better.Therefore,the proposed leak detection algorithm has certain reference significance and theoretical value for realizing real-time location of industrial gas leak detection.