Vision Based Infrared Gas Leak Detection Algorithm Research

With the development of the petrochemical industry,gas leakage in chemical production has gradually increased.Most of these gases are volatile organic compound's which are explosive and invisible to the naked eyes.It's not easy to be detected in time when leaked.Once an accident occurs,It will cause irreparable losses to chemical companies,seriously threaten to people's lives and property.In view of the shortcomings of traditional detection methods such as small detection range and low efficiency.A method is proposed in this thesis that can realize automatic detection in a larger range.The gas infrared imaging technology is used by this method,introducing visible light video on the basis of infrared video to make up for the visible light factor that has been neglected in the past infrared gas detection.Utilizing this type of gas is only visible under infrared condition,on the basis of heterogeneous image registration,the movement interference coexisting under infrared and visible light was preliminarily eliminated to obtain a more accurate suspected leakage area.Finally the gas characteristic analysis is performed on the suspected leakage area to determine whether there is a gas leakage.Compared with the traditional method,the method in this thesis can carry out a full range of gas leakage monitoring without human intervention,and has significant advantages such as large range,high efficiency and strong safety.The main works of this thesis are as follows:(1)In order to achieve the mapping between the same target in the infrared and visible light images,and to be used in the subsequent visual difference processing,the registration of heterogeneous images must be performed.Since the current heterogeneous image registration algorithm is difficult to achieve uniformity in accuracy and real-time,a heterogeneous image registration algorithm is proposed in this thesis that uses a combination of image closed contour regions and mutual information.In order to achieve contour area extraction,the completed edge information of the image needs to be obtained.An infrared image edge extraction algorithm based on infrared image features and K-means is proposed in this thesis.(2)To analyze the characteristics of gas infrared imaging.firstly,it uses the background subtraction method to extract the motion area to eliminate some interference similar to the gas characteristics,and then using the infrared and visible image registration results to differentiate the motion areas between images.Coexisting motion interference is eliminated,and the suspected gas leakage area is finally obtained.(3)The gas characteristics are summarized and analyzed under infrared conditions,proposing a multi-feature fusion gas leak detection method to analyze and quantify the gas irregular characteristics,shape characteristics,diffusion characteristics,floating characteristics and grayscale characteristics,It can accurately distinguish the leakage gas and the interference area,adapt to the complex working environment and have stronger robustness.Experiments and analysis are completed on the proposed algorithm in this thesis.The experimental results show that the gas leakage area and leakage trend can be located and grasped within specified time in the algorithm of this thesis,which provides an effective solution for volatile organic gas leakage detection.