A method for identifying ethylene gas leakage based on multi-feature joint detection and risk analysis

In view of the fact that the leakage of ethylene gas in the production process of petrochemical enterprises can easily cause major safety accidents, infrared video real-time imaging is used to monitor possible ethylene gas leakage points, and ethylene gas is achieved through real-time processing and analysis of acquired infrared video images. Real-time detection of leakage, a multi-feature joint detection of ethylene gas leakage, using the loss function to carry out risk analysis of the test results, and ultimately to determine whether there is ethylene gas leakage determination of ethylene gas leakage identification method. Firstly, by establishing a global energy function model, the interrelationships between intra-frame and inter-frame neighboring pixels in the video sequence and the spatial domain information are used, and by adding a neighborhood constraint, an integrated spatio-temporal consistency feature set, etc., a leak suspecting segmentation is achieved. Then, an ethylene gas leak detection and identification method based on multi-feature joint detection was proposed by analyzing and detecting the characteristics of area variation in suspected areas of gas leakage, wavelet domain energy variation characteristics, and transient gray value “nine-square grid” distribution characteristics. Finally, through the further determination of the leakage state based on the loss function and risk analysis, the misjudgment and missed-judgement conditions are eliminated as much as possible, and the accuracy of the recognition is improved. The experimental results show that the ethylene gas leakage detection and identification method in this paper has the characteristics of good real-time performance and high recognition rate, which can meet the real-time warning requirements.