Development Of On-line Image Monitoring System For Flame And Oil Leakage Based On CMOS Camera

With the development of society,the manufacturing industry is gradually transforming to the intelligent direction.The traditional model of after-the-fact maintenance and planned maintenance is increasingly difficult to meet the needs of cost and efficiency.At present,manual inspection is the main monitoring strategy for mixers in ironworks.However,the timeliness and accuracy of manual inspection are difficult to guarantee,and it cannot meet the high efficiency,high speed and high precision inspection requirements pursued by modern industry.Therefore,this paper has developed an on-line image monitoring system for flame and machine oil leakage faults at the working site of the ironmaking plant mixer,thereby improving the speed of fault detection.The research object of this article is flame and leaked engine oil.The flame is distinguished from background interference by extracting the static and dynamic characteristics of the flame,so as to determine whether a flame exists at the monitoring site,and the image processing algorithm is used to monitor the machine for oil leakage faults.The corresponding image processing algorithms are studied,and software and hardware systems are developed.The main research contents of this article are as follows:(1)Set up an image acquisition system,select an industrial camera suitable for field applications,and design a camera layout plan for the work site.A simple image acquisition system was built using the laboratory's existing industrial cameras and other equipment,and a camera control and data transmission system was developed through MTALAB(2)Use the built image acquisition system to collect images,and then perform noise reduction processing on the images through median filtering and morphological opening operations.Binary segmentation of the flame image by combining the RGB+HSI color model with a custom threshold method;(3)Analyze the static and dynamic characteristics of the flame,first extract the static characteristics of the flame,and remove some regular non-flame interference areas in the background through the circularity and rectangularity thresholds.The dynamic characteristics of the flame are extracted through the continuous multi-frame flame images collected by the image acquisition system,and the background interference is further removed according to the characteristics of the flame sharp angle,the area change rate,and the flicker frequency,so as to determine whether there is a flame at the monitoring site,and to detect the presence of flame in the monitoring site Extract the flame area and position information when it exists;(4)Monitor the oil leakage of the machine,and divide the cement floor and the metal test bench through image masks.The cement floor monitoring area uses the 2D-Otsu method to equalize the S channel in the HSI color model after histogram equalization.At the same time,the texture local entropy image is segmented by the 1D-Otsu method,and the above two methods are combined to accurately extract the oil leakage area on the ground without causing misjudgment.For the monitoring area of the metal experimental platform,the location of the oil spill area is first determined by combining the local entropy of the image texture and the maximum orthogonal entropy,and then the area of the oil spill area is extracted by combining the local entropy of the texture and the two-dimensional maximum fuzzy entropy.(5)Based on MATLAB 2017a software for algorithm writing and system development,a complete software system is finally formed,and the modular interface is convenient for users to use.Functions such as image collection,processing,and alarm can be realized through oneclick operation.At the same time,various step-by-step processing algorithms have been designed to facilitate users to call.