The Research On The Key Algorithm In Coke Oven Engine Intelligent Monitoring System

In the existing coking automatic control system, the heavy dust, the electromagnetic interference, the vibration of the vehicles and other site factors, lead the current used positioning technologies, such as inductive wireless technology, infrared laser beam technology, and rotary encoder technology and so on, can’t achieve the desired results. The development of coking industry has always been influenced by the low-efficiency and unreliability of artificial operation, along with the harsh site environment, the natural climate and other factors. As the machine vision technology and the embedded technology become more and more mature, under the market driven of the unmanned operation of coking vehicles, research on the key algorithms in the coke oven engine intelligent monitoring system makes a noticeable difference in reality.This paper, under the background of the application of the embedded machine vision technology in automatic coking system, invested the spot and analyzed the specific needs of the coke oven engine intelligent monitoring system, and collected a large number of images and videos as the experimental samples, and then started the study on the algorithms of the identification of the oven number and the detection of the moving body before the coke oven’s door. In the oven number identification algorithm, this paper firstly located the oven number by using the Sobel edge detection with two more direction templates, and selected the average of the four gradient magnitude to be the new gray value of this pixel; secondly, binarized the oven number image after getting the accurate edge information, and achieved the single character segmentation of the oven number by vertical projection method, and meanwhile proposed a calibration method according to the characteristics of the image to improve the reliability of the algorithm; finally, designed a BP neural network classifier for the character recognition, and trained and tested the network with a large number of spot images in normal, dusty, smoky, light-various and other scenarios, and achieved the specified error standard to0.01. In the algorithm of the detection of the moving body before the coke oven’s door, this paper, inspired by the idea of block modeling, divided the image into blocks unevenly, and established Gaussian mixture models for every block of the image, and combined frame differencing with background differencing, used the average of the differencing result of the three consecutive frames to assess changes in light and other environmental factors, which made up the background differencing method’s shortcoming of sensitivity to the light changes, and at the same time proposed a effective background model update method with the background memory parameter. From the simulation experiment results, this improved algorithm effectively inhibited the impact of environmental mutation, and reduced the risk that the moving target integrates into the background when it remains still for a long time, and improved the result of the moving body detection.The research of this paper has solved the two key problems in the coke oven engine intelligent monitoring system, and by monitoring the coke oven engine and the working condition, made up for the existing coke oven engine positioning system, and ensured the safety of coke production, and also made a good foundation for the unmanned operation of the coke oven engine.