Research Of Dynamic Information Processing And Recognition System Of Entry Coal Locomotive In Fuel Electric Plant

With the high-speed development of computer technology in recent years, digital image processing has been widely used in various industries. Dynamic information and automatic recognition system of the loading coal locomotive in fuel electric plant is an intelligent system based on image processing and character recognition. This system is applied to measure the load of locomotive, which can help operators register locomotive no. and improve quality of service.According to the characteristic of locomotive no., this work is based on the technology of digital image processing to recognize the dynamic no. of locomotive automatically, meanwhile it presents the design idea and scenario of the system, completes the system debugging and test.Dynamic information and recognition system of the loading coal locomotive is achieved by following working process:1) Image extraction of moving object. In this stage, the position and point of view of the CCD camera, setting of the sensors and the use of supplemental lighting facilities are researched, which are foundations for later recognition.2) Pretreatment for the images to be recognized. Gray scale transform, binarization and edge detection are researched in this stage. To reduce the noise caused by part binarization method, P-parameter nonlinear gray scale transform principle utilizing gray scale stretch to improve imagine quality combined with the structure image binarization arithmetic of SOFM net is adopt.3) Two times localization method is adopted to accurately localize locomotive no., at the same time interpenetration count characteristic and horizontal-vertical characteristic are used for refined segmentation and extraction. It can be effective to search for the upper and lower edge of the localized locomotive no.4) Intelligent character recognition, utilizing BP NN-network to recognize the extracted character. The main recognition process consists of three parts: First, confirming the effectivity of the character to be recognized; second, extracting the rude gridding characteristic of unified character and imagine; after that selecting NN-net classifier. It simulates the recognition character behavior of human being, according to the different states and circumstances of the object to be recognized, it adjusts the parameters of inference engine, adopts different recognition strategy and chooses different feed-back structures. Meanwhile the system gives attention to the demand of diversified recognition capabilities and targets.This system has been applied in fuel management system of Shenyang heat and power plant, more than one year's practice has proved, the designed system is feasible with high accuracy rate of recognition and high speed.