Research Of The Algorithm Of The On-line Intelligent Impurity Detector

The system used for checking and picking material on line is a rising automaticchoosing equipment. It is widely used in grading and machining food, vegetable,fruit or other farm produce, and also used in examining inferior in industrialproduction. For example, it can be used in monitoring the process of automaticchecking and machining in industry, checking the quality of byproducts, analyzingand diagnosing medical images and so on. Compared to the primal manual method,the rising automatic choosing equipment reduces the intensity of the workers' labor,increases the efficiency of the work, and also improves the precision of the grading andchecking of the products.The on-line intelligent impurity detector, which is together developed by AnhuiUniversity, Qinhuangdao machinery company in Shandong, and Guoyang tobaccofactory in Anhui, is a high-tech product involving machine vision, computertechnology, sensor technology and aerodynamics. It is the local key project of theNational "Ninth Five-Year-Plan" of department of science and technology, and hasachieved the international advanced level. This system has been completed andtentatively used in 2003. Presently, the on-line intelligent impurity detector is mainlyapplied in the industry of tobacco. Having been used in Guoyang tobacco factory formany years, it is proved that the product is of steady capability, easily handled, andhas the function of weeding many kinds of impurity.The purpose of this paper is to perfect the "on-line intelligent impurity detector".By using in tobacco industry for several years, the deficiencies of the systemgradually emerge. In this paper, we study the implementation of the system'ssoftware, include collecting signal by the computer, the pretreatment of the image,and recognizing the signal. By analyzing and concluding the current problems, fouraspects will be discussed as follows:1. Studying the models of different signals, include the models of tobacco leaf, background, impurity and noise. The emphases of this part is to abstract thecharacteristic of the color of certain tobacco leaf in order to find a betterrecognition algorithm.2. The research of the algorithm of adjusting luminance. Because of theposition and the characteristic of the light, there are two types of luminancedistortion in the system. In this part, a distortion model of asymmetricluminance is proposed, and these two types of distortion are adjusted.3. Studying the algorithm of optimizing the signal's threshold of tobacco leaf.First of all, the characteristic space of the signal is optimized by theoperation of protection and filter. Secondly, the non-tobacco region iscompressed by using KL transform.4. Designing the unit of communication. In order to manipulate the systemconveniently, firstly we design a reasonable order, then establish a smallLAN to monitor, control and recognize the material on the transmissionbelt.