Research And Implementation Of Stationery Inspection System Based On Machine Vision

Automatic detection based on machine vision is a new detection method, which combines the application of image processing and analysis, pattern recognition, artificial intelligence, precision instrument manufacturing, and other techniques. Automatic detection based on machine vision, as a kind of modern detection method, has drawn increasing attention because of its high-speed, accuracy, non-contact and so on.As a major manufacturing country, stationery production and export have occupied a considerable share in China. At present, the production process of stationery has been achieved the automation basically, but the way of testing stationery quality is still traditional manual detecting. With its increasing productivity, the traditional methods haven't been able to meet the need of production. To deal with this problem, realize the on-line testing of stationery quality and save the whole process time, this paper makes some researches as follows:1. The paper describes the working principle of automatic detection system based on machine vision, integral structure and the flow work, analyzes a variety of hardware features and selecting appropriate hardware, such as light source, image cards and industrial cameras.then,set up a decection system's hardware platform.2. After the deep analysis of stationery features, stationery was classified by a improved histogram using a combination of two characteristics of area and ratio(area/perimeter); the critical area that distinguish the same stationery was found using the difference accumulator ,and the color features was extracted.3. The paper introduces several commonly used neural network model, the basic structure, and analyze the learning rules of neural networks. In-depth study of the multi-layer feedforward network with back-propagation learning algorithm of network structure and training algorithm of the concrete steps, then,some improvements was made to adapt it this project.4. The system's software was developed in VC++ platform. By analysis of the data from simulation experiment, it shows that Auto-detection based on machine vision system used for flaw detection of stationery can achieve the expected goal.