Foil Defect Detection System Based On Machine Vision

Vision is an important means for people to observe and realize the world. People feel the world through vision, feeling, hearing and smelling, however, 80 percent of the information is gotten through vision. Machine vision is the way that people use computer to realize their vision, which is the feeling, recognition and comprehension of objective in the real world. Machine vision is a new and rapidly growing research field, and it is widely used in many areas. Machine vision techniques will show their predominance in some situation that can't be felt by people, such as the detection of danger and so on.Traditional method for detecting foil defections is that operators try to find defections with their eyes. Obviously, this method mainly depends on operators experience and their opinions and can not guarantee repeatability and stability of detect result. How to automatically detect foil defection? The author did research work about this problem. In the paper, the author worked out a detecting algorithm based on machine vision theory and achieved hardware design and solved following problems.1. Detecting system principle, system hardware structure, software structure.2. Effective extraction and accurate recognition of defection image in industry spot.3. In whole system, communication net building and communication protocol definition between central controller and subordinate controllers.4. Defection image database design.5. The design of the interface between people and computer.The system algorithm includes correcting image geometrical distortion, converting into a binary image, extraction of the image feature, defection recognition and image compress.Firstly, because of high precision requirement and single camera in the system, so, we must take image geometrical distortion into account and correct it. We correct distortion in this system by 2 steps: space counterchange and gray interpolation. In the paper, we use cubic equation for space counterchange and bilinear interpolation for gray interpolation.Secondly, the mainly shape of foil defection is like pinhole. The background of defection image is black, but defection area is white. So, color of defection area contrast clearly with background and there is only one hill in the histogram of defection image. In order to simplify the process of image feature extraction and defection recognition, author converts the defection image into binary image in the paper. The most important thing in the converting process is how to find right threshold value. The paper gives out an adaptive threshold value methode. Through finite iteration, we can canculate different threshold values for different images for converting defection image into binary image.Thirdly, when defection area is bigger than 1mm, detecting system should recognise it. In the paper, we use circumcircle diameter of defection area and central point of defection area as defection area characters. After defection image is converted into binary image, we firstly scan binary image according to 4-direction connectivity rule. After scanning, we can find different defection areas in image and then we check defection areas one by one and find out all real defection areas.Finally, system is required to save all defection images that we find in 3 years in 10 different foil product lines. All images that we get from system camera are BMP format files and size of every image is 434KB. In order to save more images with less space, author compresses all images according to JPEG standard in the paper.The whole system consists of 10 subordinate controllers and 1 central controller and it is a distributed control system. Central controller communicates with subordinate controller through RS-485 serial bus. Central controller conmmunicates with every subordinate controller one by one and spent 0.5s for building connection and data transfer with each subordinate controller. Paper describes how to build up connection and communication protocol definition.Author sets up a defection image database with SQL Server 2000 and develops application program and interface between people and computer with VC++. Through simulation test, we find that the system has good stability, high precision, high speed and achieves all requirement of system.