Machine Vision Technology And Application Of Scratch Detection

In the process of industrial products manufacturing,because of raw materials limited and the process technology problem and some other causes, product surface appear scratches, it will not only affect the appearance, the more likely to reduce the quality of the products. Feather piece was the main raw material in the production of shuttlecock, its quality directly affect the quality of the finished product of shuttlecock. The quality of shuttlecock depends directly on the strength of the feather bar, and the scratches on the feather bar was an important factor of the strength. Light scratch makes the product unbeautiful, but the heavery one makes the quality goes down, and easy to broken off. Therefore, this paper use the technology of machine vision and image processing techniques to realize the automatic detection of scratch on goose feather bar, the main raw material in the production of shuttlecock. It has important actual meaning and economic performance.The scratches on feather bar are different from ordinary object surface scratches, including tiny size, the small color contrast between scratch and surrounding tissues. It was sorted by human hands for a long time, wile general machine vision inspection method of detecting precision insufficient due to put into practice. This paper using wavelet transform method to detect feather bar scratches, mainly solved the following questions:1. According to the characteristics of feather bar scratches, this paper choose the suitable LED light source and the CCD camera,and built up the feather bar scratch test platform and get scratches features obvious being testing images.2. Feather bar scratches are different from the general object surface scratches, the different scratch test object, the different scratch characteristics, and the applicable wavelet and decomposition layers are different too. This paper analyzes the characteristics of feather bar scratches, experimental different wavelet base and the different wavelet decomposition layers, determine suitable for feather bar scratches detection based on the wavelet DB1 single-layer decomposition. 3. Wavelet decomposition detection platform to the collected feather bar images, and the scratch test effect is good. This paper also analyzed the effect of detect with the size of input image, the size of scratch on feather bar, the length of feather bar, further verified this method the testing precision and efficiency.4. According to wavelet decomposition the paper got image threshold segmentation, make scratches features emerges, summary scratches features, and put forward on the judgement of the feather bar scratches.The paper completes all the arithmetics through using the software of Matlab7.0,the software deals with the images of feather scratch defects and gets the the result of emulator analysis. The last method has been decided and the good effect has acquired from the research of many kinds of method and the influence of all the data.