The Study And Application Of The Pre-processing Arithmetic Of Computer Vision

Before products' inspected ,we must work by the pre-processing of computer vision in the ourreal inspection system, which include the noise restraining, the enhancement of image contrast andimage edge detection and so on The error and the mistake coming from The pre-processing willdirectly affect the correctness of decision~making in the our real inspection system, whereas theexactness coming from The pre-processing will help supply a reliant input data. Accordingly,there are two main purpose in this dissertation, Firstly, author study the different kindsof pre-processing approach, and succeed at programming them with VC+-I-; secondly, author applythem to the real inspection system.Chapter 2 introduce the commonly principle of edge detection to the signal with no-noise firstly,and also introduce hov to affect edge detection to the image with noise. Lastly, author depicture twomain noise-restraining approach Gaussian-smoothed filter and adaptive filter. Gaussian-smoothedfilter is a lowpass, and it controls the lilting effect by 0, and the bigger (iT is, the better thefiltering effect becomes. But because image' edge behaves high frequency, Gaussian-smoothed filtercan restrain noise efficiently and blur the image edge as well. However, adaptive filter based on gradsinformation is not so, it can both restrain noise efficiently and sharp image edges. Thereforeadaptive filtering is a effective approach in primary vision processing . Author has improved thearithmetic about adaptive filter after specializing the analysis to them because of its shortcomings,and experimentations indicate that the effect to restrain noise of improved adaptive filter becomefamous.After specializing the analysis to Lee's feature enhancement approach, author put forward it'improvement arithmetic in chapter 3.experimentations indicate that improved arithmetic betterenhances the image contrast.Author mainly study the image edge detection operators in Chapter 3anny operators.According to the validity of image edge detection and the localization reliability, canny studied theperformance of a optimal edge detector, and educed the mathematical form about optimal edgedetector .if we directly convolve image with canny operators to extract image edge, it takes muchmore time, and vhieh restrict canny operators apply to a real inspection system . Though cannyoperators can convolve image ~vith separate approach and take less time than the former, it misfit areal inspection system because it take more time none the less. After analyzing earnestly the cause,Author find cann\' operators takes more detection time still when detecting image edge points withcanny' three conditions of detecting edge and replace it with "none-most compress approach", andfinally "none-most compress approach" make canny operators apply to our real inspection system.Finally, author apply above approach to our real inspection system to inspect bottle's crackperpendicularity error air bubble and stone. experimentations indicate that the effect becomefamous.The different kinds of pre-processing approach such as Gaussian-smoothed, adaptive filterCanny operators, and so on, had hardly been found their implement code in the existing papershowever, author implement these arithmetic with VC, which assure our real inspection systemsucceed...