Research And Design Of Edge Detection In Machine Vision Based On Wavelet Transform

Possessed with the advantages of high speed, high precision, repeatability and objectivity, the machine vision, compared with manual operation, would significantly improve the efficiency and accuracy of detecting and assembling and its technology has become a hotspot.One of the cores of machine vision algorithm is image edge extraction, which provides basis for subsequent image processing algorithms.And the central parts of automation of detecting and assembling are extraction of image edge, analysis of extraction and obtaining the coordination of the detected object.In the paper,some of the classical operator in edge extraction,such as Robert operator, Prewitt operator, Sobel operator, Marr-Hildreth operator, Canny operator etc.,were analyzed firstly and conclusion were drawed on the basis of simulations.Considering the defects of classical edge detection,which is single scale edge detection and may lose the effective extraction of edge information,the paper proposes and designs a new improved algorithm.The new algorithm, based on multiscale analysis of the wavelet transform, gets rid of the complexity of multi-scale modulus maxima edge detection and become time-effective. Besides, it also avoids the loose of details resulted from single-scale modulus, reducing noise sensitivity and misjudgment rate.Furthermore, vision alignment system in machine vision had been designed so as to detect the edge detection algorithm and study technologies related to machine vision.The whole system is composed of four parts: image collection and processing, motion control and motion platform, among which the image processing algorithm is operated under VC+Open CV environment in; the motion control hardware is designed with the core of STM32F103, and the software program, such as communication program, machine control and home positioning return, are operated under Keil.Having been tested under the vision alignment system, it can be improved that the edge detection algorithm is equipped with high efficiency and reaches expected goals.