Study On Image Acquisition And Inspection System For Red Steel Rod Surface Defects

The bar is an important raw material in industrial production, the quality of its surface has been highly valued. In the process of production, defects such as cracks, scratches, pits, ears which produced by the processing equipments and processes will become the sources of the stress concentration points and performance mutation points. These would influence the products’wear resistance, fatigue strength and hardness. Realization of the bar’s online surface defects detection and finding out the causes become the key link of quality control and process improvement, besides, it can provide data support for the following quality rating, reducing the workers’labor intensity and improving the production efficiency. This topic study the online detection system of red steel’s surface defect start from the machine view, carrying out the following works:The online detection system of bar’s surface detect was designed, that is check the surface quality through the surface images’acquisition and processing. First introduce the machine vision system’s structure and working principle of surface defects’ detection, and choose relevant hardware equipment according to the red-steel bar’s own characteristics. The bar processing’s line speed is relatively fast, surface images’rapid acquisition is the preliminary requirement of real-time online detection. The high temperature of actual production line exposes infrared light and visible light; this would influence the surface image’s acquisition and weaken the defect signals of collected images. After comparing multiple light sources’characteristics, we choose high intensity laser line light source and design filter device to make the surface images’gray scale evenly as much as possible. According to testing requirements and actual conditions, we choose linear array CCD camera and a telephoto lens as the image acquisition unit. And determine the reasonable layout scheme to form clear red-steel surface image easily. That is adopting forward lighting, putting laser line light source and line array camera on the same plane, keeping these two perpendicular to red-steel’s axes.The image preprocessing algorithm of steel rod surface was studied. The steel rod surface in camera view is cylindrical. The gray level images captured by industrial camera are bright in middle and dark in two sides. Besides, the images contain much dark background information which has nothing to do with defect inspection. The genetic algorithm based on bionics was emphatically researched. The binary encode method was used to do genetic operation, and the optimal solution was used to segment images. The object area was extracted and the useless background information was deleted. Genetic algorithm simulates the evolution of population. The scale of search is big and it is difficult to enter local optimal solution. It has big speed to convergence and has strong adaptive ability. It can effectively find out the global optimal solution in the parameter space and search the optimal threshold for segmentation. The steel rod surface images are influenced in the capture and transmission parts, so the quality reduced. This enhances the difficulty of inspection. To solve this problem, the filter algorithm based on local statistical characteristic and two dimensions Gabor transformer was studied. The experiments were carried out and anglicized. Gabor transformation has good property for separating information and noise. The filter algorithm based on two dimensions Gabor transformation has good effect for keeping the edge information of images.The inspection algorithm for steel rod surface defects was studied. The characteristics of common steel rod surface defects in image were introduced firstly. Through the analysis of red steel surface images captured in actual line, the gray lever characteristics of images were accounted. The sinusoidal function which seems like the pit shape on steel rod surface is selected to make convolution with the pixels gray values. According to the response value, we can judge whether the image has defect. For scratches defects, wavelet decomposition algorithm was used. The non-sampling wavelet decomposition method was used to get the coefficients in every direction. The vertical detail coefficients were used to make convolution with Gaussian function and the edge characteristic of defects was emphasized. The mean value of energy was computed and the high-low threshold method was employed to make image binary. At last, the whole shape of scratch defect was obtained through morphologic theory.The applications of rod surface defects detection system was tested online. In order to accomplish the high-quality surface image capture of red steel, the control parameters of image capture software are debugged, such as the width of the pixel, the time of exposure and the acquisition frequency. The layout plan of hardware equipment in work site was tested. Linear laser light source must be in the same vertical line with the linear array camera for image acquisition. All of this validated the correctness of the whole layout plan from the point of theory.