Study Of Surface Flatness Detection Method Based On Machine Vision

Abstract:With the progress of society and the development of science and technology. The requests about product quality have been demanding increasingly. Consumers and manufacturers are paying more and more attention to surface flatness which is an important aspect of product quality. This paper focuses on mirror surface and non-mirror surface of polished titles and roads respectively, and studies the flatness detection system based on vision technology.Firstly, according to tested contents, this paper analyzes the technology path and divides the hardware constitution and software module of the system. Based on reflection characteristics of the detected surfaces, adopting different light sources visualizes the roughness of the two kinds of surfaces. Considering the positional relation between camera and light source and the relative motion relation between detection device and detected surface, this paper designs different image capture method to improve image quality, which results in the best image capture effect. In mirror test, point matrix projection image collection is followed by barycenter extraction. This paper proposes an anti-interference window scanning method for extraction, arranges the extracted points by row and then column, analyzes the impact of ceramic tile incline on test results, and gives an algorithm to eliminate this impact. By analyzing the distribution of lattice in bitmap under different rough modes, a direction statistical method is proposed to qualitatively analyze and judge the smoothness of ceramic tile. This method also considers the offset and migration direction of corresponding points:firstly filter the offset points by offset value, secondly perform statistical analysis of the proportion of offset points in each offset direction and compare this result with offset direction of the points under rough mode. These two similar distribution indicates the roughness of the tested brick. In non-mirror test, this paper simplifies laser triangulation measurement model and deduces a computation model for elevation value of a corresponding road. In the process of calculating model parameters, we need to mark the camera and thin the laser line. After obtaining the parameters, we calculate the elevation value and give a flatness indicator. By thinning laser line, a road network is constructed, from which we can conveniently find the position of the concave and convex region composed of line segments. According to the length of line segment, we separate and extract each concave and convex region to be measured to get size information.