Research On Measurement Method Of Geometry Size Of Profile Section Based On Image Technology

Image measurement technology has become a new research of modernized mensuration for its rapid, non-contact and so on, which has obvious advantages in the detection of small size, large size and complex mechanical structure. It is of great significance to the realization of industrial automation processing and testing. Aluminum profile plays a very important role in the modernization construction. With the rapid development of industry, construction and transportation, people put forward higher requirement for aluminum profile quantity and quality. Deviations of cross-section dimension is an important index of aluminum profile, the classical measuring methods can not satisfy the requirements of measurement because of the complexity of profile cross-section. In order to realize the automatic measurement and promote the development of image measurement technology and related theories, this paper takes ordinary aluminum profile as the research object, and uses two methods to measure the size of profile section: a single camera and dual camera, then studying the key technology. The main research contents are as follows:1 This thesis introduce the calibration technique and focuses on the computation of camera exterior parameters. According to image feature, the spot recognition algorithm is designed and the center is extracted by gray weighted centroid method with coordinates corrected. Automatic method to eliminate spots with larger error are designed, which improve the accuracy of exterior parameters. The common filtering algorithm and the boundary extraction algorithm are studied and a combined filtering method is proposed according to the practice. The Canny algorithm based on combined filtering is adopted to extract the edge of profile with good performance. An algorithm based on image skew correction was proposed to measure the proportionate relationships between image size and the actual size of profile.2 In the base case, the shape of profile is so complicated that some corners are very weakly. An improved method was proposed for detecting the corner combine automatic inspection with human-computer interactive. The auto detection algorithm based on the traditional CSS algorithm. Firstly the larger step curvature threshold method is applied to remove fake corners in small curvature region in low scale. Then fake corners are concentrated in the arc region with larger curvature, and minimal primitives are merged to remove fake corners on the arc. Lastly, the projection height method is applied in the judgment of the properties of the primitives and the separation and fusion of primitives. The results show that the improved detection algorithm can greatly improve the detection accuracy. In order to reach a 100 percentage of correct, correcting corners by means of man-machine interaction which produces good results.3 This thesis study the algorithm of feature points matching and 3D coordinates calculation when take measurements by the two cameras. According to the characteristic of image, this paper presents an algorithm of feature point matching based on translation sector in which epipolar constraint, principle of order and unicity are combined to get a good match. When considering the error, to make use of the foot position of common perpendicular of lines on different planes to get 3D points coordinate.4 Based on the above works, two different methods can be used to measure the parameters of profile, such as thickness, length, arc length, center and radius of circle. Measure the thickness and length of profile by vernier caliper with a precision of 0.02 mm as the standard value. Compared the experimental result with the standard value, then the origin of the measured error is analyzed. Experiments show that the results achieve the specified ultra high precision progressive of GB5237.1 when using both of these methods.