High-Precision Localization Algorithm And Implementation For Multi-Edge

In recent years, with the improvement of people’s living standard, demand for cars is in rapid growth, at the same time of the valve market demand is also growing. In the face of such a large market, gave birth to a lot of domestic valve manufacturing company. But the current valve detecting technology is relatively backward, rely mainly on hand detected artifacts, measure size by eye and manual sorting, in the face of such a large valve production, this is not only low efficiency and poor quality but also a large number of labor. Indications that the traditional measuring technology has been unable to meet the demand of the development of modern manufacturing industry, and the image optical detection technology with no-contact, high speed, high precision, strong anti-interference ability, etc, especially in the modern manufacturing demand for detecting technology.In this paper, based on valve image detection system as the research object. In view of the present image detection system, the accuracy of the measurement is not high, the results are unstable, the arc radius and cone angle measurement is not accurate. To study it, found that the fundamental problem is to get the original edge localization algorithm of edge position is not very precise, especially the hypotenuse edge and circular arc edge position is not very accurate. According to the above problem, this paper presents a variety of edge and high precision positioning algorithm, the algorithm for different kinds of edge, using different testing methods, and combining with the sub-pixel imaging technology and the lens distortion correction, finally realize the edge location of the accurate positioning. The main content of this thesis are as follows:(1) By median filter and gauss filter for image preprocessing, eliminate the salt and pepper noise and Gaussian noise in image acquisition and for edge detection provides the high quality of acquisition at the back of the figure.(2) The variety of edge detection algorithm is put forward, on the edge of the different types with different edge detection algorithm, so as to achieve the horizontal edge, vertical edges and circular arc edge, hypotenuse more precise positioning.(3) In order to achieve higher precision of edge location, the edge pixel level coarse location is completely not enough, therefore put forward the geometric centroid sub-pixel edge location algorithm based on polynomial interpolation, the method can realize precise sub-pixel edge location and has a good stability.(4) Lens distortion correction is to ensure high accuracy measurement is one of the key factors, this paper according to the actual project in the presence of radial distortion, obtained the camera radial distortion function, related parameters obtained by the RAC calibration algorithm. Then according to the coordinates of the location of sub-pixel edge points, through the radial distortion function to find the edge points of the distortion correction, correction of the edge points of edge distortion correction based on the obtained, so as to get more accurate edge location.(5) Error processing method is put forward, to ensure the accuracy of the measurement data. Finally, according to the edge position mentioned above, the dimension detection algorithm is converted to corresponding edge position information to the corresponding dimension data, and the dimension is measured.