Study On Image Measurement Algorithm Of Parts Geometrical Size Based On Image Mosaic

Image Measurement is a new kind of photoelectric measurement technology that uses the Charge Coupled Device, namely CCD to take photos and measurements.The paper aims at the problems of image measurement field such as the small view field and high cost, presents to take measures on the geometric sizes of parts base on image mosaic. We have improved the related image measurement algorithm. We study on the image measurement station of CTLS-1000 equipment in the Advanced Control Laboratory of Education Bureau of Guangdong Province contributed in Automation Faculty of Guangdong University of Technology.The paper first introduces the concept and the development situation of image measurement technology, points out the problems that exist in this field and presents the objective, significance and the research range. Besides, it also introduces the image measurement station of CTLS-1000 equipments and presents the overall structure of the algorithm.Second, through the pre-processing on the original images such as removing the noise pollution and blurring led by the motion of the object, the blurred image can be restored to prepared for mosaic operation. We adopts intensity correlation algorithm to make image mosaic operation and also make image fusion at joint area to eliminate false edges. Then, in order to realize the segmentation of mosaic image at the sub-pixel level, we improve the ant colony algorithm, and use gradient operator as ant colony Characteristic of Clustering, use two types of ant colony Characteristic of Clustering to make segment rapidly and precisely. At the mean time, in order to make the edge position more precise, we use the algorithm based on Lyvers Spatial Moment Operator to fix on the position at sub-pixel level. We improve the algorithm, only position the segmented points and thus reduce the time. We use Hough and the Least Square Method to fit the discrete points, and obtain the basic information about the object. Last, we make non-linear distortion compensation for CCD system, adopt quadratic calibration to complete the transform from pixel values to the real sizes, and thus we can get the geometric sizes of the object.Last, we take the mosaics of two images as an example to test the algorithm, and then analysis the precision of the algorithm.