Research on Error Correction Algorithms of Oblique Optical Axis in Visual Measurement Based on BP Neural Network

Abstract

Visual measurement technology is widely used in the field of non-contact measurement. However, in practical measurement, the measurement accuracy is reduced, because of the vertical error between the optical axis of the camera and the plane of the camera. Finally, the corresponding error angle can be obtained directly by acquiring the information of the image punctuation, and the three-dimensional coordinates of the corrected points also can be calculated. To verify the feasibility of the algorithm, a visual measurement platform is built for image acquisition. By collecting the scale relation of rectangular edges and the data set that causes the error angle, the function relation model between proportion and angle is obtained by training the afferent neural network. The BP neural network fitting value is compared with the actual measured value and the calculated value. The result shows that by using BP neutral network approximating function, the minimum mean square error can reach 0.5075mm which is close to the expected error of 0.50mm, and the oblique optical axis correction error of single camera in visual measurement can be achieved. In this paper, the visual measurement errors caused by oblique optical axis are theoretically analyzed and an oblique optical axis error correction algorithm based on BP neural network algorithm are proposed.