Summary: | 碩士 === 國立高雄應用科技大學 === 模具工程系碩士班 === 102 === The two-step method was first proposed by Roger Y. Tsai [1] in 1987 to derive the relationship between a three-dimensional world coordinate and a two-dimensional coordinate on an image. In that method, extrinsic and intrinsic parameters were used for the calibration of a camera so as to describe the camera pose and its characteristics. Another calibration approach based on the homographic matrix was later proposed by Zhengyou Zhang [2] in 2000. Both of the methods can induce some calibration calculation errors which further cause the subsequent inaccuracy in the field of reverse engineering.
In order to reduce the calculation error, the objective of this research is to combine the two-step and homography matrix method to calibrate a fixed-camera by single image. To achieve the calculated precision of extrinsic parameters, two-step method was more appropriate than the homography matrix because the absolute position of fixed-camera was explicitly represented by three-dimensional world coordinates. However, the calculated precision of intrinsic parameters can be affected by the calibration error in the extrinsic parameters calculation procedure and therefore result in the reconstruction errors in reverse engineering.
The first part in this study is to find the extrinsic parameters which describe the camera pose, and intrinsic parameters to describe the camera characteristics. In order to reduce the calibration error, this experiment reduces the calibration images numbers required in homography matrix method to calibrate the intrinsic parameters. The world coordinate of an object is then determined by using these parameters obtained to convert the image coordinate of a point intersected by the laser line and laser light plan on the object. A modified approach is proposed in this study to simplify the calculation procedure to reduce the measurement errors. It is shown that the measurement error can be reduced to less than 0.2 mm in a 30 mm diameter cylindrical object.
|