Study Of The Key Techniques Of Computer Vision Measurement

In some applications, it is required that some data associated with an object (such as its temperature, size and position) be measured by non-contact real time means. Because the manual measurement, which is laborious and time consuming, needs to contact with the probed object, it is not competent for this job. By means of computer vision measurement techniques, the data associated with the probed object can be obtained from the information extracted from the images. With the character of flexibility, fast speed, non-contact, high-precision and automation, computer vision measurement might be applied in different areas such as industry, medicine and national defense.This thesis is devoted to the study of key techniques of the infrared and visible vision measurement. The main contributions achieved are listed below:(1) This thesis presents a three-phase scheme for removing the effect of the infrared radiance coming from the interior of an uncooled infrared camera on the quality of image. First, from a set of samples and by least square fitting, the way in which the two-point-correction (TPC) data vary with the camera temperature is determined. Then, the TPC data are calculated from the camera temperature. Finally, the collected infrared image is corrected by the TPC method. Both theoretical analysis and experimental results show that, for a proper camera temperature, the scheme can remove the effect effectively.(2) This thesis presents a three-phase compensation scheme for coping with the effect of the infrared radiance coming from the interior of an uncooled infrared camera on the temperature measurement accuracy. The first phase acquires the calibration data and forms the calibration function by least square fitting. The second phase obtains the compensation data and builds the compensation function by fitting. With the aid of these functions, the third phase can determine the temperature of the object in concern from any given ambient temperature.It is known that acquiring the compensation data of a camera is very time-consuming. For the purpose of getting the compensation data at a reasonable time cost, we propose a transplantable scheme. The idea of this scheme is to calculate the ratio of the central pixel's responsivity of the child camera to the radiance from the interior and that of the mother camera, followed by determining the compensation data of the child camera using this ratio and the compensation data of the mother camera. Experimental results show that either of the two cameras can measure the temperature of the object with an error of no more than 2°C.(3) This thesis proposes a novel method for calculating the 3D coordinate of an object, which requires a single camera with a 3-axis accelerometer sensor rotated around an appointed axis. First, we formulate the rotation matrix and translation vector from one coordinate system of the camera to another in terms of the rotation angle, which can be figured out from the sensor. Second, with the camera calibration data and by coordinate system transformation, we propose a method for calculating the orientation and position of the rotation axis relative to the camera coordinate system. Finally, given the rotation angle and the images of the object at two different positions, one before and the other after camera rotation, the 3D coordinate of the object can be determined. Experimental results show the validity of our method.(4) This thesis proposes detection and area measurement methods of the defects on the surface of the material sample. First, through camera calibration, the lens distortion is removed and the mapping from image coordinate system to real world coordinate system is constructed. Second, with the aid of the weight of the edge between two vertices, which is calculated by the distance of colors of the corresponding pixels, the segmentation result of defects by graph-cut, random walker and power watershed method is improved. Third, these methods require the seeds of the defects or the background marked by manual, which is labor-consuming. We propose a novel auto-marked seeds of corruption defects method, by which defects are detected by the color and texture feature extracted from the first corruption defect processed by these segmentation methods. As a result, their seeds are marked. With the seeds, they are segmented. Repeating the process, all defects can be detected and segmented. Finally, the amount and area of defects are determined. Experimental results show the validity of our method.