Study Of Pairing Vision Measurement In Large-scale Section

Study and comparison on the principle of large-scale component section measurement system is carried out for its demand and application, based on status quo and development, so an online vision measurement system using vision sensor pairs is proposed for large-scale component. Each pair of sensors in the system is characterized by differential measurement so as to obtain higher single point measurement precision. The global calibration of sensor coordinates is accomplished by coordinates measurement arm and laser tracker, improving calibrated precision and efficiency. The geometrical shape of large-scale component section is detected quickly through this system with high precision.According to theory and practice, some research is finished deeply and corresponding experiments is summarized as follows:1. The characteristics and application is analyzed for the research of large-scale coordinate's measurement methods home and abroad. The technology of vision detection is introduced through its application in manufacturing, preparing for the large-scale section geometrical shape measurement system.2. The class of present vision sensors is investigated, including monocular sensor, binocular sensor, and multi-ocular sensor. The sensor pair using binocular sensor is proposed for sensor measurement net, considering cost, precision and so on.3. The calibration of large-scale section geometrical shape vision measurement system includes local and global calibration, while local calibration is composed of Videocon calibration and parameters of sensor structure. The field calibration with laser arm and theodolite can improve efficiency.4. The algorithm of ellipse fitting is also studied in this paper. The main methods of ellipse include Hough transform, the least-square method, and boundary fitting, which is characterized with complicated calculation, long time and low efficiency, so a method based on that the line joined by midpoints of parallel chords must cross center of ellipse is proposed, and finally the center coordinates of ellipse is calculated for fitting the figure of ellipse.5. Image processing technology with high precision plays an important role in vision measurement. The interfering error should be weakened for the whole work. Since image is collected by CCD of sensor, so image pretreatment and characteristic point's arithmetic have an effect on the whole component measurement precision. This experiment uses the methods of median filter transform threshold and tessellate to preprocess the image to stand out the characters of it. Then pick-up the pixel of unique point according to the program edited by myself.