A Real-time And Dynamic Stereo Vision Measurement System With Multiple Cameras

Compared with the binocular vision system,the multi-view stereo vision system has higher measurement accuracy in a wider range of measurement with broad application prospects,though its algorithms are complex and difficult to selection.The dynamic measurment of multi camera stereo vision is studied in this thesis.System design and algorithms research are carried out based on hardwares selection,synchronous image acquisition,image feature extraction,the image motion tracking,adaptive filtering and stereo calibration.A set of multi view stereo vision measurement system is developed.Considering the requirements of multi-view dynamic measurement for image synchronization acquisition and transmission rate,the image transmission interface and the type of the camera are selected.The multi camera synchronous acquisition system software is developed on the Lab VIEW platform.The multi-view stereo vision system model and software platform are set up.According to the relationship between the system structure and the measurement accuracy,the space layout scheme of the camera is designed.The algorithm of multi-view stereo calibration is analyzed.All the camera intrinsic parameters,distortion parameters and structural parameters are obtained by developing a new calibration toolbox.According to the actual situation of the calibration environment,a real-time algorithm is proposed to adjust the exposure time,which makes the checkerboard imaging clearer and the uncertainty of the calibration results smaller.The quadrilateral diagonal intersection is chosen as the dynamic measurement feature point,and the accurate extraction of the feature points is realized.In order to solve the problem of real-time application,the image tracking algorithm is studied based on motion detection update model.Compared with the nonuse of image tracking algorithm,the proposed algorithm can reduce the running time of 40%,and implement the real-time processing speed of 12FPS(Frames Per Second)for each camera of three cameras on the i7 four cores CPU.An IAE-DW adaptive filtering algorithm is proposed,which dynamically adjust the window size,and can obtain the better filtering effect.Experiments indicate that the average error of the trajectory measurement is less than 1.5mm in 1.5m depths when the maximum velocity is 64mm/s.