Design And Implementation Of Binocular Stereo Vision Distance Measurement System Based On DSP

Binocular stereo vision distance measurement technology is the focus and hotspot in computer vision research.Its basic principle is to use two cameras with relatively fixed position to obtain the left and right images from the same scene and different angles at the same time,then to acquire distance by using parallax.With the development of many years,the binocular stereo vision technology has been widely applied in many fields,for example,robot vision,vehicle navigation,3D reconstruction,engineering measurement,pattern recognition,etc.Based on the collection of a lot of relative information and materials,conducted in-depth study on the technology of binocular stereo vision distance measurement.Stimulated and tested the binocular stereo vision distance measurement by using the MATLAB software.Used binocular camera to collect the images of calibration and the object that need to be measured,and used the plane template calibration technology to complete the image calibration.Then the calibration parameters are used to realize the correction of the images,including the non-distortion correction and parallel correction of the images.Corrected image contains a lot of noise,and used median filter to deal with the noise.Through the comparison of Fast(Features From Accelerated Segment Test)feature points,the Harris feature points,SURF(Speed-Up Robust Features)feature points extraction algorithm,finally chose the Harris as the feature points extraction algorithm.Used the position information to narrow the range of feature extraction that to make the feature points are mainly focused on the target object.The normalized cross correlation matching algorithm is used to match the feature points on the left and right images,and then using RANSAC algorithm to remove false matching points.Finally,used the parallel binocular distance measurement principle to get the actual distance between camera and the object.In hardware development aspect,adopted the Da Vinci series ICETEK-DM6437-B development platform to achieve the image reading module,median filtering processing module,image feature points extraction module,feature points matching module,parallel binocular distance measurement module.Debugging by using the CCS3.3 software,and connected the communication between the computer and the hardware via simulator.Finally,through lots of distance measurement experiments,it is concluded that the system achieves good distance measurement effect within the camera imaging range,basically meets the requirements of distance measurement accuracy in general situation.