The Study Of Video Detection Technology For The Steel Pipe Defect Detection

The late seventies of twentieth century, video detection technique has raised up, which studies for road traffic information mainly. Video detection is a non-contact detection method, which also named image processing or artificial vision and video testing technology is a combination of video images and computer-based pattern recognition. By way of video cameras and computer to imitate the function of the human eyes, the video detection technology is the one of the newest advanced detection technology. This thesis introduces video detection technology into steel pipe defect detection, which realizes video image analysis processing for steel defect detection and provides the basis for the quantitative and qualitative analysis of the steel pipe defect. The main results are as follows:1.This thesis constructs binocular vision measurement system. The comparing of SUSAN and Harris algorithm have been analyzed in detail. To meet system needs, we introduce multi-scale theory into Harris algorithm, at this time the direction of iteration is along the small-scale to large-scale. Experimental results show that this corner detection method has high precision and accuracy .2. The corner stereo matching has been applied in this thesis. On the basis of gray correlation similarity, we obtain coarse matching. Under the condition of unknown epipolar constraint, we use robust estimate method to obtain the fundamental matrix. We obtain epipolar constraint by computing. We take advantage of the unique constraint, consistency constraints, disparity gradient constraint and the epipolar constraint to remove false matches. Experimental results show that this method can get epipolar constraint and matching rate correctly and highly, but this match speed is slow, the reason of that is many numbers of coarse matching set.So, we need to judge the coarse matching corner before recursive searching: if the left image corner and its corresponding right image corner has a difference which is not in the threshold, we judge them as a false matching and eliminated. Experiments results show that this method reduces the matching time and has a good result.3. Use of binocular vision measurement principle and three-dimensional reconstruction technology,we obtain three-dimensional coordinates of the surface feature points of steel pipe and the internal defects points .Then ,we change three-dimensional coordinates into left camera coordinates. In this way, we get the depth of the defect, which is the difference between the two z coordinates.