Research On Vision Tracking System Of 3D Textile Composite Preform Stitching Robot

To improve the flexibility and intelligent performance of the 3D textile composite preform stitching robot, a vision tracking system of 3D preform stitching robot is designed, which includes vision sensing system, image processing system, deviation recognition and control system and execution system.As an added light source, laser is projected in front of the seam. And the light stripe reflecting from the work-piece is received by a CCD camera. After being quantitated, the digital image is saved in the computer memory.Considering the characteristic of textile composite noises, an effective image preprocessing method is put forward. Image graying, contrast intensification, binarization, median filtering and opening algorithms are used to preprocess the textile composite laser image to eliminate the noises caused by burrs and braided pitches.Experimental results show that the preprocessed laser stripe on the stitching textile is discontinuous and has nonuniform width, so it is difficult to detect feature points directly. The laser midline is extracted by the way of thinning operation of mathematical morphology. Because the midline is flexural and discontinuous, the general method such as slop analyses is not practical. As a result, feature points are detected more quickly by the way of comparing vertical coordinates.Then, feature points excursion in horizontal and vertical direction, which is also the stitching head excursion, is computed based on the principle of lens image and optical triangulation. Next, the controlled quantum is developed by the excursion and outputted to the actuator, and the actuator drives the single side stitching machine to move. Finally, the automatic tracking of sewing textile seam is realized.The whole system is developed under Microsoft Visual C ++6.0 with high accuracy and good reliability. By experiment and analysis, the validity of image processing algorithm and control algorithm are verified. The whole time is less than 250ms, about four real time seam images which are processed precisely can bedistilled during one second, which can meet the demand of the robot stitching system.