Research On Object's Dimension Measurement Technique Based On Strctured-light Vision

In the modern industrial applications, three-dimensional measurement technology has been the hotspot in the measurement area, and the existed 3D measurement methods have been summed up in this paper. Structured-Light 3-D Vision Measurement Technique, which means to take the images as the carrier of information, makes use of the digital image processing technique to research the imaging principle of the objects, and obtain the measured object' size based on the theory of vision. This technique is focused on image processing and algorithm, which is less dependent on the external devices. It develops very fast because of many advantages, such as non-contact, higher precision, high speed, easy to abstract the information from laser stripe image, and on-line measurement.The key problems of Structured-Light Visual Measurement System are cameral calibration, character corners extraction, image processing and 3D information calculation etc. Firstly, the principle of vision sensor with the linear Structured-Light and the structure model of Structured-Light Vision System have been discussed. Meanwhile, a compare and analysis about the three popular calibration methods—typical linear calibration, Tsai's two-stage calibration and Zhang's planar pattern calibration, has been presented. Finally, a direct calibration method of the vision system with laser stripe is proposed based on the analysis. The concrete method is that the mathematical model of the visual sensor measurement system is established, and then the image of the planar calibration board which is in the same plane with the laser light plane is taken. Next the sub-pixel corners whose 3D word coordinates are known on the board are extracted based on the improved Harris's algorithm, and the corresponding relationships between the image coordinates and the 3D word coordinates of the corners are directly established. Finally, the visual sensor's model parameters are obtained through the known 3D coordinates and the corresponding image coordinates.In Structured-light Visual Measurement System, the key of ensuring accuracy is to be able to identify the center of the laser strip accurately. The laser light stripe images have been analyzed and processed in this paper, meanwhile the existed light stripe center extraction methods have been summed up. Finally, a new method to extract the light center coordinates is presented. The concrete step is that the red color components of the pixels on the color images are extracted according to the red laser characteristic, and then the laser strip is obtained based on the average algorithm. After filtering and segmenting the image, the binary image is scanned line by line, and the pixels which equal to 1 are detected, and their vertical coordinates are arranged in ascending sort; The lighted area is consisted of these pixels. If the number of the pixels in every line of the lighted area is odd, the accurate center coordinate of the lighted area is equal to the coordinate of the middle arranged pixel; if the number is even, the center of the lighted area is got by computing the average of the two pixels' coordinates which are in the middle of the increasing sort. Comparing with the existing methods, the experiment has demonstrated that this method not only can increase the precision obviously, but also can minish the error and reduce the noise. Meanwhile, the laser strip is extracted moreconveniently and this method has strengthened the integrality of laser stripe.In this paper, the software of the calibration of vision sensor with Structured-Light is designed and the processing of the laser stripe image is finished. The 3D coordinates are obtained by the image coordinates of the measured object based on the proposed model, and the radius of the cylinder is achieved according to the computation model. Base on an amount of experiments, this method has operated stably and can meet the requirements of the measurement system.