Research On The Key Technologies For On-Site Dimension Measuring Of Large Forging Based On Binocular Stereo Vision

Large forgings are the fundamental workpieces of heavy machines and equipments. Dimension measuring is viewed as a crucial technology in forging processing. However, the abnormal phenomenon to handle these hot parts often occurs due to the huge volume. Nowadays, artificial detection using mechanical instruments is the general way. So larger errors, longer cycles and higher intensity for the labors are caused inevitably, while it indeed exist potential safety hazard by exposing the operator to such hot parts. Furthermore, heavy burden to forging enterprises was also brought by the serious material losing. Therefore, more accurate measuring equipment and technology in forging process are desirable. On the basis of the domestic and international research, this paper proposed an on-site dimension measurement of large forgings based on binocular stereo vision. The crucial technologies including spatial location of double-CCDs, field calibration of structural parameters, immediate image processing and 3D reconstruction are in-depth researched.Firstly, the corresponding binocular stereo vision model with wide baseline was established, which was reliable and could satisfy such measurement requirements as remote distance, non-contact, large size, and the intrinsic parameters were calibrated with off-line. The intense shock pressure on site is apt to alteration of structure parameters of the model, which may cause erroneous measuring results. On account of large-sized hydraulic press more advantageous in the respects of the stability, the shape with parallel structure, orthogonality and coplanarity, geometric constraint in forging scene was constructed, which provided technical support to achieve structure parameters calibration in the field. In this foundation, the regular octagon target with sparse marked points was proposed and a field calibration method based on the geometric constraint in forging scene was suggested.Secondly, the reason was analyzed that vision measurement precision was influenced by high temperature objects. Image acquisition scheme by low-pass filter was presented, which effectively integrated spectrum radiation characteristics of high temperature objects, the spectral response characteristics of CCD. In any case, real time measuring should be took into account for guarantee the plastic deformation of forgings. So the automatic extraction algorithm was proposed following the visual attention frame, which was more advanced in the items of focusing the subsequent images on the interesting region rather than global image, and it could resolve excess computation problems validly when processing high resolution image. According to a biologically motivated center-lateral model of non-classical receptive field of border ownership cell, the edge detection algorithms was designed to suppress the trivial texture detail which often caused by oxides on hot forging surface. Besides, stereo matching about the points and lines feature with wide baseline were studied. Then, line segment matching in topology consistency constraints was realized, and there straight lines were described by neighborhood texture information.Thirdly, 3D reconstruction optimization algorithm about point coordinates and overall reconstruction about line and curve feature were researched. Based on this, reference-plane-measuring method in stereo vision was presented, which made an analysis on forging process characteristics, considered pressed surface as object to reconstruction and the vanishing point plays an important role as constraint condition.Finally, measurement space with geometric constraint was built for experiment and several geometric shape objects were measured by proposed method in the space.