Search

Owing to the complex structures of welding materials, special welding conditions, and challenges during the automatic welding of the liquefied natural gas (LNG) ship Mark III’s membrane tank, a series–parallel–series hybrid structure mobile welding robot having sufficient adaptability for welding corrugated plates in membrane tanks was designed in this study. The configuration of the hybrid robot had good workspace characteristics, and it could always maintain a certain distance and angle between the end of the welding torch and the weld line because of its detection and control system coordination. In this study, degrees of freedom, kinematic characteristics, workspace, and adaptability analyses were conducted for the hybrid mechanism. A simulation verification was performed, and a ripple-trajectory-following experiment was conducted using real objects. The simulation and experimental results showed that the welding robot had a reasonable mechanism design, smooth motion, and good terminal distance and angle control, thus meeting the requirements for automatic welding of corrugated plates in membrane tanks.

Aiming at the problem that the traditional spatial arc fitting has low robustness and low fitting accuracy, a new method of space circle fitting is proposed in this paper. Firstly, based on the Lagrangian multiplier method, the objective function is established based on the constraint of the plane condition, and then the scheme of the fitting equation directly based on the spatial arc is derived. Secondly, the error tracking point is eliminated by using the RANSAC (random sample consensus) method, thereby improving the fitting accuracy. In this paper, three different algorithms are used to carry out experiments. The theoretical basis of the three algorithms is respectively based on least square method, based on Lagrange multiplier method and based on Lagrange multiplier method added RANSAC algorithm. Experimental results show that the method of spatial circle fitting based on Lagrange multiplier method is greatly superior to the traditional method based on least square method, and adding RANSAC algorithm on the basis of Lagrange multiplier method can further improve the accuracy of spatial circle fitting.

In binocular vision measurement systems, image matching is an indispensable part. However, in the measurement of simple workpieces, due to the high similarity of the workpiece edges and the complicated distribution, it is often difficult to match. The traditional shape context matching algorithm only considers the position distribution of contour points, ignoring its own gradient features, which not only causes a high mismatch rate, but also is computationally complex and takes a long time. Therefore, based on the texture features at the edge contour points and the joint polar line constraint, an improved shape context matching algorithm is proposed. On the one hand, the DLBP (Double Local Binary Pattern) texture feature weight value is introduced, and the edge point gradient information is used to make the algorithm more robust to noise and illumination changes, and reduce the mismatch rate. On the other hand, by combining the polar constraint, the search space is limited to a small rectangular frame, which reduces the computational time and further reduces the mismatch rate. Finally, the effectiveness and practicability of the proposed algorithm are demonstrated by experiments.

This paper improved two-dimensional Otsu algorithm in order to segment the Cold rolling thin strip defects of low contrast and complex noise. In order to avoid searching threshold in a large two-dimensional space, two-dimensional Otsu algorithm of two-dimensional histogram had been decomposed. The algorithm had solved threshold problem of the image pixel gray histogram and that of gray histogram neighborhood pixels respectively. The paper added balance factor to the pixel neighborhood grayscale in solving the segmentation threshold problem and used the threshold as the best segmentation threshold. Finally, the paper analysed the segmentation result through a large number of experiments, compared with two-dimensional Otsu algorithm, QPSO and improved algorithm. The results show that the method can split out various defects on the surface of cold rolled thin strip quickly and efficiently.

To solve the problem of precise edge localization and measurement on low-resolution images, statistics characters of edge gradient magnitude of both high and low resolution images along the gradient direction are extracted and analyzed. The gradient magnitude statistics is described with General Gaussian Distribution (GGD). The relationship between statistics of high and low resolution images is established via GGD analysis on training samples, based on which a high resolution gradient image is constructed from a low resolution image. Precise edge location and measurement are then achieved by double threshold edge detection and interpolation. Experiments show that the proposed method has better performance on precise measurement on two dimensional work-pieces than sub-pixel method only based on high precision interpolation.

Animation design for computer simulation technique of cylindrical centerless grinding process is proposed based on analysis on the processing character of centerless grinding and the techniques of computer simulation. 3DS file and model structures are applied to reconstruct the simulation model of the centerless grinding. Two-dimension animated simulation for the grinding process is realized combining OpenGL technology which is used to present three dimensional appearance changing of workpiece during processing. Testing on the system show that the simulation system presented in the paper has good real-time performance, friendly interactivity and ability to simulate two or three dimensional animation in practice.

In a three-dimensional (3-D) model-based objects tracking and recognition system, the key problem of objects location is to establish the relationship between 2-D objects image and 3-D model. Based on 3-D model projection and 2-D image feature extraction, a modified Hausdorff distance is used to establish the matching function. The relationship between matching parameters are described with a probability model, and the distribution of parameter evolves towards the direction of dominant character through probability model learning and the corresponding operation, which is proposed to solve the problem of overmany iteration and slow constringency velocity The experiments show that the optimal matching parameters between 3-D model and 2-D image feature can be found accurately and efficiently, and then the accurate object location is completed.