Search

This paper presents a robot autonomous calibration approach using a probe. The robot grips a simple probe automatically to touch constraint planes in a workspace (the locations of the constraint planes are not necessarily known exactly). The robot internal sensor measurements are recorded for kinematics calibration while the tip-point of the probe is in contact with the constraint plane. The kinematic constraint conditions are obtained from the known shape of the constraint surface, rather than from the measured reference locations in a workspace. This method eliminates any use of external measuring devices, and is suitable for a periodic robot re-calibration. Simulation results on a polishing robot show that this method can achieve the same calibration accuracy as that of the other traditional calibration approach such as geometrical parameter calibration.

Mechanism models used in cold continuous rolling mills can reflect the general trend of production process. But the precision of models can't satisfy the more requirements of production quality, variety and performance. In order to improve the effective of rolling models, data mining technology can be used to study and find the valuable information and knowledge, which is the newly item and research direction in rolling process field. At present many cold rolling production lines have finished the basic modification works, can acquire and save amounts of process data from the production spot. Various methods such as SVM are used to study the rolling process by model and data from another direction in the paper. The results show that data mining method can help to find more knowledge to improve the production control.

The moving target detection is the important area and key technology in machine vision, image understanding and terminal guidance. It finds a wide application in industry and military field. This paper presents a new approach to comprehensively detect the moving multi-targets in dynamic scenes. The method extracts the contour of each moving object based on the fusion of motion segmentation technique using difference of image and region growing process, and then the final result can be achieved after the optimization with active contour model. The advantages of this method are as follows: the complexity of the background information and the size of the target cannot affect the result of detection; viz. the detection method is of very robust and doesn't need motion compensation in the detecting process. The result of the simulation experiment has proved that the algorithm is reliable, practical and robust.