Excavator Manipulator’s Attitude Detection System Based On Vision Measurement

As automation technology is advancing and the working environment of excavators is becoming more and more complex, the study of excavator automation has become more popular in the excavator research area. The conventional contact-type sensors are usually used to detect the attitude of excavator’s manipulator while they can be easily damaged because of the collision and vibration. For this matter, a novel approach based on vision measurement is proposed to detect the attitude of excavator’s manipulator without contact.Firstly, a sheme of the excavator manipulator’s attitude detection system based on vision measurement is proposed in this thesis. In this system, an industrial camera with an infrared filter is used to grab the perspective images of the manipulator under the irradiation of a LED infrared light source; and the man-made target which contains three X-corners is fixed on every arm of the excavator’s manipulator. Furthermore, the kinematics of the excavator’s manipulator is analyzed and the attitude computation method in this vision measurement system is derived.Then, the camera calibration method applied in the system proposed before is studied; also, two X-corner subpixel detection algrithms are studied, which are intensity profile based algrithm and Hessian matrix based algrithm.Further, by analyzing the characteristics of the target images in this vision measurement system, a target matching approach based on geometric constraints is presented. And a moving target tracking algrithm based on neighborhood searching is adopted to track the man-made targets.Finally, the vision measurement software is programmed by using the Hessian matrix detecter; and the static and dynamic experiments have been conducted with the whole vision measurement system. Experiment results show that, the measuring accuracy and the real-time performance of this vision measurement sytem satisfy the requests of the excavator’s manipulator attitude detection. In concrete terms, comparing with the contact-type sensors under the static conditions, the vision-based method’s deviations of the measured attitude angle of the boom, the arm and the bucket rod are all within ±0.2 degrees. While, under the dynamic conditions, the deviations of the boom are within ± 1 degrees; and the the deviations of the arm are within ±2 degrees in most time, but within ±3 degrees occasionally; and the deviations of the bucket rod are within ±1 degrees in most time, but within ±3 degrees occasionally. Experimental data also demonstrates that the process time of each frame is less than 80 milliseconds.