Research And Implementation Of A High-precision Visual Identification And Positioning Method For Specific Components

Usually we use machine vision to identify and help locate the target. But for the identification and location complicated components in certain conditions, there is no feasible general way. In the process of relative motion of camera and target, there are some difficulties in identification and location.Traditional methods of recognition apply only to regular objects, so these methods can not identify the complicated components fastly and accurately. It is difficult to achieve accurate calibration in a state of motion by traditional methods due to their necessities to calibrate reference objects. Though the camera self-calibration in motion has better flexibility and practicality than the traditional method, it is still long before it really addresses actual engineering issues for its positioning accuracy and robustness still be in experimental stage.For certain complicated components, a high-precision visual identification method was brought up, which is based on edge extraction and geometric calculation. By extracting the reference line which records the original pictures and shooting positions, comparing the reference line with datum line for different positions, and finally calculating the results using mathematical modelling one can accurately identify the target, and use the location result to achieve auxiliary location. The precision of this method reaches 10μm,it takes about 600ms to treat a picture of 2452*2056px, which meets the requirements of high-precision. By placing a variety of different positions of practical components, combined with micro adjustment, one can accurately, quickly and reliably identify the data needed. It has been proved by experiment that this method is stable and reliable, and could be applied in high-precision identification of complicated components.For positioning complicated components in camera motion state, an approach to a high-precision recognition and positioning of camera in a state of motion was brought up by transforming tough positioning to accurate one. This solution can achieve high-precision recognition and positioning of camera in a state of random movement. By experiments and tests, the contribution of the paper is the giving of a practical solution for high-precision recognition and positioning of components in the operations of the complex system. The accuracy of recognition system reached 0.01mm which is accurate and stable made it possible.