Development Of Automobile Cluster Auto-Calibration System Based On Vision Technology

As one of the measurement instruments, the gauge with pointer is gennrally used in the manufacturing process as present. Automobile cluster is the interface between automobile and human being. It can supply the driver with necessary information of running status such as malfunction, speed and so on. That makes it an essential part in an automobile. There are speedometer, tachometer, temperature gauge, fuel gauge and some warning lights. Traditionally, the gauge is recognized by human eyes. However, some of the subjective factors suck as:observation angle, observation distance and fatigue may have effects on the value. This methord will lead to not only gross error, small reliability but also low efficiency and body fatigue.Instead of human vision, we can test clusters automatically by using computer vision technology. This way, the errors made by operators will be avoided, and the efficiency and accuracy will be dramatically improved.In this paper, computer vision technology is used to develop auto-testing system for Automobile cluster. At the beginning, I introduce the developing history of computer vision technology. And then I indicate the thinking in the design of the whole system including hardware and software, and give some knowledge about CAN--the most commonly used way for communication with in an automobile. Then I illustrate some fundamental theories of vision image processing and the fundamental methods and theories of recognition. Chapter 3 is one of the central part of the paper. In this chapter, I indicate how to test and recognize pointers, scale line and warning lights. The main research work of this chapter is generalized as following:To get the axis of pointer by using Morphological image processing technology, I point out that when use closing operation at first and then thinning operation, I can get a much better result, which is confirmed by the practical result;After Hough transform, I suggest getting the area near the straight line which has been detected by Hough transform. Then, in this area, I use regression analysis to get the more accurate parameter of equation of the straight line; I make a study and propose some new fast processing methods. These methods meet the real need in testing automobile cluster in a factory;I propose a new method to detect the chromatism of warning light by converting colors from RGB color model to HSI color model;Methods proposed in this paper have been proved by practice. The method to develop the software by using cluster template is a good way which can improve efficiency of software developing and cluster testing. At the end of the paper, I introduce the developing tool and environment also the thinking in developing this system.