Car Dashboard Computer Vision Inspection System

Automobile dashboard centralizes the meters of cars. It displays the operation status of cars. Although every meter has its own feature, there are speedometer, tachometer, temperature gauge, fuel gauge and warning telltales in the cars. Before automobile dashboard from the workshops fix into the vehicle, there must be some checking. Commonly, these tests are accomplished by workers who detect the position of pointer and the indication error of the gauge pointer. Workers judge the color and brightness of them, and find out other visible defect. In this way, the result will be infected by several subjective elements such as the difference of angle and distance when watching instrument, and the fatigue of operator. This method will lead to gross error and small reliability. Instead of computer vision, this thesis can test meters automatically by using computer vision technology. This way, the errors made by operators will be avoided, and the efficiency and accuracy will be improved.This paper relies on science and technology development planning item of Jilin Province (serial number20080535), research on the detecting of automatic automobile dashboard by computer vision, to design auto-testing system for automobile dashboard by using computer vision technology.The main design work of this paper includes:Designing signal source to drive automobile meters pointer, motion-capturing automobile meters image and it is placed in a computer. In the image preprocessing process, use median filtering method to removing noise, and use Lap lace operator to do edge detection, by otsu method for binary image. By using morphological image processing, the paper point out that when use closing operation at first and then thinning operation, remove the interference of image acquisition to get a much better result, which is confirmed by the experimental results. Using process optimization Hough transform algorithm extracting automobile meter pointer linear, improves the testing precision of dashboard meters pointer and processing time. Calculate meter reading by using distance method, to raise farther the detection precision of meters pointer.Methods proposed in this paper have been proved by experiments. This method is very effective in the pointer on a gauge extracting, which possess the strong capacity of resisting disturbance and stabilization. Using the computer vision technology to automobile meter on-line detection will gradually replace artificial detection, become the mainstream of testing meters.