Study Of Circle Detection In Complex Condition

Computer vision regards image processing technology as the core, it makes the machine has ability of environment understanding. Vision measurement and location is applied widely because it has the characteristics of untouchable, fastness and high precision.This paper discusses the technology of circle detection and location in the complex condition of low illumination, asymmetry illumination and noise disturbing. And then this technology is used to detect the entrance of the oil train.Circle is often used in application of computer vision. It is the most essential and most important task to detect and locate circle in measurement based image. This paper discusses the detection in complex condition. But gray-scale image is influenced by illumination. So we must detect the edge of circle first in image preprocess and later location is based on edge image. For improving the detection speed, circle location is completed through tow steps: circle recognition and circle detection. These tow steps are also named approximate location and precise location. Approximate location is to confirm weather the circle target is existed in the special area. This step can be completed through template match. Cross-correlation algorithm, matrices match algorithm, SSDA and FFT correlation based on phase. FFT correlation method is frequency field algorithm. It has robustness and can be used to circle detection in noise condition. Golden tower search algorithm can improve the searching speed. Hough transform is used widely in circle detection. Its detection result is precise and robustness is good. But its calculate speed is slow and can not be used in real time detection. This paper discusses the least squares estimator circledetection method which is based on statistical theory. It gets the position through edge tracing and uses the edge point to estimate the center and radius of circle. And the result is sub-pixel accuracy. The least squares estimator circle detection algorithm is faster than Hough transform. It still can be used when edge is absent due to low illumination.FFT correlation and least squares estimator algorithm can be applied to locate the entrance of oil train and its detection result is good.