A Color Channel Ratio Image based object detection method in complex scenario

In complex scenario such as underwater imaging or rough weather, local textures are unavailable due to image blurring and background clutters. Global features like color and contours became the important clues for object detection. However, it is challenging to extract the object region with regard to the variances of color and massive disruptions of object contour. In this paper, we proposed a new method which used Color Channel Ratio Image (CCRI) for edge detection. The CCRI-based edge detection method can effectively extract the object contour in complex scenario where traditional grayscale-based edge detectors often fail. Moreover, we design a fast object detection algorithm based on the CCRI image and chamfer matching. Performing k-means clustering on the CCRI image can obtain the candidate regions of the target object. Thus applying directional chamfer matching only in the candidate regions can efficiently speed up the detection procedure.