Edge Detection of Growing Citrus Based on Self-Adaptive Canny Operator

A self-adaptive canny operator was developed to detect edges of growing citrus images. RGB color images were obtained and linear transformed into R-B chromatic aberration space at first. In R-B space, width of Gaussian filter fast calculated using integral images and the high and low threshold values obtained by OTSU algorithm were extracted to improve automatic edge detection. It is shown that the method we proposed can obtain connected and complete edges even the weak edges and can suppress the noise well. Moreover, it is less sensitive to lighting variations. Therefore, our method is more effective than other traditional methods in application of citrus edge detection. Once image edges are detected with this adaptive canny operator, the following identification and localization of citrus may be applied.