Analysis of the Performance of Various EdgeDetection Techniques in Detecting Prominent Edges inPlant-based Images

Edge detection is a process of detecting the sharp intensity discontinuity in digital images. More commonly these discontinuities are found on the boundary of the objects in images. So edge detection is the significant step in identifying the objects in the digital images or in segmenting the image. Edge detection in digital image processing is achieved by convolving a 2-D image with a spatial filter which may be based on first order or second order derivatives. There are many classic edge detecting operators like Canny, Sobel, Roberts, Prewitts..etc. The goal of this paper is to analyze the performance of various edge detecting techniques in detecting prominent edges in plant-based images with the intention of getting clear boundaries of the leaves. That is, in this case we are interested only detecting the prominent edges which form the boundaries of the leaves. Many plant-based images particularly agricultural images consists lots of overlapping. These overlapping may be complete or partial. For instance, the leaves of crop may be partially overlapped on the weed plant or weed leaves. So applying edge techniques on these images and analyzing their performance gives us good understanding of these edge detecting techniques and how well these techniques can be used as initial processing steps in computer vision system in segmenting the weeds among crops.