A Generic Algorithm for Segmenting a Specified Region of Interest Based on Chanvese's Algorithm and Active Contours

Image processing and recognition is a modern field which is gaining popularity due to its capability to automate certain mundane object recognition tasks and provide unparalleled accuracy and precision. Computer graphics have evolved sufficiently so as to cater to a wide array of applications ranging from categorizing mechanical parts of a machine to identify foreign objects/tumors inside the human body [1]. In this particular application, we are working toward identifying cotton plants from a heap. Traditionally, this task has been performed by laborers who identify the cotton plants from a collection and classify them based on certain characteristics. This task can take anywhere from a day to a week depending upon how good the yield is. We propose that this entire task from identifying cotton to classifying it based on the prerequisite characteristics can be performed automatically. We have developed an algorithm that can provide the exact position of cotton buds from pictures of the harvested plant. From the hardware aspect, this is the scope of automating the harvesting task. We have used the techniques of image acquisition, segmentation, and feature extraction. A highly modified iterative version of Chanvese's algorithm is utilized to chalk out a starting boundary (contour) and then work successively on it to reach at a final segment that defines the cotton buds.