Detection and Counting of Connected Lentil Grains using Convex Deficiency for Quality Estimation

Abstract

Proposed here is a simple and effective method for identification of connected grains in an image. The grains considered here are the whole lentil grains which are circular in shapes. These circular shapes can be considered as convex polygons. When two or more such convex polygons are connected or partially overlapped, there are concave regions along the boundary line of the connected region. Such concave regions form the convex deficiencies for the convex hull of the connected or overlapped grains. Finding the number convex deficiencies in the convex hull of connected object boundaries allows us to identify whether connected component is formed by connecting two or more objects or grains. Also, it allows us to exactly count the number of connected objects and simplifies the finding the segmentation boundaries. The method is simple, fast and gives good accuracy of counting connected grains without actually segmenting the connected grains.

Keywords: Convex hull, Convex Deficiency, YCbCr color model, Image segmentation, Connected components