| Abstrakt: |
Rice milling industries are seen in large numbers in the eastern and southern parts of India. Like the other industries, this is also going through changes because of the introduction of automation using machine vision technologies. In the consumer market, buyers consider that a good quality rice bag must not contain broken kernels. Considering this commercial aspect, during rice production, the estimation of broken rice kernels mixed with whole rice kernels is an essential grading task in the husking mills. In this paper, we propose a method for monitoring the presence of broken rice kernels in the rice outputs during milling. In our proposed method, during milling, rice kernel samples are captured, and image understanding-based processing is applied to realize broken and whole rice kernels. A whole-grain kernel is distinguishable from a broken-grain kernel in terms of its size, perimeter length, etc. But many times, several kernels get attached to one another during image capturing. These attached kernel groups need to be separated into single kernels so that the shape and size information of kernels may be used to estimate the amount of presence of broken rice kernels. Hence, the proposed method works in two stages. In the first stage, attached rice kernels are accurately separated using geometric analysis of the boundary edges of the attached kernel groups, and then, in the second stage, all individual kernels are clustered into either broken or whole kernel categories based on their shape and size information. The proposed method identifies the broken kernels with accuracy of 98.25 % , 96.54 % , and 98.52 % for Shamashri, Lalswarno, and Basmati rice kernel respectively. Finally, we present statistics of the estimation depicting the percentage of broken and whole rice grains in the sample images. Our test results indicate the effectiveness and applicability of the proposed method. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink