Skip to main content
ARS Home » Midwest Area » East Lansing, Michigan » Sugarbeet and Bean Research » Research » Publications at this Location » Publication #286401

Title: An image segmentation method for apple sorting and grading using support vector machine and Otsu's method

item Mizushima, Akira
item Lu, Renfu

Submitted to: Computers and Electronics in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/17/2013
Publication Date: 4/1/2013
Citation: Mizushima, A., Lu, R. 2013. An image segmentation method for apple sorting and grading using support vector machine and Otsu's method. Computers and Electronics in Agriculture. 94(1):29-37.

Interpretive Summary: Optical imaging technology is now widely used for sorting and grading agricultural and food products. It relies on the implementation of an appropriate image processing algorithm to achieve accurate classification results. Image segmentation, which separates objects from the background, is the first critical step in image analysis that can directly affect image processing outcomes like estimation of product size and shape. A number of image segmentation algorithms have been developed, but they suffer some major deficiencies, such as being inconvenient and time consuming in training or being unable to adjust automatically to changes in the lighting condition or variations in the color of products. This paper reports on the development of a new image segmentation method for sorting and grading apples. The method is based on linear support vector machine and Otsu’s thresholding method, two widely used methods in imaging segmentation, to achieve automatic adjustable segmentation of products from the background. A computer algorithm was developed, and it was then evaluated for 300 ‘Delicious’ apples of three color classes (i.e., orange, stripe and dark red). Results that showed that the new method achieved consistent and accurate results with the segmentation errors of less than 2%, compared with a popular fixed support vector machine method, which yielded errors between 3% and 25%. The new method is fast, easy to train, and suitable for automatic sorting and grading of apples. It can also be used for automatic inspection of other agricultural and food products, and for automatic recognition of targets in precision agriculture applications.

Technical Abstract: Segmentation is the first step in image analysis to subdivide an image into meaningful regions. The segmentation result directly affects the subsequent image analysis. The objective of the research was to develop an automatic adjustable algorithm for segmentation of color images, using linear support vector machine (SVM) and Otsu's thresholding method, for apple sorting and grading. The method automatically adjusts the classification hyperplane calculated by using linear SVM and requires minimum training and time. It also avoids the problems caused by variations in the lighting condition and/or the color of the fruit. To evaluate the robustness and accuracy of the proposed segmentation method, tests were conducted for 300 ‘Delicious’ apples using three training samples with different color characteristics (i.e., orange, stripe, and hard red) and their combination. The segmentation error varied from 3% to 25% for the fixed SVM, while the new adjustable SVM achieved consistent and accurate results for each training set, with the segmentation error of less than 2%. The proposed method provides an effective and robust segmentation procedure for sorting and grading apples in a multi-channel color space, and it can also be used in other imaging-based agricultural applications.