Determining Banana Size Based on Computer Vision

Abstract

An automatic algorithm based on computer vision to determine three size indicators of banana, namely length, ventral straight length, and arc height, respectively, was developed in this article. The automatic algorithm calculated these indicators by three steps. First, banana was marked by image pre-processing. Second, the Five Points Method as the core part of the automatic algorithm was used to locate five points at the edge of banana. Finally, the Euclidean distances between two certain points were calculated to determine these indicators. The three size indicators of 28 bananas with slightly curved, curved, and end-straight shape were determined using the manual method, semi-automatic method, and automatic method, respectively. Results demonstrated that the automatic method was more precise with lower standard deviations and more accurate with a percent difference within 16 and 22% for the length and the ventral straight length, respectively. In conclusion, the automatic algorithm was acceptable for banana size determination.

Keywords:

• Computer vision
• Machine vision
• Banana
• Image processing
• Size
• Shape