Development of Visual System for Watermelon Harvesting Robot (Part 1)

Relationship between Lighting and Detectability

Abstract

To develop an automatic robot harvesting system for the detection of watermelon fruit in the field, we used an algorithm under various lighting conditions and investigated the relationship between detectability and the average of grays. Results showed that 1) the fruit was undetectable by fixed threshold algorithm under various contrast and gray averages. 2) variable thresholding, however, to calculate the binary value of the image was effective regardless of gray average and contrast. 3) the combination of Gaussian Laplacian and local process improved the shaded image. Thus, the algorithm would be applicable regardless of lighting conditions.