Detection of Fruit Skin Defects Using Machine Vision System

External appearance is one of the most significant attributes for fruits when consumers decide to choose or reject them, thus packinghouses need to adopt appropriate systems that are capable of detecting the skin defects for fruits before packing them into batches and reaching the end consumers. For this purpose, this paper proposes a new method to detect fruit skin defects by using machine vision system, which is proved to be more accurate, more robust to color noise and has more modest calculation cost. The color histogram is extracted in the local image patch as image feature, while the Linear SVM (Support vector machine) is used for model learning. In a case of orange inspection, this system realizes a recall rate of 96.7% and a false detection rate of 1.7%.