Computer Vision News - March 2017

Every month, Computer Vision News reviews a successful project. Our main purpose is to show how diverse image processing applications can be and how the different techniques contribute to solving technical challenges and physical difficulties. This month we review RSIP Vision ’s automated algorithm able to perform accurate grading of agricultural produce , used by precise agriculture companies to assess the quality of each item. RSIP Vision’s engineers can assist you in countless application fields. Contact our consultants now! Background Grading of fruits and vegetables is a classic problem in the agriculture industry: produce needs to be sorted according to size and quality, as determined by observing each item’s features. Originally, the task was performed by humans, with the obvious downfalls of limited throughput and low accuracy, due to the tediousness and the subjectivity of the task. Mechanical sorters Mechanical sorters can sort produce by weight and size. However, also this sorting technique has clear downfalls, as it does not detect defective items as needed. Farmers need to isolate produce with defects for various reasons. One of them is that they can increase revenues by selling goods of higher quality. Another is due to urbanization of populations and globalization of food distribution and consumption, which requires produce to travel long journeys that only the best products can survive unhurt. Detection of internal and external defects (in ways which we will later see) is key to assess fruits and vegetables life expectation and select the fittest produce for overseas shipments. Cameras used for this task can be RGB, IR and UV cameras, the latter being the standard for internal scanning, while the former ones can check external defects in the skin which are clearly visible on it. At this stage, algorithms were needed to classify images, often more than one for each individual item, in order to cover the whole surface of the skin. They were simple image processing algorithms: they had good capabilities but needed to be used together with human expert, while tuning needed to be done individually for each type of produce and even each source needed individual tune. That made the accuracy of the system highly dependent on the expertise of the human operator. Project 20 Computer Vision News Project Fruits and Vegetables Grading