HIGH-SPEED VISION OF PLANTS

Abstract

The challenges in machine vision specific to the food and agricultural sectors include the requirements of high speed, variable product shape, and surface inconsistency. Object identification such as plants, has not been developed to as great extent in agriculture as in other industries, due primarily to the complexity of plant images. Out of the enormous volume of experimental work that has been published during the past decade, two major problems which have not been fully resolved yet, emerge: There is a limitation of algorithms execution speed, and most of the image processing algorithms used to recognize the determinant features require very expensive hardware to execute in real-time. In addition to these obvious problems, digital image processing encounters another difficulty. In order to analyze an object, it must be presented in the input image as a single item. This is not a real situation in the field. This paper proposes a solution to these problems by using an optical system which is not only fast but also cost-effective. Optical pattern recognition (OPR) techniques work with low cost components and at the same time are fast and operate at almost the speed of light. The system is based on a real-time element for displaying images. A computer simulation has been developed to demonstrate this concept. This system would be of great benefit to automatic identification of plants. Another important aspect is the embedding of fast image processing systems as a component of quality management systems.