A Method for Color Classification of Fruits Based on Machine Vision

A dominant color histogram matching method for fruits classification was presented in this paper. In classification of fruits based on machine vision, image was acquired with a color CCD camera that outputted color information in three channels, red, green, and blue. Because traditional RGB color space couldn’t meet subjective color sensation of human being, so color image needed to be transformed from RGB to HSV color space which represented human being's subjective color knowledge. However, the conversion result was still three-dimensional information that made determining color grades very difficult. A new color space conversion technique that could be implemented for high-speed real-time processing for color grading was introduced in this paper. The result of this technique was a simple one-dimensional array that represented different color levels .These colors were known as dominant colors of fruits. The technique reduced computation consumption greatly. Color histogram as a statistical feature had visual invariance and high robustness. The dominant color histogram matching method was used for color grading. Grade judgment result was given by calculating and comparing the similarity between the inspected sample histogram and standard template histogram for every grade, fruit sample would be assigned to the grade whose template had the biggest similarity with it. Experiment results show that dominant color histogram matching method has high accuracy in fruits’ color classification. Key-Words: fruit classification, color grading, color classification, histogram matching, dominant color, machine vision, feature extraction 1 Instruction China produces a large number of fruits every year. However fruits quality inspection is still executed manually now, which is characterized as labor intensive, slow, sometimes inconsistent due to operator fatigue and high staff turnover caused by boredom, so machine vision based fruits inspection system has large market potential in China. Color is an important quality characteristic of fruits. It represents the degree of maturity, sugar content, acidity,taste etc. Some colors are more preferable than others. For example, in fresh fruit market such as apples and peaches, darker red represents higher quality than the light red colors. Fruits with preferred color generally can be sold for higher price. Color grading based on machine vision technology has become a primary way to maintain consistent quality and increase the fruit value . Some applications for color grading in Agricultural Products such as fresh market apples , peaches , tomatoes , potatoes , peppers [13] and cucumber [14] have been developed. The objective of this research is to introduce a new WSEAS TRANSACTIONS on SYSTEMS Changyong Li, Qixin Cao, Feng Guo ISSN: 1109-2777 312 Issue 2, Volume 8, February 2009 color grading technique which is simple and easy to implement. In color machine vision applications, image is acquired with a color CCD camera that outputs color information in three channels, red, green, and blue. Because traditional RGB color space can not meet subjective color sensation of human being, so color image need be transformed from RGB to HSV color space which can represent human being's subjective color knowledge. However, the conversion result is still three-dimensional information that makes determining color grades very difficult. A new color space conversion technique that can be implemented for high-speed real-time processing for color grading is introduced in this paper. The result of this technique is a simple one-dimensional array that represents different color levels.These colors are known as dominant colors of fruits. The technique reduces computation consumption greatly. Color histogram characterizes the distribution frequency of each color in images, as a statistical feature, it is not sensitive to the changes of scale, translation and rotation. So it has a visual invariance and a high robustness. The dominant color histogram matching method is used for color grading. Grade judgment result is given by calculating and comparing the similarity between the inspected sample histogram and standard template histogram for every grade, fruit sample will be assigned to the grade whose template has the biggest similarity with it. Figure 1 shows the classification process with histogram matching method. Compared with the traditional qualitative classification method, the classification method based on the dominant color histogram matching does not need specific training, and the design of the classification is simplified. The classification system is able to meet the practical requirements. In this paper, tomatoes were selected as samples to illustrate how the method for color classification of fruits based on machine vision works. Fig1. Schematic diagram of histogram matching method 2 Image pre-processing 2.1 Background segmentation The digital image of fruits (on the production line) grabbed by camera is composed of fruit and background. In order to obtain fruit’s feature, the fruit need to be separated from background. It is very important to choose a proper color space for effective image segmentation. The OHTA color space used in this paper was proposed by Ohta , who analyzed more than 100 color features which were thus obtained during segmenting eight kinds of color pictures and found a set of orthogonal color features. Compared with other traditional color spaces, the conversion from RGB to OHTA color space is linear and computation is inexpensive. OHTA color space can be effectively applied on color image segmentation. OHTA color space has two different kinds of expression as shown in equation (1). In this paper, Fruit segmentation is achieved by a threshold algorithm with the I2 feature (R-B). ( )