Shape and weight grading of mangoes using visible imaging

https://doi.org/10.1016/j.compag.2015.05.006Get rights and content

Highlights

  • Automatic grading of Harumanis mangoes by its shape and weight analysis.

  • The Fourier descriptor method was developed to grade the shape of Harumanis mango.

  • The cylinder method was developed to grade the weight of Harumanis mango.

  • Multi-classification using Support Vector Machine and Discriminant analysis.

Abstract

This paper presents the work on the use of visible imaging as a tool in grading the mangoes. A Fourier-descriptor method was applied on mango images acquired by a CCD camera, to grade the fruits by their shapes. The method was able to correctly classify 98.3% using DA and 100% using SVM. It is also possible to estimate the weight of the mangoes from their images by applying the Cylinder approximation analysis method. The scatter plot between the estimated and actual values of the weight shows high correlation, with R2 equal to 94.0%. The high prediction accuracy obtained shows that this simple formula is adequate for the prediction of fruit weight and volume (measured volume using the cylinder method). The correlation formula derived based on the collected data is determined as w = 2.256 V  157.7 where w is estimated weight in grams and V is estimated volume. Overall result for weight grading using our proposed method yields 95% accuracy.

Introduction

Mango (Mangifera indica L.) belongs to the family Anacardiaceae. It is grown extensively and commercially in India, Philippines, Thailand, tropical Australia, the lowlands of South-East Africa, Hawaii and in the lowlands of Central and South America. The ‘Harumanis’ variety is mainly grown in the northern part of Malaysia, particularly in the State of Perlis. Harumanis is considered the “King of Mangoes” and is very popular in Malaysia because of its sweet and aromatic fragrance (Musa et al., 2010).

Since the 1970s, the automatic visual inspection systems have become useful tools in industrial and agricultural process. They are used to detect malfunctions during the manufacturing process and control product quality (Gonzalez and Woods, 2002). Grading is an important operation to measure size, color, shape and defect of agricultural product. Image pre-processing is an important step for agricultural product quality system. Presently, there are many methods available for analyzing shape of an object, ranging from a simple multiple point features method to a complicated geometric features approach. The method used in this project was conceptualised by Zahn and Rookies (1972). It was based on Fourier Descriptors (FD). They provide detailed mathematical explanation of FD for object recognition, matching and registration. One unique feature of this method is that it uses global image descriptors instead of the local ones, making it more applicable to real-world images in which simple multiple point features may be difficult to extract, and eliminating the need for feature matching between the reference and observed images.

The image processing final step of agricultural external grading system is decision making which is the result of image processing steps for the sample. In the other hand it is sample classification and ranking method compared to the known samples which trained to the system before base on specific factors. Generally, there are different techniques used to train the reference samples to the grading system and then classify the new sample base on the training stage which are; statistical classification, neural network classification, fuzzy logic classification and then neural-fuzzy classification. Statistical approaches are generally characterized by having an explicit underlying probability model, which provides the probability of being in each class rather than a simple classification (Du and Sun, 2004). SVM is a supervised machine learning method that performs classification based on the statistical learning theory. Essentially, SVM is based on fitting a separating hyperplane that provides the best separation between two classes in a multidimensional feature space. This hyperplane is the decision surface on which the optimal class separation takes place. In order to represent more complex shapes than linear hyperplanes, a variety of kernels including the polynomial, the radial basis function (RBF), and the sigmoid can be used. Also, a penalty parameter can be introduced to the SVM classifier to allow for misclassification during the training process. Finally, SVM classifiers can be extended to more than two classes by splitting the problem into a series of binary class separations. The aim of this study is to develop vision technique to grade the mangoes by its shape and weight using visible imaging and classifying the mango using statistical classification then be applied to post-harvest handling.

Section snippets

Elements of machine vision system

The procedures and methods are implemented on machine vision workstation, conveyer show in Fig. 1, which included an illumination system Philips LED lighting, 4.5–5 watt, 6500 K color temperature and a Basler acA1600-20gc GigE camera with the Sony ICX274 CCD sensor delivers 20 frames per second at 2 MP resolutions. The RFID module is used to trigger the camera and the belt speed for the conveyer is adjustable. The conveyer that use in this study is flat bed and the purple cup is use for fruit

Shape measurement

The experiment classified the mangoes into three grades, from grade A, representing the best quality grade, to grade C. Each fruit was inspected by a human panel, who inspects the mangoes one at a time and made judgment on the quality of the fruit.

Approximately 300 fruits were sampled; the first 180 samples were used as training set and the remaining 120 samples as the testing set consist of 100 samples for each grade. In the same way, the inspectors looked at the mangoes one at a time and

Conclusion

In this paper, the algorithm have successfully been developed for the quality control inspection of Harumanis mangoes using machine vision application. Only one population which is Harumanis mango has been used in this paper, The sample was divided into 180 for training. Both the Discriminant analysis and Support Vector Machine have been investigated for shape classification of Harumanis mangoes. Training and testings have been performed using all component |F(m)|. The results indicated that

Acknowledgment

This research has been supported by the Fundamental Research Grant Scheme (FRGS) of the Ministry of Education, Malaysia entitled “Multi Modalities Sensor Fusion for Quality Assessment of Agro based product (9003-00250).

References (6)

There are more references available in the full text version of this article.

Cited by (0)

View full text