Automated System for Crops Recognition and Classification

Automated System for Crops Recognition and Classification

Alaa M. AlShahrani (Taif University, Saudi Arabia), Manal A. Al-Abadi (Taif University, Saudi Arabia), Areej S. Al-Malki (Taif University, Saudi Arabia), Amira S. Ashour (Taif University, Saudi Arabia & Tanta University, Egypt) and Nilanjan Dey (Techno India College of Technology, India)
Copyright: © 2017 |Pages: 16
DOI: 10.4018/978-1-5225-1022-2.ch003


Marketing profit optimization and preventing the crops' infections are a critical issue. This requires crops recognition and classification based on their characteristics and different features. The current work proposed a recognition/classification system that applied to differentiate between fresh (healthy) from rotten crops as well as to identify each crop from the other based on their common feature vectors. Consequently, image processing is employed to perform the statistical measurements of each crop. ImageJ software was employed to analyze the desired crops to extract their features. These extracted features are used for further crops recognition and classification using the Least Mean Square Error (LMSE) algorithm in Matlab. Another classification method based on Bag of Features (BoF) technique is employed to classify crops into classes, namely healthy and rotten. The experimental results are applied of databases for orange, mango, tomato and potatoes. The achieved recognition (classification) rate by using the LMSE for all datasets (healthy and rotten) has 100%. However, after adding 10%, 20%, and 30% Gaussian noise, the obtained the average recognition rates were 85%, 70%, and 25%; respectively. Moreover, the classification (healthy and rotten) using BoF achieved accuracies of 100%, 88%, 94%, and 75% for potatoes, mango, orange, and tomato; respectively. Furthermore, the classification for all the healthy datasets achieved accuracy of 88%.
Chapter Preview

Researchers are interested with applying various image processing techniques in the agriculture applications due to its accuracy. Various studies were conducted that use image processing to analyze one type of crops by extracting certain feature such as the color, size or texture. Jiménez et al. (1999) introduced a review of various vision systems to recognize fruits for automated harvesting. The survey concluded that under difficult conditions there is a feasibility of practical implementations of computer vision systems for the analysis of agricultural scenes to locate natural objects. The authors discussed the basic considerations about the distributions and characteristics of the fruits in natural orange crops.

Complete Chapter List

Search this Book: