Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Identification of citrus greening disease using a visible band image analysis

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan

Citation:  Paper number  131591910,  2013 Kansas City, Missouri, July 21 - July 24, 2013. (doi: @2013
Authors:   Alireza Pourreza, Won Suk (Daniel) Lee, Eran Raveh, Youngki Hong, Hyuck-Joo Kim
Keywords:   Classification HLB Image Processing Polarized Light Textural Features.

Abstract. Huanglongbing (HLB) or citrus greening is an extremely severe disease in citrus trees which is incurable and causing a huge loss in the citrus industry in Florida. However, early detection and removal of the infected canopies may decrease spreading of the disease and avoid an enormous loss. The disease symptoms are not clear in the early stages of infection. Subjective disease detection methods such as ground scouting and other objective means are also either inaccurate or costly and time consuming. This paper introduced an easy, inexpensive, fast, and accurate method of HLB detection which is more applicable and affordable for citrus growers. A customized image acquisition system was developed to acquire images of citrus leaves (Valencia) at a waveband of 591 nm. Polarized filters were used in both illumination and imaging system to highlight the HLB disease symptoms. Several types of textural features were extracted from the leaf images and the best sets of features which could describe the infection characteristics were ranked using five different feature selection methods. The performances of seven classifiers were evaluated in a step by step classification approach. Since HLB symptoms are similar to some nutrient deficiencies, magnesium and zinc deficient samples were also included in the classification process. Healthy and HLB symptomatic samples were identified with an accuracy of 100%; however, some of the nutrient deficient samples were misclassified into other classes using this method. The overall accuracies of 86.5% and 89.6% were achieved in five-class identification and two-class (healthy or HLB) detection, respectively.

(Download PDF)    (Export to EndNotes)