Objectives: To develop an algorithm to classify the fruit damage caused by the Aonidiella aurantii pest in a Citrus × tangelo crop. With this detection and prediction is intended to provide a tool for estimation of pesticide application doses on the affected plants. Methods/Statistical Analysis: A system based on digital image processing was implemented, to recognize the fruits that pass through a conveyor belt, thus to select the quality fruits. As of Digital techniques as 2D Fourier transform, for calculate the intensity changes on the image, digital filtering, Circle Hough transform as method to detect circular contours and HSB color model related to represent the colors based on three parameters: Hue, Saturation and Brightness. Findings: The algorithm was capable to identify and classify Citrus × Tangelo fruits, using as parameter the recognized affected surface. After the image segmentation, a new mask was applied to recognize the free A. aurantii area of the fruit, as of second histogram information, was possible to obtain the healthy area of the fruit. One of the factors to obtain a feasible identification of the health areas on the fruits was the quantity and quality was the light in the work environment, due to a degradation of these factors could lead to a bad interpretation of the data when reflections or shadows are present. Application/Improvements: The main contribution of this system is the implementation of artificial vision methods for detecting affected areas on the surface of the fruit and determines its quality in order to accomplish the task of fruit selection. Also this work contributes to the development of agro industry, proposing new non-invasive methodologies to improve the production process in the agricultural sector in Colombia.
Read full abstract