Identification of Cocoa Pods with Image Processing and Artificial Neural Networks

Abstract

Cocoa pods harvest is a process where peasant makes use of his experience to select the ripe fruit. During harvest, the color of the pods is a ripening indicator and is related to the quality of the cocoa bean. This paper proposes an algorithm capable of identifying ripe cocoa pods through the processing of images and artificial neural networks. The input image pass in a sequence of filters and morphological transformations to obtain the features of objects present in the image. From these features, the artificial neural network identifies ripe pods. The neural network is trained using the scaled conjugate gradient method. The proposed algorithm, developed in MATLAB ®, obtained a 91% of assertiveness in the identification of the pods. Features used to identify the pods were not affected by the capture distance of the image. The criterion for selecting pods can be modified to get similar samples with each other. For correct identification of the pods, it is necessary to take care of illumination and shadows in the images. In the same way, for accurate discrimination, the morphology of the pod was important.