JSEG Algorithm and Statistical ANN Image Segmentation Techniques for Natural Scenes

Abstract

Computer vision, when used in open and unstructured environments as in the inspection of crops for natural scenes, demands and requires complex analysis of image processing and segmentation algorithms, since these computational methods evaluate and predict environment physical characteristics, such as color elements, complex objects composition, shadows, brightness and inhomogeneous region colors for texture. Several segmentation algorithms proposed in literature were designed to process images originally characterized by the above-mentioned items. Additionally, agricultural automation may take advantage of computer vision resources, which can be applied to a number of different tasks, such as crops inspection, classification of fruits and plants, estimated production, automated collection and guidance of autonomous machines. Bearing the afore-named in mind, the present chapter aims the use of JSEG unsupervised segmentation algorithm (Deng et al., 1999a), Statistical Pattern Recognition and Artificial Neural Networks (ANN) Multilayer Perceptron (MLP) topology (Haykin, 2008) as merging processing techniques in order to segment and therefore classify images into predetermined classes (e.g. navigable area, planting area, fruits, plants and general crops). The intended approach to segment classification deploys a customized MLP topology to classify and characterize the segments, which deals with a supervised learning by error correction – propagation of pattern inputs with changes in synaptic weights in a cyclic processing, with accurate recognition as well as easy parameter adjustment, as an enhancement of iRPROP algorithm (improved resilient back-propagation) (Igel and Husken, 2003) derived from Backpropagation algorithm, which has a faster identification mapping process, that verifies what region maps have similar matches through the explored environment. To carry through this task, a feature vector is necessary for color channels histograms (layers of primary color in a digital image with a counting graph that measures how many pixels are at each level between black and white). After training process, the mean squared error (MSE), denotes the best results achieved by segment classification to create the image-class map, which represents the segments into distinct feature vectors. Several metrics (vector bundle) can be part of a feature vector, however, a subset of those which describes and evaluates appropriate classes of segments should be chosen.