Microscopic Image based Object Detection and Classification of Indian Herbal Plant

Abstract

This report relates to the attempt in learning unusual characteristics of herbal plants like licorice, rhubarb, datura and to successfully classify them. Application of microscopic images plays an essential function for detailed detection and diagnosis of object characteristics. The main aim of this report is to address the problem of object classification by multiple features of plant root cell object. For given cell images the classifier has to decide to which class the image belongs. In this report, the classification is done based on shape and texture features of manually segmented images. The task of classification is a bit difficult for similarly characterized cell objects and because of variation in size, shape, orientation or texture occurs within classes. For classifying different classes same feature will not work properly, so to overcome this problem, the concept of using feature combination in order to distinguish each class image from many is taken into account. The currently available techniques are combined by their application and by doing an approximate description of each method, we present an applicable image processing methods in the problem of microscopic image classification. This analysis is helpful for use of the existing method and enhancing their performance as well as designing new ones. Classification performance of different classifiers is analyzed in terms of accuracy. Accuracy is measured with combination of shape features, GLCM features, Wavelet decomposition features, local binary pattern and HOG features. Accuracy of 100% is achieved for SVM classifier. Using KNN classifiers, 90% accuracy was obtained and with the use of decision trees 98% of accuracy was gained.From that, the capacity of the classifier for each object feature is obtained and from that creating a successful classification of microscopic images, the objective is to use it in area of interest like adulteration in medicine or grading and analyzing of samples.