Font Size: a A A

A Study Of Feature Extraction Technique Based On Color Image & Plant Classification

Posted on:2007-02-27Degree:DoctorType:Dissertation
Country:ChinaCandidate:Z K HuangFull Text:PDF
GTID:1118360212960451Subject:Pattern Recognition and Intelligent Systems
Abstract/Summary:PDF Full Text Request
Computer vision technology has been widely used in modem life. Highly efficient image processing algorithms determine whether computer vision could have a successful application or not. Compared with industrial image processing, the size, the shape and the color of plant are important appearance features. However, illumination usually alters quickly in plant growth environment. Plant images have poor imaging condition, complicated objects and backgrounds, and at the same time, they are not so much regular and easy description as industrial parts, which make up of a great difficulty for classifying and recognizing plant species. Human being's eyes can only identify gray image with below hundred level, but can identify thousands upon thousands colors. Particularly, color images can provide more plenty information. So the methods based on digital image processing technique, which are suitable for complex environment, able to process complex images and highly efficiently at processing color images are the targets people have been in pursuit of. This thesis investigates on effective and efficient color image processing algorithms that could process complicated images and fit to complex environment.This thesis studies the issue of feature extraction and classification from color images. Through analyzing bark texture image, Support Vector Machine (SVM) and radial basis probabilistic neural network (RBPNN) classifiers are used to perform the classification of plants. The target is to develop a computer-aided plant classification system so that an automatic plant classification system can be achieved.First of all, color space was deeply studied. We introduced all existing color space models, and analyzed the transformation relation between different color spaces. Because the color of the plant could alter in different growing environments, for example the leafage changes its color with seasons. At the same time, the changes of the environment, such as moisture, nutrient, sunlight time, can also lead to different colors which have been shown. So the features of color of image have been seldom applied to recognize plant species. Nevertheless, compared with other image features (e.g, texture features), color features are very stable, and they are robust and not sensitive to rotation, shift and size changes. Particularly, color features are easily computed. In this thesis, we try to apply the color features and the color feature...
Keywords/Search Tags:Computer Vision, Digital Image Process, Pattern Recognition, Color Image, Radial Basis Probabilistic Neural Network (RBPNN), Supporting Vector Machine (SVM), Wavelet Transform, Generalized Gaussian Model, Contourlet Wavelet, Gabor Filters
PDF Full Text Request
Related items