عنوان البحث(Papers / Research Title)
Object Classification By Neural Networks
الناشر \ المحرر \ الكاتب (Author / Editor / Publisher)
توفيق عبد الخالق عباس الاسدي
Citation Information
توفيق,عبد,الخالق,عباس,الاسدي ,Object Classification By Neural Networks , Time 5/22/2011 6:46:25 AM : كلية تكنولوجيا المعلومات
وصف الابستركت (Abstract)
Object Classification By Neural Networks
الوصف الكامل (Full Abstract)
Object Classification using a neural networks with Gray-level Co-occurrence Matrices (GLCM)
Dr. Tawfiq A. Al-Assadi Department of Computer Science Babylon University
Mehdi Ebady Manaa Department of Computer Science Babylon University
Nawfal Turki Obies Department of Computer Science Babylon University
Abstract:
This paper describes a hybrid method in the object classification for computer digital images. method in this paper has been designed and developed to recognize a typical texture features for certain object. The basic approach used here is that the textures features values that extracted from gray level co-occurrence matrices (GLCM) can show the typical values for features analysis in classification. An artificial neural networks using error multilayer back propagation network has been used for texture analysis and object classification. The obtained results of different types of images areas like "seas" "non-seas" and "background" as unknown images was characterized in a good range.Keyword: Object Classification, Gray-Level Co-occurrence Matrix(GLCM), Texture features.
Introduction:
This paper introduces a new approach of object classification for certain type which is a part of image processing using Gray level co-occurrence matrix (GLCM) as an example the class of seas images has been used for classification and the approach can be applied for single class image in different patterns. The simplest way practice in this paper is a classification of any class images into patterns using adaptive segmentation with the use of their textures features in different direction of GLCM matrix to train the artificial neural networks (back propagation neural network used here). This association between local trained features values and recognized class sea as an example led to obtain a good results using this method [1]. Another direction in this paper is extracting the texture feature for unknown image and let the neural detect the type of this image using a neural network and the approach applied for variety images. Patterns features may be applied for realizing a wide pattern in different texture without imposing any restriction on their distribution [2]. Based on a topicality of the given approaches this paper present the texture segmentation for a new approach by using Gray level Co-occurrence Matrix (GLCM) [3]. Machine classification vision based on gray level co-occurrence matrix of class classification is necessary and important method. As it lead to obtain a good results for classification in certain features characteristics. There are different approaches in computer vision, Angelo Zizzari and Udo Seiffert used detection of tumor in digital images of the Brain[4]. Commander K.Velu [5] used an Intelligent Segmentation of Industrial Component Images. Virendra Pathak and Onkar Dikshit [6] used Segment based classification of Indian urban environment. P. Tymkow, A. Borkowski [7] introduced land cover classification using airborne laser scanning data and photographs. When we propose a class classification based on Gray level co-occurrence matrix (GLCM) with a neural network to recognize a certain class, it is necessary to allocate following points. a. Choice the patterns of textures attribute for a large numbers of variety Images. b. adaptive Image segmentation for the input image. c. Texture Features extraction using GLCM Matrix in different Direction. d. Train a neural network on different patterns for certain class and the seas patterns used here as an example. d. Test unknown image by calculate the texture features by GLCM and used a neural network to detect it.
Dear visitor,For downloading the full version of the research/article click on the pdf icon above.
تحميل الملف المرفق Download Attached File
|
|