Content-Based Image Retrieval System for Silk Pattern Information on Mobile Phone

Abstract

In this research, we present a content-based image retrieval system on the mobile phone for searching silk pattern information in Sisaket province. The system comprises mobile application, web application and pattern recognition application. The mobile application is developed based on Apache Cordova framework and used for taking a suspicious silk pattern by built-in camera to be an image, sending the image to server and displaying search results. Next, the web application is developed by PHP language and applied for receiving and saving the image as a temporary file, calling pattern recognition application and returning search results as web service in the form of REST format. Finally, the pattern recognition application is developed by C# language cooperate with OpenCV framework and it comprises speed-up robust feature algorithm for searching interesting points and k-nearest neighbor algorithm for evaluating similarity distance with the sample. The performance for classification of the proposed system is evaluated through confusion matrix. The result shows that the average accuracy is 88.34 %