Technology Study On The Retrieval And Recognition Of Seawater Pearl Based On The Integration Of Multiple Features

Seawater pearl have been appreciated by numerous Chinese and foreign customers for its brilliant, beautiful appearance. But the identification of seawater pearl is still at the artificial recognition stage, which is largely depended on personal experience, and it is hardly possible to finish this task by most customers. Hence, to design a system that able to automatically identify seawater pearls is particularly important.This article designed an automatic identification system for seawater pearl. First, find out the regularity on characteristics of the seawater pearl through common features on color, texture and shape; then, extract characteristics of sea pearls; lastly, fuse various characteristic parameters together to identify the seawater pearl. Feature extraction module, In view of the color feature extraction, the fusion of information entropy and block color histogram is not only can effectively extract color features, but also can avoids information redundancy caused by the average weight distribution, which will be More conducive to the accuracy calculation of color similarity; For texture feature extraction, mainly use the Grey-Level Co-occurrence Matrix to extract texture features of seawater pearl, and select best parameters for the Grey-Level Co-occurrence Matrix; For shape feature extraction, on the basis of getting shape characteristics for gray image by the invariant moment method, joint the concept of block unit entropy, put forward seven moment invariants of the entropy matrix to represent shape features of seawater pearl, experiments have proved that the algorithm is very effective. Feature fusion module, distribute weight coefficients according to various characteristic’s information entropy. Pattern recognition module, use SQL to create a sample image library for seawater pearl; Extract color, texture and shape features of pearl via corresponding feature extraction methods respectively, then cluster extracted characteristics using k-means clustering method, and build a tree index structure to facility system detection and identification, and build the sample feature database successfully; Use Matlab tool to implement system operational interface and feature extraction module, and link Matlab database through the ODBC data source; When the system identify for the pearl image, extract the characteristics of the pearl image under test Directly, and calculate similarity between pearl image and the sample characteristic library, finally, feedback the maximum similarity information to customers.