Wood Species Retrieval Based On The Similarity Matching Of Digital Images

With the coming of the information age and the rapid development of digital technology, the traditional keyword-based wood retrieval method can no longer meet the research and production for species identification retrieval requirements. Therefore, retrieval technology based on digital image processing as one of the key technologies to solving this problem have the potential to be applied to the field of wood science.This study chooses 120 representative species in China, each specie used two section images of tangential and radial as the object of study. Extract digital image color, histogram and texture three feature of wood species by using MATLAB, use the matching algorithm for comparing the image features of wood species with wood image measurement data in the database to find a group of images most similar to the being retrieval image , according to the close similarity shows that rank. Users can search for the closest surface image to the being retrieval image through the retrieval software, judge the retrievaled image if the retrievaled image is same to type of being retrieval image, to achieve wood species matching retrieval, and in certain degree achieve identification retrieval purposes for species.Research and experiment results:(1) Feature Extraction: Extract the hue, saturation, value and color moments as color feature parameters and the 16 gray feature of statistics histogram, the cumulative histogram of gray-scale image. On the basis of analysising color features of wood images and proposed color model for wood, and the HSV non-quantitative color space for wood. Experiments show that, HSV color space is a better retrieval performance, combined with feature extraction method of color histogram, the experiment has produced good results. Analysising texture algorithm and texture features of the wood extracted contrast, angular second moment, sum of variances, long run emphasis, fractal dimension, and wavelet horizontal energy proportion, give expression from different angles of wood texture features.(2) Pattern matching: on the basis of analysising more than 10 kinds of color and texture feature, proposed retrieval feature extraction methods and a combination of matching algorithm for wood. Used the distance, absolute distance matching method for histogram features matching. Similarity matching of feature vector absolute distance method on texture features is effective.(3) Through the method of MATLAB / VB / SQL mixed programming, overcame the limitations of interface design by MATLAB, used advantages of VB designed friendly interface. In the process of retrieva, used single-feature and multi-feature. Among them, step-by-step search method is used as multi-feature, you can choose in results database of color feature retrieval for a second search. Experimental analysis of existing color, texture extraction and matching algorithm, color feature images of wood preliminary search can be more effective detection retrieval wood species corresponding images, use texture features can be detected further, achieve to be more precise search results, and experimental results show that the effectiveness of the method.(4) By effective confirming the features of the color, texture, similarity matching methods, and exploring step-by-step search method, the establishing for the database of wood images feature, and other steps of the study, by mixed programming construct Retrieval System. The system effectively deal with matching retrieval for the wood species and has features of highspeed, more accurate and easy-to-use, achieve the expected goals of the study.