Image Retrieval Model Based On Immune Algorithm

The artificial immune algorithm (AIA) is a new kind of intelligence searching algorithm based on the biological immune theories. AIA has much ability in study, memory, and auto-adapt. It shows perfect performance in the field of searching global optimal solutions. Nowadays AIA is becoming hot in the algorithm research field, and has been widely used in the field of machine learning, fault diagnosis, network invading detection, pattern recognition etc.CBIR (Content-Based Image Retrieval) is formed based on the text-based image retrieval system. CBIR can describe the image content objectively and completely without human being's much involvement, and has a bright future promised by the researchers. Recently, CBIR is still at the beginning stage. The key of CBIR is to pick up the features of color, stripe, shape etc. in images to do similarity matching, and the problem is how to retrieve the features in image in that the semantics of the image can be described more clearly, so that the accuracy of the image retrieval can be improved remarkably.In order to describe semantic content clearly, we introduced AIA to image retrieval and designed an effective image retrieval model. This retrieval model has memory mechanism of artificial immune algorithm, and can study the retrieval demand of users in a long time which will enhanced the relationship of the image features and the semantics.The main work includes the following components.1)In the case of summarizing the principle and developing situation of the immune algorithm and analyzing the immune response type and principle, we studied and analyzed some typical immune algorithm, what's more, introduced the application of the immune algorithm in some relative fields.2) Expatiated the developing situation and retrieving steps of CBIR, studied and analyzed the technique of retrieving the features in images, the measure methods of image similarity and relative feedback types; discussed some famous CBIR search system and the main research aspect in the future.3) With the combination of AIA and CBIR, we designed a kind of new image retrieval model based on the immune algorithm. This model includes four parts: (1) presentation of image features and measurement of image similarity; (2) database of features and features cluster; (3) database of immune memory; (4) the steps of the immune algorithm in image retrieving procession.4) This image retrieval model used the HSV color model and seven fixed matrix of image areas to present image features and used the affinity of the immune algorithm to calculate the features similarity. At the same time, we combined the memory mechanism of artificial immune algorithm in the relative feedback of retrieving result, and put the relationship of the feedback information and the retrieving samples into the memory database, and then we can use the memory database to rectify the output result when we retrieve the image next time, in this way, the accuracy of the retrieving result can be improved steadily.5) Under the Windows environment, we used Visual C++ to carry on the simulation test for the above image retrieval model. The images in this experiment were from Corel image database, which included 27 subjects and 4000 pieces of images. The result of the experiment indicated that the result of the beginning retrieving operation can meet the customers' requirements very well, because of accumulating a lot of relative feedback information when we continue to retrieve the same image several times, the accuracy of the retrieving result will be improved obviously.