| Nowadays,the content-based video retrieval system has become a hot spot in multimedia retrieve field. Since the conventional text way of video retrieval has been unable to meet the user demand that using a unknown video screenshot or images that contains video content to retrieve videos, content-based video retrieval technology is growing fast in this demand.This thesis analysis and research the content-based video retrieval technology firstly, focusing on technologies included video frame feature extraction, shot segmentation and key frame extraction.In order to adapt to the characteristics of the video retrieval system,the thesis improves and optimizes the above-mentioned technical algorithm, and finally add in the form of plug-in interface of the improved algorithm into SSH framework. Ultimately achieve the function that users making video search online. Details are as follows:1. The thesis adds histogram barycenter algorithm on the basis of the color histogram algorithm in color feature extraction of video frame feature,and then integrates the two algorithms, using the integration of the extracted features as frame color feature.2. This thesis improves the adjacent interframe difference method for shot segmentation techniques, proposing a variable interval sampling method to deal with the shots switching of a gradient mode.3. The thesis designs and implements a video retrieval system based on SSH framework that may connect the plug-in interface of the above-described improvement algorithm to the system. When the users upload the videos or pictures to the server, the framework will call the video processing algorithms and retrieval algorithm. Meanwhile, the system design a database to store a variety of video information and finally returns the corresponding results.4.Using experiments to verify the integrity of the improved algorithm and system functions. |