一种基于遗传算法自学习的图像检索方法

A GA based self-learning algorithm of image retrieval

目的根据传统的相关反馈图像检索的不足,结合遗传算法的优越性,提出了基于遗传算法自学习的图像检索方法,以改进图像检索性能。方法首先通过相关反馈中的人-机交互过程,进行遗传算法的初始群体构造,再通过遗传算法进行自学习,获得满足用户语义要求的最优解。结果实验证明,该方法能够提高检索的性能,查找出更多表达用户查询意图的图像。结论给出了遗传算法在相关反馈图像检索中的应用方法。利用遗传算法自学习的过程,能够发现用户潜在的需求,改善查询结果。

Aim Traditional relevance feedback, because of complicated relationship between the visual features and the semantic concept, has limited ability to improve content based image retrieval. After analyzing that genetic algorithm can find the optimal results in multi-space quickly, it is necessary to propose a GA based self-learning al- gorithm. Methods Firstly, this approach constructs initial generation through interactive retrieval process in relevance feedback; Then according to self-learning by GA, the optimal results are obtained, which meet users need. Results The experimental result shows that the proposed approach returns more pictures which represent users semantic concept. Conclusion A method of image retrieval using GA based self-learning is presented. It can get more user＇s potential requirements and improve search greatly by using self-learning process of GA in image retrieval.