论文部分内容阅读
基于内容的遥感影像检索已经成为遥感领域的研究热点,因此,本文提出了一种综合视觉词袋模型和颜色直方图的遥感影像检索方法,利用尺度不变特征算子提取影像的局部不变特征,通过视觉词袋模型组合局部特征,生成每幅影像的金字塔直方图,接着结合每幅影像的颜色直方图生成更有区分性的特征向量,利用新的特征向量集训练支持向量机分类器,通过分类器输出与查询属于一类的影像,完成遥感影像检索。试验结果表明,本文方法不仅提高了影像检索的查准率和查全率,并且验证了该方法能有效克服影像光照、噪声、方向等变化,鲁棒性较好。
Content-based remote sensing image retrieval has become a research hotspot in the field of remote sensing. Therefore, this paper presents a comprehensive visual bag-of-vision model and color histogram remote sensing image retrieval method, using the scale-invariant feature operator to extract local invariant features of the image , The pyramidal histogram of each image is generated by combining the local features of the visual word bag model, then the more discriminative feature vectors are generated according to the color histogram of each image, the SVM classifier is trained by the new feature vector set, Through the classifier output and query belongs to a class of images, complete remote sensing image retrieval. The experimental results show that this method not only improves the accuracy and recall of the image retrieval, but also verifies that this method can effectively overcome the changes of image illumination, noise and direction, and has better robustness.