基于多特征融合的花卉图像检索

以植物花卉图像为研究对象，对植物花卉图像在区域分割、特征提取、基于内容的雷同图像过滤以及基于SVM的植物花卉图像检索等方面进行了系统、深入和较为全面的研究。首先为保证检索效果，提出了基于Canny边缘的雷同图片过滤算法对花卉图像库中的雷同图片进行过滤。然后提出基于2RGB混合颜色模型的自适应阂值分割算法来对花卉图像进行分割。对特征提取采用多特征融合的方法，其中针对形状特征提出了基于HSV颜色模型的加权不变矩，并提出了结合形状特征与纹理特征的边缘LBP算子。通过在花卉图像库进行的实验表明，提出的若干方法都是有效的。

Flower Image Retrieval Based on Multi-features Fusion

This paper had systematic and overall researches on flower images, including regional segmentation, feature extraction, content based duplicate images filtering and SVM-based image retrieval, etc. Firstly, in order to ensure retrieval results, we proposed duplicate images filtering algorithm based on Canny edge to detect duplicate images. Then we proposed adaptive threshold segmentation algorithm based on 2RGB mixed color model to segment flower images.Using multi-feature fusion strategy for feature extraction, we proposed weighted invariant moment shape based on HSV color model, as well as edge LBP feature which both combined texture feature and shape feature. Finally, we executed experiments on flower image library, comparative results show that above algorithms are effective.