花朵图像自动提取方法研究

Research on image auto extraction method of flower

采用改进的主动轮廓模型算法,实现从蝴蝶兰花簇图像中提取单个花朵图像,为自动识别蝴蝶兰的生长状态奠定基础。首先利用改进的骨架算法和轮廓重生算法,生成蝴蝶兰花簇的初始轮廓;然后利用含有形状能量的主动轮廓模型进行轮廓的演化,使其更接近真实的蝴蝶兰花簇边缘,最后根据蝴蝶兰花蕊位置获得相应的蝴蝶兰花朵。实例验证和比对实验结果表明,该模型能够较好的分割和提取蝴蝶兰花簇中单个花朵,并具有较强的抗噪能力。利用该方法,可以较好的提取蝴蝶兰花簇中的单个花朵,与人工提取效果进行比对,正确率达到了91.5%。

Improved active contour method is used to extract single flower object from phalaenopsis amabilis cluster, which lays good foundation to identify the growth state of phalaenopsis amabilis automatically. It firstly uses the improved skeleton algorithm and contour repossession algorithm to create preliminary contour, and then makes contour evolution by active contour model with shape energy to make it close to real object edge. Finally, the corresponding flower will be obtained according to the position of pistil/stamen. Both the instance validation and the comparison experiment prove the model can segment and extract single flower object from phalaenopsis amabilis cluster very well, with good anti-noise capacity. The algorithm model put forward in the article can extract single flower object from phalaenopsis amabilis cluster with a accuracy rate as high as 91.5%, compared with artificial extraction method.