基于先验知识水平集方法的草莓图像分割

在实际应用中,当目标本身含有一些固有的颜色纹理特征时,可将这些特征作为一种先验信息,这样可以大大提高分割的准确性.为此,本文提出了一种基于先验信息的改进水平集图像分割方法.首先,利用传统的C-V模型能量项的构造思想构建了基于颜色信息的局部能量项,该项是用于处理彩色图像;然后将颜色分量引入到传统的结构张量中构建出新的扩展型结构张量,该项是用于处理纹理信息;最后,将上述新构造的能量项以及Li模型约束项引入到传统C-V模型中得到新的水平集模型.鉴于草莓果实所具有的颜色信息和纹理信息,本文将上述改进水平集方法应用到农业自动化应用中草莓果实分割中.对实验室环境与草莓生长环境下的草莓图像进行分别实验,结果显示该方法能够不仅能够分割出草莓果实且能够很好地处理草莓表面的纹理信息.另还与OTSU算法、传统C-V模型、改进C-V模型对草莓图像作对比实验,结果表明本文算法均比上述三种算法具有更好的分割效果.

Strawberry Image Segmentation Based on Level Set Method

In practical applications, when the object itself contains some inherent color or texture features, these features can be used as a priori information, which can greatly improve the accuracy of the segmentation. Therefore, this paper proposes an improved level set method which is based on prior information. Firstly, we use the thought of the energy structure of C-V model to construct an energy function that contains of color information to segment color image. Then, we put the color component into the traditional structure tensor for the texture image segmentation. Finally, we get the new level set model from the traditional C-V model with the new energy structure and the penalty term of Li model. In view of the color and the texture information of strawberry fruit, the improved level set method has been applied to the segmentation of fruit image in agriculture. By experiment on laboratory and nature, the result shows that the improved can not only segment out strawberry, but also segment texture on the surface of strawberry. Comparing with the OTSU algorithm, the traditional C-V model and improved C-V model, the experimental results show that the proposed method has better segmentation result.