基于多元线性回归的昆虫图像分割方法

INSECT IMAGE SEGMENTATION ALGORITHM BASED ON MULTIPLE LINEAR REGRESSION

昆虫图像自动鉴定是一种快速鉴定昆虫的方法,图像分割则是其中的关键步骤。为了使昆虫图像分割达到更理想的效果,提出基于RGB三色板的多元线性回归算法对昆虫图像进行分割。算法的核心是首先建立图像的多元线性回归模型,根据RGB三色板的背景样本确定回归模型参数,并采用范数对回归模型进行优化,优化后的算法有效解决了昆虫图像分割中部分区域分割不理想的问题。实验证明,提出的算法对昆虫图像分割具有良好的应用价值。

Automatic identification of insect images is a quick approach for identifying the insects, and in it the image segmentation is the key process. In order to make the insect image segmentation reach a more ideal effect, we propose an algorithm to segment the insect images based on multiple linear regression algorithm of RGB colour palette. The core of the algorithm is to establish the multiple linear regression model of images first. The parameters of regression model are decided according to the background swatches of RGB colour palette, and by using the norm the regression model is optimised. The optimised algorithm is effective in solving the problem that the segmentation of part of region is not ideal in insect image segmentation. Experiments verify that the proposed algorithm has good application value on insect image segmentation.