基于改进水平集的菌落图像智能计数算法

Intelligent Counting Algorithm for Colony Image Based on Improved Level Set

针对现有的菌落自动识别计数方法对背景敏感、对多菌种菌落分割普适性差的缺点,提出一种基于改进水平集的全自动菌落分割、计数方法.该方法利用偏置场对背景进行建模以消除背景灰度不均影响;构造含有终止条件的多相水平集算法实现菌落目标的自适应分割;通过极坐标空间中凹点检测实现粘连目标计数修正.由天津市食品安全检测技术研究院提供300例多菌种混杂菌落样本做为测试集,以专家人工鉴定结果为金标准,将本方法与传统形态学方法、迅数icount10两种定量方法进行对比,菌落密度在300 CFU内时,本方法计数准确率达到92.7%,对多菌种混杂菌落的计数精度、分割效果都优于其他两种方法.

The current colony counting methods are sensitive to background and are inefficient from the viewpoint of the universality of multi-species colony segmentation. To overcome this limitation,an automatic colony segmentation and counting method based on an improved level set is proposed. Here in,the bias field is used to model the background to eliminate the influence of intensity inhomogeneity. A multiphase level set algorithm with termination condition is constructed to realize adaptive segmentation of colony target. Through the concave point detection in polar coordinates, the correction of adhesion targets counting is achieved. The algorithm is tested on 300 samples of multispecies colonies provided by Tianjin Food Safety Inspection Technology Research Institute. The algorithm is compared with two quantitative methods:traditional morphological algorithm and Shineso icount10 algorithm,and the results of expert artificial identification are used as gold standards. When the colony density is under 300 CFU,the counting accuracy rate is up to 92.7%. Both the counting accuracy and segmentation effect of multi-species colonies using the proposed method are superior to those of the other two algorithms.