Intelligent Counting System For Composite Colony Based On Improved Level Set

Colony counting is one of the important tasks for quality testing in food,health care,environment and other industries.The existing automatic colony counting systems still have some unsolved problems.For instance,the instrument is too large and expensive and the software has low accessibility.At the same time,the embedded colony detection algorithm is sensitive to the background shadow of culture dish,poor in the universality of multi-species colony segmentation,lacking in segmentation ability of adhesive colony,impossible to realize adaptive segmentation with good intelligence.In order to solve these problems and improve the efficiency of colony counting,this paper has designed a set of intelligent colony counting system.The main works are listed as follows:In the part of system hardware,the CCD sensor is selected as the photosensitive element to improve the image clarity.The upper and lower projection light source and the optical darkroom structure are added to improve the uniformity of illumination.And the constant current circuit is designed to realize the linear light intensity so as to ensure effective light compensation under different light environments.In the PC version software,the utilitarian functions like selection of multi-mode counting area,information statistics of single colony target and differential counting of colonies with different size simplify the operation process of the first-line employees.In the critical part of image processing algorithm,an automatic colony segmentation and counting method based on improved level set has been proposed in this paper.The bias field is used to model the background to eliminate the influence of intensity inhomogeneity.The 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 counting correction of adhesive targets is achieved.Tianjin Food Safety Inspection Technology Research Institute provided 300 samples of multi-species colonies as the test set and the gold standard is artificially made by experts.Firstly,the algorithm is compared with two segmentation methods,level set algorithm of improved C-V model and colony analysis algorithm of Shineso Icount10,our method has obvious advantages in both qualitative and quantitative aspects.Secondly,the counting result of the proposed algorithm is compared with other two quantitative methods traditional morphological algorithm and Shineso Icount10 colony counting algorithm in the range of four different colony densities(0-100,100-150,150-200,200-300).The average counting accuracy rate of the proposed method is up to 92.7%.The results of experiment show that both counting accuracy and segmentation effect of multi-species colonies are superior to other current algorithms.Briefly,compared with the existing colony counting system,the system designed in this paper owns many advantages and has broad application scenarios in the fields of medical inspection and environmental monitoring.The system has been delivered to the cooperative company for subsequent industrialization.