Research On Multi-Cell Contour Estimation And Tracking Using Ant System

As computer vision technology develops rapidly,multi-object tracking technology has been successfully applied in many fields,especially in the fields of intelligent transportation systems,video monitoring and biological pharmacy.In the field of biological pharmacy,the image analysis software CellProfiler can help biologists analyze complex morphological features such as cell number,size,contour and so on.In recent years,although some progress has been made in related field,there are still many theoretical and technical breakthroughs to be made.For example,the phenomenon of cell division,rotating,and adhesion,as well as the randomness of cell movement in low signal-to-noise ratio and intensive situations have brought great challenges to cell tracking.To solve the above problems,this paper mainly focus on the method of multi-cell contour estimation and tracking based on contour pheromone field and the multi-Bernoulli random finite ant colony for multi-cell contour estimation and tracking.In addition,a multi-cell contour tracking software system is designed.As a typical representative of swarm intelligence method,ant colony algorithm(ACO)has gradually opened a new world for the research of artificial intelligence with its powerful search function.A multi-cell contour estimation and tracking method based on contour pheromone field prediction and update is proposed,this can accelerate the construction of contour pheromone field in the current frame.Considering the predicted position of the contour pheromone field may has greatly deviated,the direction selection probability model is designed in the process of ant colony decision.In addition,the pheromone release and diffusion model are designed based on the gray gradient to help the accumulation of pheromone on the contour.Experiment results indicate that the method has higher accuracy than other cell contour tracking algorithms.Since the random finite set(RFS)has been successfully applied in the field of multi-object tracking,it has been favored by multi-object tracking.This paper combines RFS with ant colony algorithm and proposes a multi-cell contour estimation and tracking method based on multi-Bernoulli random finite ant colony.This method using the contour pheromone field and heuristic function to calculate the existence probability and probability density distribution of the corresponding Bernoulli,and the existence probability of the cell corresponding to the position of the ant colony is judged by the existence probability of Bernoulli,the contour and state of the cell are extracted by the contour pheromone field corresponding to the ant colony with high probability of existence.Experiment results indicate that the method has higher tracking performance.Finally,a multi-cell contour tracking software system is constructed,it includes two ant colony-based multi-cell contour tracking methods and a mixed threshold-based detection and tracking method which is based on a series of morphological operations such as grayscale stretching,expansion,filtering,and watershed transformation.The system can estimate and track two specified sequences,store the tracking results in the form of an excle report,analyze and compare the performance of these methods.