基于双重随机有限蚁群建模与调控的多细胞参数估计方法研究

On the basis of two theories, i.e., random finite sets and swarm intelligence, this project aims to develop a novel theory, namely, dual random finite ant colony, which will be applied to solve the problem of micron multi-cell parameter estimation in a series of low-contrast image sequences. Through modeling multi-cell states with the random finite ant colony, we investigate the mechnism that how the pheromone field is formed in ant colony system, and a recursive estimate technique is proposed to propagate the pheromone intensity　over time to obtain the accurate and real-time estimate of multi-cell states. For each ant colony, through modeling random finite ant individuals and defining the constraint mechanism of each ant behavior, the occupied regions of multiple cells are effectively covered by interested ants and their contours of each cell are well identified, and finally the morphological parameters such as the areas and contours of all cells (including existing, newly entering and spawning) in the current image are estimated in an accurate way. As a consequencce, this project gives a feasible solution to the challenging problem of multi-cell parameter estimation under these uncertainties such as cell overlapping, splitting and spatially adjacency. It is believed that, according to our recorded all cells’ data, our research will contribute to the development of pharmacology and pathology, and provide some crucial and fundamental reference values as well.

在随机有限集理论与群智能理论基础上，本项目提出了一种双重随机有限蚁群理论，并将其应用到低对比度图像序列下的微米级多细胞参数精确估计问题中。通过对多细胞状态的随机有限蚁群建模，研究蚁群系统中信息素场形成机制，从而构建基于信息素强度信息的递推式多细胞状态实时精确估计方法；通过对单个蚁群蚂蚁个体进行随机有限建模，引入蚂蚁行为的约束机制来对细胞进行有效覆盖与边缘标识，实现对图像中所有细胞（含已存在的、新进入的、分裂的）面积、轮廓等形态特征参数的精确估计，从而最终解决在视频图像中多细胞覆盖、衍生、近邻等诸多不确定因素条件下的多细胞多参数精确估计难题，所获取的多细胞历史特征数据能为药理学和病理学的研究提供重要的基础参考价值。

在随机有限集理论与群智能理论基础上，本项目提出了一种双重随机有限蚁群理论，将其应用到低对比度图像序列下的多细胞参数精确估计问题中。针对多传感器多目标跟踪问题，提出了基于标签多贝努利随机有限集的目标（细胞）跟踪方法；在此基础上，通过对随机有限蚁群和蚂蚁个体搜索行为建模，针对蚁群系统中信息素场形成机制，研究了基于信息素强度信息的帧间递推多细胞跟踪方法；通过对侦察蚁群个体混沌至确定性行为的建模、对觅食蚁群个体决策行为建模和信息素的持续更新，研究了基于双层多贝努利随机有限蚁群多细胞状态与轮廓参数估计方法；定义了标签随机有限蚁群模型，给出了标签蚁群帧间进化模型，研究了基于标签蚁群的细胞状态及其形态学参数联合估计方法。基于上述研究，实现了在噪声、杂波、细胞分裂、黏连等诸多不确定因素条件下的多细胞自动跟踪以及细胞面积、轮廓等形态特征参数的精确估计，所获取的多细胞历史特征数据能为药理学和病理学的研究提供重要的基础参考价值。