基于改进型蚁群算法和Gauss-Markov随机场的植物病斑自适应分割

针对植物病害图像成分复杂、病斑排列无规则等特点,提出了基于改进型蚁群算法和Gauss-Markov随机场的自适应病斑分割算法。该算法采用自适应信息素更新策略,对信息量进行有差别的动态更新,克服了标准蚁群算法容易陷入局部最优造成的早熟、停滞现象。同时,利用Markov随机场的局部相关特性并结合Gauss分布组成线性平稳自回归模型,针对植物病斑特征建立分割模型。最后,采用改进型蚁群算法对其进行优化,并结合Gauss-Markov随机场最大后验概率估计,实现对植物病斑的自适应分割。仿真试验表明,改进后的算法能够针对植物病斑特性实现自适应分割,鲁棒性较好。然而,对于蚁群算法与Markov的最佳耦合方式及参数初始值的设置仍需作进一步研究。

Adaptive Segmentation of Plant Disease Spot Based on Improved Ant Colony Algorithm and Markov Random Field

According to the characteristics of complex components of plant disease images and random alignment of disease spot,this paper introduced adaptive segmentation algorithm with improved ant colony algorithm and Gauss-Markov random field.By adopting self-adaptive pheromone update strategy for dynamic update of information flow differentially,this algorithm avoided the premature convergence and stagnation induced by local optimum which traditional ant colony algorithms often make.Meanwhile,for the characteristics of plant disease spot,segmentation model was created by utilizing the local correlated characteristic of Markov random field and linear stationary auto-regression model formed by Gauss distribution,and self-adaptive segmentation of plant disease spot was realized through optimizing the model with improved ant colony algorithm and maximum a posteriori probability estimation of Gauss-Markov random field.Simulation experiment showed that the improved algorithm had better robust and could realize the self-adaptive segmentation of plant disease spot.However,the optimal coupling method between ant colony algorithm and Markov and initialization of parameter values need to be further researced.