基于改进袋獾算法的破片图像多阈值分割

Multi-threshold segmentation of fragment sequence images based on a modified tasmanian devil algorithm

针对静爆场中所拍摄图片存在破片目标小、背景复杂的问题,提出一种基于改进袋獾算法的破片图像多阈值分割。基于Tent混沌映射初始化种群,加入自适应权重策略以提高算法全局搜索能力,结合精英反向学习策略来避免算法过早陷入局部最优。基于优化袋獾算法求解Tsallis相对熵的最小值,作为目标函数值计算最佳阈值对破片图像进行目标分割。仿真结果表明,ITDO在12类基准函数上相较于其他算法表现出更强的收敛性和稳定性。ITDO-Tsallis算法与其他两种先进算法相比,收敛时间更快、求解目标更精准,说明该算法能有效解决静爆场破片图像目标分割问题。

To tackle the challenge of segmenting small debris targets against complex backgrounds in static explosion imagery,we've refined a multi-threshold segmentation technique using a tasmanian devil algorithm.This method leverages Tent chaos mapping for population initialization and an adaptive weight strategy to bolster global search efficiency.It also integrates an elite reverse learning strategy to evade local optima traps.Using ITDO to solve for the minimum value of Tsallis relative entropy as the target function value to calculate the optimal threshold for debris image segmentation.Simulations reveal that the ITDO algorithm outperforms others in convergence and stability across 12 benchmarks.The ITDO-Tsallis algorithm,notably,offers swifter convergence and more precise target resolution than its counterparts,proving its efficacy in debris image segmentation within static explosion fields.