基于自适应步长人工鱼群算法的仿生机器鱼目标检测研究

Research on Target Detection of Biomimetic Robotic Fish based onArtificial Fish Swarm Algorithm with Adaptive Step Size

摘要:

以全局视觉仿生机器鱼的视觉子系统为研究对象,为准确检测与跟踪目标(水球),提出了一种基于自适应步长人工鱼群算法的目标识别方法。该方法首先对鱼群个数进行初始化,以图像像素梯度值为目标函数,随机分布在像素矩阵中；然后计算初始鱼群中各机器鱼当前位置的食物浓度,选取其最小值作为公告板值,并将此鱼当前状态赋值公告板；在此基础上,各机器鱼分别模拟追尾行为和聚群行为,进行评价并选取目标函数值较优者为实际行为,通过不断更新公告板值,直到找到最优结果。实验结果表明该方法可减少计算复杂度,提高系统实时性,有效应用于全局视觉仿生机器鱼的目标检测与追踪。

Abstract:

Using the vision subsystem of global vision artificial fish as research object, to detect and track target(polo)accurately,an target recognition approach based on Artificial Fish Swarm Algorithm (AFSA) with adaptive step size for biomimetic robotic fish is proposed. Firstly, initialize the number of artificial fish, and randomly distributed in pixel matrix which using gradient values of image pixels as objective function. Then calculate the food concentration of every artificial fish’s current position, and choose the minimum as bulletin board value which is assignment to this fish. On this basis, by evaluating the rear-end and cluster behavior, select the optimum for the actual behavior. Update the bulletin board value until find the optimal results. The simulation experiments show that this method can simplify the complexity of calculation and improve system real-time performance. So it is feasible to be used in target detection and tracking of global vision artificial fish.