基于PSO优化改进的Snake模型煤矿环境目标检测

摘要

针对Snake模型寻优过程中抗噪能力差、不能向凹处收敛等问题,提出适合煤矿等复杂环境的目标轮廓检测新算法。算法对Snake模型进行改进,使其自动分配蛇点,具有拓扑自适应性,并将粗收敛结果作为粒子群算法的初始轮廓。同时针对粒子群优化过程中易丧失群体多样性和易收敛于局部极值的问题结合遗传算法中育种和变异思想改进,淘汰适应度低的粒子,增加了相邻粒子间约束,通过自适应惯性权重非线性调整方法提高收敛精度。实验中将单峰、多峰测试函数和图像仿真与传统方法进行对比,证实了改进算法的有效性,在照度低、分辨率差的井下视频目标检测中有良好的应用前景。%Proposed a new algorithm for coal mine complex environment object contour detection.Solved the traditional Snake model has poor anti-noise ability and cannot converge to the concave in optimizing process.Algorithm improved Snake model to make it automatically assign the snake points with a topology adaptive.The crude convergence results as the initial outline of PSO.Particle swarm optimization process is easy to lose swarm diversity and converge to local extremum.Combined with genetic breeding and mutation thought algorithm can solve the problems above.Eliminated particles with low fitness,increased constraints between adjacent particles,improved the convergence accuracy through nonlinear inertia weight adaptive adjustment method.Experiment compared the unimodal,multimodal functions and simulated images with traditional methods,confirmed the effectiveness of the improved algorithm.It has a good prospect in object detection under low illumination and poor resolution coal mine environment.