摘要
对巡检船主动避障算法进行研究,提出一种方向抑制的GOA(gradient of averages,平均梯度)改进算法,解决了传统直线段提取算法计算量大的问题.采用抑制横向纹理的GOA算法强化障碍物特有的竖(斜)向纹理,利用特殊形态学模板进行去噪,并结合时域滤波处理连通分析结果,完成目标特征检测.通过与LSD(Line Segment detection)算法进行实验对比,结果表明:改进算法边缘特征提取完整,计算复杂度低.针对TI公司的Davinci芯片进行算法核心函数优化,满足实时性要求.
By studying the algorithm of patrol boat automatic obstacle avoidance,an improved GOA（gradient of averages）algorithm based on the direction suppression was proposed,which had lower computational complexity than traditional line extraction algorithm.Using the GOA algorithm to suppress transverse texture and enhance the unique vertical（oblique）texture of obstacles,and utilize the special morphological template for noise rejection,the detection of objective characteristics has been accomplished together with the result of the time-domain filtering connectivity analysis.Compared with the LSD algorithm experimentally,the GOA algorithm can significantly improve the integrality of edge extraction,and reduce the computational complexity.On Davinci chip of TI the core function of the algorithm was optimized,and meet real-time requirements.
出处
《河北大学学报（自然科学版）》
CAS
北大核心
2014年第4期434-438,共5页
Journal of Hebei University(Natural Science Edition)
基金
河北省教育厅科学技术研究计划项目(Z2011252
Z2012126)
