摘要
复杂动态路表和室外光照条件下,鲁棒精确地识别图像中的激光线是激光结构光在公路平整度测量应用中亟需解决的一个基础性关键问题。针对公路表面的强反光特性地标线和复杂光照的条件,文章提出了一种基于刚性项约束主动轮廓模型的复杂路面激光线鲁棒识别方法。采用三次样条曲线进行更为精细的激光结构光光条曲线初始拟合,以降低主动轮廓模型对于初始值的敏感性,并在能量泛函中加入了以曲线弧长二阶导数表示的刚性项,保证曲线光滑演化的同时能够有效抑制曲线拉伸。在有效性验证实验中,不改变算法参数的情况下,分别选取了包含车辆阴影、白色标志线以及平整度极差路面共3种主要扰动因素的路面实际图像,与已有算法进行对比测试,实验结果表明该文方法具有良好的识别精度和鲁棒性。
Under the unstructured outdoor lighting environment, robust recognition and accurate inter- polation of structured laser light were still challenging in the application of road flatness measure- ment. For obtaining more robustness to the disturbance of shadow and landmarks, a laser line recog- nition method was proposed by utilizing rigid item constraint active contour model. To reduce the sen- sitivity of the initialization of the active contour model, the cubic spline algorithm was employed for more accurate curve interpolation. A rigid item constraint described by the second derivative of arc length was appended to the energy functional definition to ensure the smooth curve evolution and ef- fectively restrain the curve stretch. With fixed parameters, the experimental results of real images that included the extreme disturbances of shadows, white indicator line and poor road flatness valida- ted the efficiency and robustness of the proposed method.
作者
杜英魁
刘成
田丹
韩晓微
原忠虎
DU Yingkui LIU Cheng TIAN Dan HAN Xiaowei YUAN Zhonghu(School of Information Engineering, Shenyang University, Shenyang 110044, China Shenyang Institute of Automation, Chinese A eademy of Sciences, Shenyang 110016, China)
出处
《合肥工业大学学报（自然科学版）》
CAS
北大核心
2017年第3期321-326,共6页
Journal of Hefei University of Technology：Natural Science
基金
国家自然科学基金面上资助项目(71672117)
辽宁省自然科学基金资助项目(2015020158
2015020037)
机器人学国家重点实验室开放基金资助项目(2015008)
辽宁省高等学校创新团队资助项目(LT2013024)
辽宁省博士启动基金资助项目(201601213)
沈阳市社会发展科技攻关资助项目(F16-155-9-00)
关键词
平整度
激光结构光
各向异性扩散
三次样条
主动轮廓模型
flatness
structured laser light
anisotropic diffusion
cubic spline
active contour model
