Lin Ciyun, Sun Wenjian, Sun Ganghao, Gong Bown, Liu Hongchao
Department of Traffic Information and Control Engineering, Jilin University, Changchun 130022, China.
Jilin Engineering Research Center for Intelligent Transportation System, Changchun 130022, China.
Sensors (Basel). 2025 Jun 21;25(13):3875. doi: 10.3390/s25133875.
Pavement markings, as a crucial component of traffic guidance and safety facilities, are subject to degradation and abrasion after a period of service. To ensure traffic safety, retroreflectivity and diffuse illumination should be above the minimum thresholds and required to undergo inspection periodically. Therefore, an onboard light detection and ranging (LiDAR) and camera deployment optimization method is proposed for pavement marking distress detection to adapt to complex traffic conditions, such as shadows and changing light. First, LiDAR and camera sensors' detection capability was assessed based on the sensors' built-in features. Then, the LiDAR-camera deployment problem was mathematically formulated for pavement marking distress fusion detection. Finally, an improved red fox optimization (RFO) algorithm was developed to solve the deployment optimization problem by incorporating a multi-dimensional trap mechanism and an improved prey position update strategy. The experimental results illustrate that the proposed method achieves 5217 LiDAR points, which fall on a 0.58 m pavement marking per data frame for distress fusion detection, with a relative error of less than 7% between the mathematical calculation and the field test measurements. This empirical accuracy underscores the proposed method's robustness in real-world scenarios, effectively mitigating the challenges posed by environmental interference.
路面标线作为交通引导和安全设施的关键组成部分，在使用一段时间后会出现退化和磨损。为确保交通安全，反光性和漫射照明应高于最低阈值，并需定期进行检查。因此，提出了一种车载激光雷达（LiDAR）和摄像头部署优化方法，用于路面标线病害检测，以适应复杂的交通状况，如阴影和光线变化。首先，基于传感器的内置特性评估了LiDAR和摄像头传感器的检测能力。然后，针对路面标线病害融合检测，对LiDAR-摄像头部署问题进行了数学建模。最后，通过结合多维陷阱机制和改进的猎物位置更新策略，开发了一种改进的红狐优化（RFO）算法来解决部署优化问题。实验结果表明，所提出的方法在病害融合检测中，每数据帧有5217个LiDAR点落在0.58米的路面标线上，数学计算与现场测试测量之间的相对误差小于7%。这一经验精度凸显了所提方法在实际场景中的鲁棒性，有效缓解了环境干扰带来的挑战。
