基于BANSAC的水下抗环境光光斑定位

Underwater Light Spot Localization Resistant to Ambient Light Based on BANSAC

摘要:

信噪比是衡量水下无线LED光通信系统性能的一个重要参数,为了使通信系统获得最大信噪比,需要通过光路对准跟踪来使接收端接收到最大光功率。而光斑中心定位精度会直接影响通信链路的对准跟踪环节的实施难度。因此,为克服水下环境光导致接收端难以准确且快速定位光斑中心坐标的困难,研究了光斑偏移量对通信链路对准跟踪性能的影响,并在此基础上提出了一种基于BANSAC圆拟合算法的光斑中心定位方法。首先对光斑图像进行预处理以提升光斑边缘特征的准确性和完整性。随后,对获得的边缘信息应用BANSAC算法生成候选圆模型,并将拟合度最高的模型圆心作为最终的光斑中心位置。实验结果表明,与传统的光斑中心定位方法相比,BANSAC圆拟合算法有效降低了环境光对光斑中心定位的影响,测量误差在x方向约为0.343pixel,在y方向约为0.279pixel,平均定位时间约为45ms。相较于传统光斑定位方法，尽管执行效率存在轻微降低，但其定位精度得到了明显优化。

Abstract:

Signal-to-noise ratio (SNR) is a crucial parameter for evaluating the performance of underwater wireless LED optical communication systems. To ensure the communication system achieves the maximum SNR, it is necessary to implement optical path alignment and tracking so that the receiver can capture the maximum optical power. The positioning accuracy of the light spot center directly affects the difficulty of implementing the alignment and tracking process of the communication link. Therefore, to overcome the challenge that ambient light in underwater environments makes it difficult for the receiver to accurately and quickly locate the coordinates of the light spot center, this study first investigates the impact of light spot displacement on the alignment and tracking performance of the communication link. On this basis, a light spot center localization method based on the BANSAC circle fitting algorithm is proposed. Firstly, preprocessing is performed on the light spot image to improve the accuracy and integrity of the light spot edge features. Subsequently, the BANSAC algorithm is applied to the obtained edge information to generate candidate circle models, and the center of the model with the highest fitting degree is taken as the final position of the light spot center. Experimental results show that compared with traditional light spot center localization methods, the BANSAC circle fitting algorithm effectively reduces the influence of ambient light on light spot center localization. The measurement error is approximately 0.343pixel in the x-direction and 0.279pixel in the y-direction, with an average localization time of about 45ms . Compared with traditional spot localization methods, though the execution efficiency is slightly compromised, the positioning accuracy has been significantly optimized.