融合中值滤波与光流算法的云团跟踪算法

Research on cloud tracking by fusion of median filtering and optical flow algorithm

【目的】针对现有的云团跟踪算法存在跟踪速率低、云团滤波边缘信息丢失、云团跟踪定位不准等问题,结合优化的中值滤波算法,提出一种融合中值滤波与光流算法的云团跟踪算法。【方法】首先对太阳的位置进行定位,采用红蓝阈值比分割图像,用优化的中值滤波算法滤除噪点后,检测云块及其质心坐标;然后对云图序列进行光流处理,实现对云块的精确检测与实时跟踪;最后在不同云量情况下进行云团跟踪试验,并且与传统光流算法、块匹配算法、ViBe(visual background extractor,视图背景提取)算法进行比较。【结果】本文算法在云团边缘检测和云团实时跟踪上表现优越,目标跟踪效率比其他3种算法的平均值提高了约6.32百分点。【结论】本研究结果有效解决了云团滤波导致边缘信息缺失的问题,对预测短期太阳辐照度的准确性有重要意义。

[Objective]Aiming at the existing cloud tracking algorithms afflicted by such problems as low tracking rate,loss of cloud filter edge information,and inaccurate cloud tracking and positioning,a cloud tracking algorithm that integrates the median filtering and the optical flow algorithm was proposed in combination with the optimized median filtering algorithm.[Method]Firstly,the position of the sun was located,the image was segmented using red and blue threshold ratios,and the cloud blocks and their center-of-mass coordinates were detected after the noise was filtered out with an optimized median filtering algorithm;then,the sequence of cloud maps was processed by optical flow to achieve the accurate detection and real-time tracking of the cloud blocks;finally,cloud tracking experiments were carried out in different cloud amounts and compared with the traditional optical flow algorithm,the block matching algorithm,and the ViBe(visual background extractor)algorithm.[Result]The proposed algorithm is superior in cloud edge detection and real-time cloud tracking,and the target tracking rate is improved by about 6.32 percentage points compared with the other three algorithms.[Conclusion]The problem of missing edge information due to cloud filtering is effectively solved,which is of significance for the accuracy of short-term solar irradiance prediction.