激光雷达作为大气主动遥感探测的重要手段之一，被广泛用于气溶胶和云层探测。基于简单多尺度层检测算法，通过添加尺度个数阈值，提出了一种在时均可提高云底反演高度准确性的云检测算法。该算法通过仿真云层内部多峰时的回波信号，得到了最优化尺度个数阈值范围；同时通过添加特征段合并有效减少了单层多峰情况下的漏判。分别使用微分零交叉法、简单多尺度法和改进后的算法对532 nm米通道激光雷达在2019年5~7月探测的20组探测时长75 min以上、结构稳定、时序上无突变的有效数据进行检测，并以微分零交叉法的云底高度为准，通过改进算法得到的均方根误差平均减小32.65%，不确定度平均减小33.80%，验证了改进算法在提高云底高度准确度上的有效性。

As one of the important means of active remote sensing of atmosphere, lidar is widely used in aerosol and cloud detection. Based on the simple multiscale algorithm for layer detection, a cloud detection algorithm was proposed that could improve the accuracy of the cloud-base height when the SNR was greater than 5 by adding the threshold of the number of scales. By simulating the echo signal of multi-peak inside the cloud, the algorithm obtained the threshold range of the optimization scale. At the same time, by adding feature segment merge, the disadvantage of the simple multiscale method, missing the single-layer with multipeak, could be improved. The effective 532 nm Mie lidar data was processed which had more than 75 min detection time and stable structure from May to July 2019 by using differential zero-crossing method, simple multiscale method and improved algorithm were proposed respectively, and based on the cloud-base height of differential zero-crossing method, the average error of the mean square root obtained by the improved algorithm decreased by 32.65%, the uncertainty average decreased by 33.80%. Then the effectiveness of the improved algorithm in improving the accuracy of the cloud-base height is proved.