Abstract:As the core facility of mining production, the stability of tailings dams has a significant impact on mine safety and the surrounding environment. Due to the influence of surface vegetation coverage and complex terrain, airborne laser point cloud data often faces issues of uneven density and multi-scale noise interference during the acquisition process, leading to deviations in deformation estimation using traditional methods. Therefore, a displacement and deformation monitoring method for tailings dams based on airborne laser point cloud data filtering is proposed. A spatial index is established through K-nearest neighbor search algorithm to partition multi-scale noise, and a dual constraint mechanism of spatial distance weight and geometric similarity weight is introduced, combined with bilateral filtering algorithm to effectively suppress noise interference. At the same time, object segmentation technology is used to divide the monitoring area into 3D grid units, achieving high-precision monitoring of horizontal deformation and vertical settlement of tailings dams. The experimental results show that this method significantly reduces the average absolute error in horizontal deformation and vertical settlement monitoring, with low displacement rate fluctuation and a maximum error of only 0.4%, providing millimeter level perception capability for the safety of tailings dams throughout their entire life cycle. Compared with traditional DS InSAR technology and temporal decomposition models, this research method exhibits higher monitoring accuracy and stability in complex vegetation cover and terrain undulating areas, especially suitable for long-term safety warning and dynamic management scenarios of tailings dams.