Study on the method of estimating the volume of fragmental rockfall based on image recognition

Abstract

There are frequent cataclastic rockfall disasters in southwest mountainous areas. To accurately simulate the process of rockfall movement and quantify the risk of rockfall, it must first determine the volume of rockfall. However, to date, there is no effective and reliable method to estimate the volume of rockfall. This study constructed an effective method of estimating cataclastic rock rockfall volume based on image recognition. The application and verification of this method was conducted at the rockfall of Yaohe Dam in Shimian of Yaxi Expressway in 2020. Through in-site measurement and close photogrammetry of UAV, the partition and sampling area of rockfall accumulation area are determined. Using the open-source software of image processing (ImagePy), a quick block identification step is established, and characteristic parameters, such as equivalent particle size, perimeter, and area of blocks, are extracted. The method of estimating the volume of cataclastic rock rockfall based on the volume distribution of blocks was applied and verified by rock rockfall in the field. The results show that: (1) ImagePy software has high speed and accuracy in block image recognition; (2) The obtained distribution curve of rock volume is nearly consistent with that measured in the field; (3) The estimated volume of rock rockfall in Yaohe Dam accounts for nearly 80% of the volume obtained by three-dimensional point-cloud difference method. Thus, it is feasible to use image recognition technology to extract the particle size and estimate the size of cataclastic rock mass, and it has the advantages of high efficiency and accuracy. This method can be applied to rapid assessment of cataclastic rock mass and quantitative risk assessment. Statistical block volume distribution can provide data support for fragmentation research.

Keywords

• rockfall deposits
• image identification
• particle size distribution
• detailed estimate