| description abstract | A hidden karst cave has a prominent influence on the design and construction of underground engineering. In order to improve the automatic identification accuracy of hidden karst cave areas and the acquisition accuracy of characteristic parameters in borehole radar signals, combined with the reflection characteristic relationship between radar electromagnetic waves and hidden karst caves, research methods on borehole radar image pre-processing and feature parameter extraction of hidden karst caves are carried out. First, based on the original signal of a borehole radar, the reflection characteristic relationship between the electromagnetic waves of the borehole radar and a hidden karst cave is established, and the geometric model of borehole radar detection of hidden karst caves is formed. Then, the reconstruction of borehole radar signals is realized by searching the peak position of Pmusic function combined with the subspace MUSIC method. Combining the continuity of hyperbola and the gray level difference of a nonreflective region, an improved variance method is proposed, which is more suitable for the actual borehole radar image segmentation. It realizes the suspected hidden karst cave inversion by setting a threshold of continuous pixels, a simultaneous gradient operator, and a maximum interspecific variance method. Finally, according to the geometric model of borehole radar detection, the energy-weighted fitting of a hyperbolic profile is carried out, and the improved Hough transform method is used to solve the multiple characteristic parameters of the hyperbolic model, so as to extract the characteristic parameters of the hidden cave from the borehole radar signal. The example analyzed proves that the method is feasible and accurate, which can provide important data support for underground engineering. | |
