In order to realize on-line detection of formation water

a geotechnical humidity detection system based on image processing and machine learning is designed. The collected digital pictures are preprocessed by high frequency filtering to remove high frequency noise interference. The mixed Gaussian model is used to find the geotechnical area in the picture

and then the geotechnical area in the picture is transformed from the RGB color space to the LAB color space. The DBSCAN clustering algorithm finds the center point of the LAB color space

and uses the LAB value as the elastic network input

and use the predicted geotechnical humidity value as output. The feasibility of the method is verified by building a simulation environment and conducting experiments. Compared with the traditional probe method

it can realize large-area rock and soil moisture detection.