Aiming at the uneven settlement which is easy to occur in the process of soft ground construction of geotechnical engineering, this paper introduces genetic algorithm into the neural network algorithm, which makes the population converge to the optimal solution by controlling the crossover probability and variance probability. The kd-tree is constructed using the foundation building point cloud to get the plane area of the foundation building point cloud. Alpha-shape algorithm is used to extract the contour lines of the foundation area, and the feature lines of the foundation plane area are obtained by fitting the results of contour line extraction. The feature line matching method is combined with the ICP algorithm to perform point cloud coarse alignment and point cloud fine alignment. The optimized neural network prediction model is used to predict and analyze the point cloud alignment results of the foundation based on the ICP algorithm. The stability of the GA-BP settlement prediction algorithm designed in this paper is verified by combining the error results of the GA-BP algorithm on soft ground settlement prediction in engineering examples. Comprehensive K161+050, K162+872, K175+600 section in the pile embankment settlement observation data and prediction data, GA-BP algorithm prediction of the maximum error, the minimum error of 9.53%, 0.66%, respectively, the model prediction data and the actual observation data similar to the prediction model has a good prediction accuracy.
