LAUSR

The problem of module discrimination and identification in the field of landscape design is the focus of researchers. Based on multimodal intelligent computing, this paper constructs a landscape design system based on deep neural network. The article first uses a deep neural network to train multimodal garden landscape images, and then performs pooling and convolution operations on garden landscape images on the multimodal training model of convergence speed on the edge and solve the problem of low model accuracy. In the simulation process, the neural network module of MATLAB software is used to extract the spatiotemporal features of the dynamic garden landscape image from the three directions of the bottom block of the garden to achieve feature complementarity. This method only uses 15% of the features of the original feature set. The complexity of the recognition system also reduces the recognition error rate. The experimental results show that by adopting the design of feature series fusion, maximum value fusion, and multiplicative fusion in the score layer, the feature series fusion achieves a high accuracy rate under the multiplicative fusion of the three modalities, reaching 77.1%, and the test error is within 0.118, which effectively improves the multimodal characteristics of the integrated landscape and makes the modeling results more accurate.