Abstract Video action segmentation is a key technology for understanding videos. Due to the limitations of existing video action segmentation methods, this paper proposed a video action segmentation network based on an improved ASFormer with multiscale temporal features. Aiming at the problem of local temporal feature disappearance caused by the multilayer structure of ASFormer network, we first introduce the multi-scale temporal feature extraction module and the multi-scale attention module to improve the network's temporal feature extraction capability of actions. Second, to improve the network’s ability to localize action boundaries, action boundary detection loss is introduced to enhance the network’s attention at action boundaries. Then, a smoothing post-processing operation is introduced to suppress over-segmentation caused by mutated frames. To verify the effectiveness of the method proposed in this paper, an experimental study is finally performed on a publicly available dataset.

Abstract Hybrid flood defense systems combining structural and nature-based components are increasingly used to mitigate flood hazards in vulnerable regions. Although hybrid flood defense systems combining dikes, moats, and vegetation have been widely explored, the influence of engineered roughness on their hydraulic performance has received little to no attention. This study introduces roughness as a new design parameter to enhance energy dissipation and reduce hydraulic forces within such systems. To address this gap, a scaled physical model of a dike-moat-vegetation system was developed using Froude and geometric similarity to evaluate how engineered roughness alters hydraulic forces under controlled flood conditions. Various hydraulic responses, including fluid force index (FFI), moment index (MI), hydrostatic and hydrodynamic forces, and velocity reduction, were investigated under different hydraulic and structural conditions. The results showed that introducing roughness to the dike and moat substantially reduced flood-induced hydraulic forces. The FFI and MI decreased by 25% and 29%, respectively, at a Froude number (Fr) of 0.60 and roughness height (Ke) =24.47mm compared to the baseline condition without roughness. Similarly, the hydrostatic and hydrodynamic forces were reduced by up to 44% and 20% under the same roughness configuration. The added roughness induced to the body of the dike and moat helped stabilize vortices, redistribute the flow field, thereby improving the resilience of the downstream area. Overall, the findings confirmed that optimized roughness on the dike and moat effectively reduces hydraulic forces and enhances flood mitigation performance of hybrid defense systems, offering an adaptable and sustainable solution for flood-prone regions.