The magnetic adsorption wall-climbing robot is expected to be the future choice for surface processing of large complex steel structures due to its strong load capacity, high adsorption stability, no additional energy consumption, flexible configuration and low running noise. However, there are still many challenges to be solved in terms of mobile processing efficiency and integration of the electromagnetic adsorption wall-climbing robot. The lack of control ability of magnetic adsorption force and power density of mobile adhesion unit are the key issues which limit the processing efficiency and load weight ratio. In order to solve the pain points of mobile adsorption unit of magnetic adsorption wall-climbing robot, a mobile adhesion integration unit based on stator structure of external tooth slotted BLDC is proposed in this paper. Through the organic combination of electromagnetic adsorption and permanent magnet adsorption, the active control of adsorption force and the integration of self-driving function are realized. The adsorption and motion capacity of mobile adhesion integration unit are verified by FEM and experiment.

Learning 3D global features by aggregating multiple views is important. Pooling is widely used to aggregate views in deep learning models. However, pooling disregards a lot of content information within views and the spatial relationship among the views, which limits the discriminability of learned features. To resolve this issue, 3D to Sequential Views (3D2SeqViews) is proposed to more effectively aggregate the sequential views using convolutional neural networks with a novel hierarchical attention aggregation. Specifically, the content information within each view is first encoded. Then, the encoded view content information and the sequential spatiality among the views are simultaneously aggregated by the hierarchical attention aggregation, where view-level attention and class-level attention are proposed to hierarchically weight sequential views and shape classes. View-level attention is learned to indicate how much attention is paid to each view by each shape class, which subsequently weights sequential views through a novel recursive view integration. Recursive view integration learns the semantic meaning of view sequence, which is robust to the first view position. Furthermore, class-level attention is introduced to describe how much attention is paid to each shape class, which innovatively employs the discriminative ability of the fine-tuned network. 3D2SeqViews learns more discriminative features than the state-of-the-art, which leads to the outperforming results in shape classification and retrieval under three large-scale benchmarks.