Kezheng Zhang

An analytical calculation methodology utilizing 3D laser scanning data is presented for the assessment of strain in smoothly dented pipelines, offering enhanced precision. This approach adopts cubic B-spline interpolation to reconstruct a smooth dent surface from preprocessed scanning point cloud data, subsequently organizing these into regular grid node coordinates via data gridding. Displacement and strain at each grid point, resulting from pipeline deformation, are determined by employing thin-shell theory alongside geometric deformation analysis. The accuracy of this method is substantiated through comparisons with results obtained from both the finite element method (FEM) and the ASME B31.8 standard. Additionally, specialized evaluation software for assessing dented pipelines has been developed, leveraging this analytical method. Comparative analysis with the finite element method reveals that the average relative errors for maximum equivalent strain on the pipeline's outer surface are 7.73% and 13.16%, respectively, underscoring the superior accuracy of the proposed method over the ASME B31.8 standard for strain calculations on the outer surface of dented pipelines.

Object detection is one important task in automatically analyzing eye tracking video data. This paper presents a real-time object detection method of retail products based on deep learning for eye tracking system. In the proposed approach, an eye tracking based Convolutional Neural Networks is constructed to obtain the feature of original images. Then, a weighted bounding box selection strategy based on gaze location is used for object detection. Besides, parameters are adjusted according to gaze location of eye tracking. Experimental results show that our method can achieve better accuracy for eye tracking than other existing methods in the detection of retail products.