A Novel Sparse Representation Based Visual Tracking Method for Dynamic Overhead Cranes

Abstract

Efficient tracking of heavy loads is the overall goal for overhead cranes in the workplace. This article presents a new way to simply and effectively track overhead cranes at higher speeds. A real-time tracking method is presented here to make tracking the overhead crane load not only accurate, but also faster. First, to observe and extract details and information to use in the appearance model, a measurement matrix is constructed here. Second, a sparse representation is also adopted to track dynamic overhead crane features effectively. Finally, a Naïve Bayesian classifier has been used as a binary classification in a compressed domain. Experiments on the VOT2013 benchmark and constructed data Cranes40, the results demonstrate that the presented sparse representation tracking method can successfully track cranes in real-time.

Cascade Network Based Bolt Inspection In High-Speed Train

“…Onestage methods STM [5] Strong Low Pixel-wise and histogram similarities [6] Low Adaboost [7] Low Sparse Representation [8] Low Unknown target search in an unknown environment [10]…”

“…In [7], an Adaboost based algorithm was used for detection. In other target detection fields, [8] adopted a sparse representation to track dynamic overhead crane features effectively. In [9], the object-based human identification stage and spatial feature-based human identification stage were fused using the fuzzy holoentropy for accurate human identification.…”