Project Related Entities Tracking on Construction Sites by Particle Filtering
Xiaoning Ren, Zhenhua Zhu, Zhi Chen, Fei Dai · Construction Research Congress 2016 · 2016
Vision-based tracking for project related entities has attracted practitioners’ interests and attentions; it can provide beneficial data for productivity analysis and safety monitoring. Some studies on tracking workforce and equipment using video cameras placed onsite have proved the feasibility and efficiency of vision-based tracking methods. However, existing tracking techniques have difficulties in tracking objects when occlusions occur. This paper presents a visual tracking method based on particle filters to resolve this issue. The method includes two main stages, prediction, and update. Initially, the target object is located with a rectangular window, and particles are generated from a normal distribution. Then, particles are propagated, and the weight of each particle is determined by the observation likelihoods. Particles are resampled to localize the target object based on weights. In this way, the personnel or construction equipment can be traced. The jobsite of Roccabella residential project in Montreal was selected as the test bed. A high definition camera was placed onsite to record the construction activities. Then, the collected videos were used to evaluate the tracking performance of this method. The results indicated the method was effective to track the object of interest in the complex situation of occlusions.