A Hybrid Kinematic-Acoustic System for Automated Activity Detection of Construction Equipment.

摘要: Automatically recognizing and tracking construction equipment activities is the first step towards performance monitoring of a job site. Recognizing helps managers to detect downtime/idle time in real-time framework, estimate productivity rate each based on its progress, efficiently evaluate cycle activity. Thus, it leads project cost reduction schedule improvement. Previous studies this topic have been single sources data (e.g., kinematic, audio, video signals) for automated activity-detection purposes. However, relying only one source not appropriate, as selected may be applicable under certain conditions fails provide accurate results. To tackle issue, authors propose hybrid system multiple equipment. The integrates two major data-audio kinematic-through implementing robust fusion procedure. presented includes recording audio kinematic signals, preprocessing data, extracting several features, well dimension reduction, feature fusion, activity classification using Support Vector Machines (SVM), smoothing labels. proposed was implemented case (i.e., ten different types models operating at various sites) results indicate that capable providing up 20% more results, compared cases individual data.