A computer vision and reasoning based video interpretation system for rapid productivity analysis of construction operations

On construction jobsites, digital cameras are inexpensive devices that are widely used as an effective site observation method (Oglesby et al. 1989; Kim and Bai 2009). This opens the door for conducting scientific method studies on complex operations for productivity improvement through examining recorded videos. However, tedious manual reviewing is currently still required to extract productivity information from the recorded videos (Su and Liu 2007; Navon and Sacks 2007). The delay and expense caused by such intensive manual efforts can cause high data collection expenses and the dilution of the effectiveness of such a program. Consequently, jobsite leadership frequently considers construction cameras as just a contractual tool for site surveillance, dispute resolution, and site demonstration. Intelligent computer vision systems that can autonomously observe the behavior of construction objects and infer their working status have the potential to enable automated productivity assessment and therefore reduce some of the shortcomings of the manual methods (Gong and Caldas 2009; Zou and Kim 2007).