Key points are not available for this paper at this time.
The construction industry is still considered among the riskiest industries in the world because workers are continuously exposed to injury from falls, slips, or trips or being struck by falling objects. Hence, safety programs have been according great emphasis on enforcing proper use of personal protective equipment (PPE) by deploying safety officers on construction sites. However, the current practice of supervising large construction areas is still manual, tedious, and ineffective. Therefore, this study aims at creating an integrated framework that can automatically and efficiently detect any noncompliance with safety rules and regulations, in particular a failure to wear a hardhat, using computer vision techniques applied on videos captured from construction sites. This is mainly achieved by (1) isolating mobile workers or construction personnel from the captured scene by means of a novel motion detection algorithm and a human classifier and (2) detecting the hardhat in the identified region of interest using an object detection tool coupled with a color-based image classification one. Several experiments were conducted and results highlighted that the proposed framework proved accurate, fast, and robust under different conditions and identified hardhats with high precision and recall. More specifically, the newly developed motion detection algorithm showed an improved accuracy compared to common background subtraction methods; the human classifier performed well and was able to identify several human postures, unlike support vector machine classifiers; and the hardhat detection algorithm achieved high precision and recall simultaneously.
Mneymneh et al. (Mon,) studied this question.