Contextual Information and Covariance Descriptors for People Surveillance: An Application for Safety of Construction Workers

Abstract

In computer science, contextual information can be used both to reduce computations and to increase accuracy. This paper discusses how it can be exploited for people surveillance in very cluttered environments in terms of perspective (i.e., weak scene calibration) and appearance of the objects of interest (i.e., relevance feedback on the training of a classifier). These techniques are applied to a pedestrian detector that uses a LogitBoost classifier, appropriately modified to work with covariance descriptors which lie on Riemannian manifolds. On each detected pedestrian, a similar classifier is employed to obtain a precise localization of the head. Two novelties on the algorithms are proposed in this case: polar image transformations to better exploit the circular feature of the head appearance and multispectral image derivatives that catch not only luminance but also chrominance variations. The complete approach has been tested on the surveillance of a construction site to detect workers that do not wear the hard hat: in such scenarios, the complexity and dynamics are very high, making pedestrian detection a real challenge.

Computer vision techniques for construction safety and health monitoring

“…Object detection can be directly used to identify unsafe conditions and acts at construction sites (e.g., failure to wear personal protective equipment and damaged area of building or defected material). For example, detection of safety gears such as safety vests and hard hats helps to distinguish construction workers from others (e.g., supervisors, engineers or pedestrians) [10,[41][42][43]85,86], as well as to detect construction workers who do not wear safety gears for safety observation [18,87]. In addition, detection of construction workers and material surrounding equipment helps operators who have limited visibility on job sites to prevent collisions [14].…”

Hardhat-Wearing Detection Based on a Lightweight Convolutional Neural Network with Multi-Scale Features and a Top-Down Module

“…In the past few years, some efforts have been made by researchers to solve the hardhat-wearing detection problem based on traditional computer vision and machine-learning techniques [1,[3][4][5][6][7]. Most of these methods employ multi-step strategies.…”

Computer‐Aided Optimization of Surveillance Cameras Placement on Construction Sites

“…Surveillance camera is nonintrusive to general construction tasks without attaching tag to workers or their proactive equipment (Park et al., ; Rezazadeh Azar and McCabe, ). With the advancement of computer vision techniques on construction sites, cameras become significant in numerous implementations such as (a) detecting and tracking construction‐related entities to calculate productivity (Chi et al., ; Gong and Caldas, ; Brilakis et al., ; Chi and Caldas, , ; Gong et al., ) and avoiding collisions (Chen et al., ; Hamledari et al., ); (b) recognizing working condition and environmental context to monitor construction progress and obtain safety context (Gualdi et al., ; Park and Brilakis, ); (c) tracking workforce and detecting motions to prevent proximity to hazards (Teizer and Vela, ; Yang et al., ) and preventing muscular injuries from awkward postures or ergonomic risks (Ray and Teizer, ; Han and Lee, ; Han et al., ; Yang et al., ); (d) monitoring and identifying damages and quality issues (Salem et al., ; Yeum and Dyke, ; Cha et al., ; Cha et al., ; Kong and Li, ), especially visual cracks (Chen et al., ; Zhang et al., ); (e) inspecting structural conditions in hazardous environment (Zhu et al., ; Oh et al., ; Park et al., ); and (f) reconstructing building models (Fleishman et al., ; Olague and Mohr, ). Based on the diverse applications of surveillance cameras, intelligible information is extracted automatically in real time, offering an effective solution to time‐ and labor‐consuming inspections on construction sites (Mirchandani et al., ).…”