Identification of Construction Cumulative Trauma Disorders: A Machine Vision Approach

Cumulative trauma disorders (CTDs) are a leading cause of disability which generates huge direct and indirect costs in the construction industry. Existing efforts identify the risk of CTDs through survey or observatory studies which are subjective and expensive. This paper attempts to automatically identify the risk of CTDs on construction sites with the aid of a video surveillance system. Specific emphasis is placed on low back disorders (LBDs), which is one of the major work-related CTDs. To this end, a risk identification method is proposed based on location changes of region of interests in the resulting video streams at minimum supervision. It is used to detect repetitive trunk flexion-extension motions, awkward trunk postures, and to assess the risk of job-related LBDs. Preliminary experimental analysis on lab datasets demonstrates the promise of the proposed method.