Hard hat wearing detection based on head keypoint localization

Abstract

In recent years, a lot of attention is paid to deep learning methods in the context of vision-based construction site safety systems, especially regarding personal protective equipment. However, despite all this attention, there is still no reliable way to establish the relationship between workers and their hard hats. To answer this problem a combination of deep learning, object detection and head keypoint localization, with simple rule-based reasoning is proposed in this article. In tests, this solution surpassed the previous methods based on the relative bounding box position of different instances, as well as direct detection of hard hat wearers and non-wearers. The results show that the conjunction of novel deep learning methods with humanly-interpretable rule-based systems can result in a solution that is both reliable and can successfully mimic manual, on-site supervision. This work is the next step in the development of fully autonomous construction site safety systems and shows that there is still room for improvement in this area.

MCA-YOLOV5-Light: A Faster, Stronger and Lighter Algorithm for Helmet-Wearing Detection

“…According to the statistics from Eurostat and IBS in 2020, the number of non-fatal and fatal accidents in the construction industry always exceeds that in the manufacturing industry, especially head injury accidents at construction sites. Even though the number of head injuries is only about 7% of non-fatal accidents, they account for over 30% of fatal accidents, making it a significant issue and critical to the safety of construction workers [1]. The most common head injury on construction sites is traumatic brain injury (TBI), which can be fatal and which occurs mainly by the rapid acceleration or deceleration of the brain moving and colliding with the skull.…”

“…Due to the limited computational overhead, we adopt the multispectral channel attention (MCA) strategy, use the discrete cosine transform (DCT) in the channel attention mechanism to compress the channel and use the model-pruning method for sparse training to achieve higher detection accuracy with smaller models in the hat-wearing detection task. The reasons are as follows: (1) The global average pooling operation cannot express rich feature information [26]. GAP is a special case of DCT, and it is embedded in the MCA framework, adding more frequency component information obtained from redundant channels, which can extract more useful information; (2) The CNN model has the problems of model size, running time occupying memory and large amount of calculation.…”