An intelligent highway construction monitoring system based on remote wide area networks and multi-technologies.

The existing highway construction monitoring systems are independently deployed. This paper proposes an intelligent monitoring system for highway construction based on remote wide area networks. By employing the tracking devices equipped with inertial measurement units and global positioning systems, the workers' activity space is monitored. People and equipment entering the site are registered by radio frequency identification tags with protective measures. Motion sensors are employed to monitor the changes in scaffolding, building structures, and highway low-grade inclinations. In addition, a light-weight object detection (LOD) algorithm is proposed to detect construction workers who fail to wear safety helmets. The LOD algorithm replaces the standard convolutional networks with depthwise separable convolution networks. It introduces the receptor field module, employs the LOD-NMS algorithm and Mish activation function, and sets an appropriate prior box according to the characteristics and ratios of people's heads in Asia, thus well balancing the precision and prediction efficiency. Experiments are conducted on the WIDER FACE open source data set and the real data sets. The results show LOD algorithm not only detects workers' safety helmets in real time with high precision, but also achieves higher recognition accuracy than traditional algorithms despite the low-resolution input images. [ABSTRACT FROM AUTHOR]