Face Mask and Social Distance Monitoring via Computer Vision and Deployable System Architecture

The coronavirus(COVID-19)is a lethal virus causing a rapidly infec-tious disease throughout the globe.Spreading awareness,taking preventive mea-sures,imposing strict restrictions on public gatherings,wearing facial masks,and maintaining safe social distancing have become crucial factors in keeping the virus at bay.Even though the world has spent a whole year preventing and curing the disease caused by the COVID-19 virus,the statistics show that the virus can cause an outbreak at any time on a large scale if thorough preventive measures are not maintained accordingly.Tofight the spread of this virus,technologically developed systems have become very useful.However,the implementation of an automatic,robust,continuous,and lightweight monitoring system that can be efficiently deployed on an embedded device still has not become prevalent in the mass community.This paper aims to develop an automatic system to simul-taneously detect social distance and face mask violation in real-time that has been deployed in an embedded system.A modified version of a convolutional neural network,the ResNet50 model,has been utilized to identify masked faces in peo-ple.You Only Look Once(YOLOv3)approach is applied for object detection and the DeepSORT technique is used to measure the social distance.The efficiency of the proposed model is tested on real-time video sequences taken from a video streaming source from an embedded system,Jetson Nano edge computing device,and smartphones,Android and iOS applications.Empirical results show that the implemented model can efficiently detect facial masks and social distance viola-tions with acceptable accuracy and precision scores.

Large-scale deformation monitoring in mining area by D-InSAR and 3D laser scanning technology integration

Large-scale deformation can not be detected by traditional D-InSAR technique because of the limit of its detectable deformation gradient,we propose a method that combines SAR data with point cloud data obtained by 3D laser scanning to improve the gradient of deformation detection.The proposed method takes advantage of high-density of 3D laser scanning point cloud data and its high precision of point positioning after 3D modeling.The specifc process can be described as follows:frst,large-scale deformation points in the interferogram are masked out based on interferometric coherence;second,the interferogram with holes is unwrapped to obtain a deformation map with holes,and last,the holes in the deformation map are flled with point cloud data using inverse distance weighting algorithm,which will achieve seamless connection of monitoring region.We took the embankment dam above working face of a certain mining area in Shandong province as an example to study large-scale deformation in mining area using the proposed method.The results show that the maximum absolute error is 64 mm,relative error of maximum subsidence value is 4.95%,and they are consistent with leveling data of ground observation stations,which confrms the feasibility of this method.The method we presented provides new ways and means for achieving large-scale deformation monitoring by D-InSAR in mining area.

Learning from the crowd:Road infrastructure monitoring system

The condition of the road infrastructure has severe impacts on the road safety, driving comfort, and on the rolling resistance. Therefore, the road infrastructure must be moni- tored comprehensively and in regular intervals to identify damaged road segments and road hazards. Methods have been developed to comprehensively and automatically digitize the road infrastructure and estimate the road quality, which are based on vehicle sensors and a supervised machine learning classification. Since different types of vehicles have various suspension systems with different response functions, one classifier cannot be taken over to other vehicles. Usually, a high amount of time is needed to acquire training data for each individual vehicle and classifier. To address this problem, the methods to collect training data automatically for new vehicles based on the comparison of trajectories of untrained and trained vehicles have been developed. The results show that the method based on a k-dimensional tree and Euclidean distance performs best and is robust in transferring the information of the road surface from one vehicle to another. Furthermore, this method offers the possibility to merge the output and road infrastructure information from multiple vehicles to enable a more robust and precise prediction of the ground truth.