| Air pollution monitoring is an important problem in environmental protection and pollution control.Recently,using image sensing devices to analyze air quality has attracted much attention of researchers.To real-timely detect factory smoke to achieve the efficient and universal social supervision,this paper proposes a mobile-platform-running efficient smoke detection algorithm based on image analysis techniques.This light-weighted detection method could serve as as an important auxiliary means of environmental protection department to monitor factory smoke emission.In the era of big data,to meet multi-scene-task needs of massive data,deep learning theory and technology have been widely studied and applied.Nowadays,For the massive smoke images’ detection task,the object detection method based on deep learning can be effectively used to carry out that.Since most smoke images in real environment have a large distribution scene,the existing object detection methods may have the problems of missing detection and false detection.To this end,this paper proposes the two-stage smoke detection(TSSD)algorithm based on the lightweight detection framework YOLO-V3,in which the prior knowledge and context information are modeled into the designed relation-guided module to reduce the smoke search space,which can therefore significantly improve the shortcomings of the single-stage detection method.Since this research direction is relatively new,there is none dataset for evaluation.This paper designs a specific dataset with 960 high-quality images,then to use multiple data augmentation strategies to expand the dataset for analysis and evaluations.Experimental results show that the proposed TSSD algorithm can robustly improve the detection accuracy of the single-stage method and the model has a good compatibility for different image resolution inputs.Compared with other state-of-the-art detection methods,the accuracy APmean of our proposed TSSD model reaches 59.24%,even surpassing the current detection model Faster RCNN.In addition,the detection speed of our proposed model can reach 50 ms(20 FPS),which meets the real-time requirements,and can be deployed in the mobile terminal carrier.The proposed TSSD algorithm has the advantages of high stability,high accuracy,and fast detection speed.It can be widely used in some scenes with smoke detection requirements,providing the great potential for practical environmental applications. |
