Helmet Detection under the Scene of Construction Based on Image Analysis

• 1、Harbin Engineering University,College of Information and Communication Engineering,Harbin,150000

Abstract：This paper mainly focuses on the problem of whether the construction workers wear the helmets correctly under the construction site scene, and uses the image analysis method to detect the helmets under the construction scene. Traditional manual detection methods are time-consuming, inefficient, and waste resources. The use of images can achieve the purpose of automated detection. The main target of detection is the helmet and the head. The main method used is the YOLOv3 target detection method. This method is used to fine-tune the self-made data set of the construction scene. By adjusting the super-parameters and continuously training, a model capable of stably detecting the helmet is obtained. In order to monitor the compliance of the workers in the construction scene, the construction personnel are guaranteed to ensure the personal safety of the construction personnel. After experimental verification, the final model has a detection rate of more than 95% for the helmet, which can be applied to the actual construction scenario.

Keywords： Image analysis Construction background YOLOv3 Helmet