임베디드 시스템에 인공지능을 적용한 안전모 인식
Automatic safety helmet detection with Machine Learning
1
조회
0
다운로드
1
인용
0
참고문헌
키워드
초록
산업별 재해상황 통계로 나타나는 작업현장에서의 안전사고는 추락, 넘어짐 등의 원인이 50% 이상이며, 이에 따른 두부 손상을 방지하기 위한 안전모 착용이 요구되고 있다. 그러나 작업현장의 특성상 안전모의 미착용 및 작업의 편의성에 따라 작업모의 불완전 착용 등으로 안전사고에 노출되고 있다. 현재 작업 감독자가현장을 순찰하면서, 작업모의 착용 여부 및 안전상황을 점검하는 제도가 시행되고 있지만, 계속적인 작업 감독의 어려움이 있다. 이러한 환경에서 안전사고를 예방하기 위해서는 작업현장에서 개인보호구 착용을 시스템에 의한 자동 관리 감독시스템을 구축하여, 감시와 경고 신호를 발생하는 시스템이 요구된다. 본 연구에서는 라즈베리파이에 인공지능 모델을 포팅한 안전모 착용 여부 감지 시스템을 개발하였다. 객체 검출( Object Detection) 컴퓨터 비전 기술을 이용하여 안전모 착용 여부를 탐지하는 모델을 구성하고, 실시간 카메라 영상분석을 통해 데이터를 입력받은 후, 안전모 미착용이 탐지되는 경우 경고음이 발생하는 시스템을 구축하였다. 이동체에 탑재된 실시간 영상을 통한 검출 정확도는 87%를 달성하였으며 TensorFlow lite 파일로 변환하는과정을 통해 기존의 TensorFlow 모델보다 20% 향상된 fps 성능 결과를 보였다.
Abstract
More than 50% of safety accidents at work sites, by the statistics of accident situations in the industry, have been caused by falling off and falling down. Accordingly, it is required to wear a safety helmet to prevent head injury. Currently, a system in which work supervisors patrol the site and check whether the helmet is worn and the safety situation is being implemented. However, it is difficult to continuously supervise work. In order to prevent safety accidents, it is necessary to establish an automatic management and supervision system by system for wearing personal protective equipment at the work site. Also a system is required to generate warning signals. In this study, a safety helmet-wearing detection system was developed by porting an artificial intelligence model to Raspberry Pi. A model was constructed to detect whether or not a helmet was worn using object detection computer vision technology. The system generates a warning sound when not wearing a helmet is detected. The system was mounted on a mobile body and analyzed images. The detection accuracy was achieved 87%, and through the process of converting to TensorFlow lite file, fps performance improved by 20% compared to the existing TensorFlow model.
참고문헌 (17)
- "Construction site safety hat" (2021)[URL]
- "Construction site safety hat" (2022)[URL]
- Rajalingappaa Shanmugamani, "Deep Learning for Compuer Vision : Expert techniques to train advanced neural networks using Tensorflow and Keras", pp.147~175 (2018), Packt Publishing
- Wei Liu, "SSD: Single Shot MultiBox Detector", Computer Vision – ECCV 2016, pp.21~37 (2016), Springer International Publishing[DOI]
- Zheng Zeyu, "Google Deep Learning Framework", TensorFlow Practice, ~ (2018)
- Rowel Atienza, "Advanced Deep Learning with TensorFlow2 and Keras - Second Edition" (2020), Packt Publishing
- Mark Sandler, "MobileNetV2: Inverted Residuals and Linear Bottlenecks", 1~2 (2019)
- Junhua Gu, "Joint Pedestrian and Body Part Detection via Semantic Relationship Learning", Applied Sciences, 9(4), pp.752~ (2019), MDPI AG[DOI]
- Hua Gao, "Parts Semantic Segmentation Aware Representation Learning for Person Re-Identification", Applied Sciences, 9(6), pp.1239~ (2019), MDPI AG[DOI]
- Ronneberger, O., "U-net: Convolutional networks for biomedical image segmentation", International Conference on Medical Image Computing and Computer-Assisted Intervention, pp.234~241 (2015)
- Samir Bouindour, "An On-Line and Adaptive Method for Detecting Abnormal Events in Videos Using Spatio-Temporal ConvNet", Applied Sciences, 9(4), pp.757~ (2019), MDPI AG[DOI]
- Ren, S., "Faster r-cnn: Towards real-time object detection with region proposal networks", Adv. Neural Inf. Process. Syst, 28, pp.91~99 (2015)[DOI]
- Radac, M.B., "Data-Driven Model-Free Tracking Reinforcement Learning Control with VRFT-based Adaptive Actor-Critic", Applied Sciences, 9(9), pp.1807~ (2019), MDPI AG[DOI]
- Changyun Wei, "Obtaining Human Experience for Intelligent Dredger Control: A Reinforcement Learning Approach", Applied Sciences, 9(9), pp.1769~ (2019), MDPI AG[DOI]
- M. I. Jordan, "Machine learning: Trends, perspectives, and prospects", Science, 349(6245), pp.255~260 (2015), American Association for the Advancement of Science (AAAS)[DOI]
- A. Ng, "Machine learning yearning:Technical strategy for ai engineers in the era of deep learning", ~ (2019)
- Y. LeCun, "Convolutional networks and applications in vision", Proc. IEEE Int. Symp. Circuits Syst, pp.253~256 (2010)