Abstract：Automatic recognition of construction worker activities is key to real-time management of construction quality, safety and efficiency. An effective system requires the integration of many areas such as construction, information and management. Taking steel benders as an example, acceleration data from the wrists of steel benders was collected using accelerometers during the construction process. All activities were divided into three categories. The features were extracted from the acceleration data. Then classifiers were used to conduct machine learning with different features. The accuracies of different classifiers were compared to find the best classifiers and features. Tests demonstrate that the best features are related to the data segment length. Within a certain range, longer data segments give better recognition accuracy. This method gives more detailed classifications of the steel benders' activities with a recognition accuracy rate of 85.9% that is higher than previous methods. It provides a method for recognition of typical worker activities on site, and establishes a foundation for automated, real-time monitoring of construction workers to improve the quality, safety and efficiency of construction management.
Mathie M J, Coster A C F, Lovell N H, et al. Detection of daily physical activities using a triaxial accelerometer[J]. Medical and Biological Engineering and Computing, 2003, 41(3):296-301.
Pärkkä J, Ermes M, Korpipää P, et al. Activity classification using realistic data from wearable sensors[J]. Information Technology in Biomedicine, IEEE Transactions on, 2006, 10(1):119-128.
王昌喜. 基于加速度信息的上肢动作识别系统设计及动作质量评价方法的研究[D]. 合肥:中国科学技术大学, 2010.WANG Changxi. Design of Recognition System and Study of Quality Evaluation for Upper Limb Movement Based on Acceleration[D]. Hefei:University of Science and Technology of China, 2010. (in Chinese)
Zappi P, Stiefmeier T, Farella E, et al. Activity recognition from on-body sensors by classifier fusion:Sensor scalability and robustness[C]//3rd International Conference on Intelligent Sensors, Sensor Networks and Information, 2007. Melbourne, Australia:IEEE, 2007:281-286.
Koskimäki H, Huikari V, Siirtola P, et al. Activity recognition using a wrist-worn inertial measurement unit:A case study for industrial assembly lines[C]//17th Mediterranean Conference on Control and Automation, 2009. Thessaloniki, Greece:IEEE, 2009:401-405.
Joshua L, Varghese K. Accelerometer-based activity recognition in construction[J]. Journal of Computing in Civil Engineering, 2011, 25(5):370-379.
Joshua L, Varghese K. Automated recognition of construction labour activity using accelerometers in field situations[J]. International Journal of Productivity and Performance Management, 2013, 63(7):841-862.
薛洋. 基于单个加速度传感器的人体运动模式识别[D]. 广州:华南理工大学, 2011.XUE Yang. Human Motion Patterns Recognition Based on Single Triaxial Accelerometer[D]. Guangzhou:South China University of Technology, 2011. (in Chinese)
Joshua L, Varghese K. Selection of accelerometer location on bricklayers using decision trees[J]. Computer-Aided Civil and Infrastructure Engineering, 2013, 28(5):372-388.
汪仲伟. 基于智能手机中传感器的用户理解[D]. 合肥:中国科学技术大学, 2014.WANG Zhongwei. The Understand of User Based on Sensors Build-in Smart Phones[D]. Hefei:University of Science and Technology of China, 2014. (in Chinese)
Mizell D. Using gravity to estimate accelerometer orientation[C]//Proceedings of the 7th IEEE International Symposium on Wearable Computers (ISWC 2003). Washington DC, USA:IEEE, 2003:252-253.
杨殿阁, 何长伟, 李满, 等. 基于支持向量机的汽车转向与换道行为识别[J]. 清华大学学报(自然科学版), 2015, 55(10):1093-1097.YANG Diange, HE Changwei, LI Man, et al. Vehicle steering and lane-changing behavior recognition based on a support vector machine[J]. J Tsinghua Univ (Sci & Technol), 2015, 55(10):1093-1097. (in Chinese)