基于Unity的水电站安全操作仿真平台设计

Design of hydropower station safety operation simulation platform based on Unity

文中基于虚拟现实(VR)技术对水电站安全操作仿真实现进行了研究。在软件层面上,借助Unity在LED-CAVE上完成了水电站虚拟现实平台的搭建,实现了沉浸式环境下的设备学习、设备拆装、水电站虚拟漫游和培训考试等功能。在算法层面上,为了保证仿真环境下设备操作的准确性,对系统中手势的识别跟踪算法进行特殊设计。算法基于检测前跟踪(TBD)的思路,在保留手部完整采集信息的基础上,对数据进行联合处理。算法借助采集的手部数据提取出结构特征、序列特征和空间特征,通过支持向量机(SVM)算法完成手部动作的跟踪识别。为了验证平台的识别准确度,建立了包含12360个视频序列的手势数据集。测试结果表明,系统手部动作识别的平均准确度可达87.2%,而对于抓取、抓放等动作的识别准确度可达95%以上。

In this paper,based on Virtual Reality(VR)technology,the realization of safe operation simulation of hydropower station is studied.On the software level,with the help of Unity,the virtual reality platform of hydropower station is built on LED-CAVE,which realizes the functions of equipment learning,equipment disassembly,virtual roaming,training and examination in the immersive environment.On the algorithm level,in order to ensure the accuracy of the device operation in the simulation environment,the gesture recognition and tracking algorithm in the system is specially designed.The algorithm is based on the idea of Tracking Before Detection(TBD).On the basis of retaining the complete information collected by the hand,the data is combined to process.With the help of hand data collected,the algorithm extracts structural features,sequence features and spatial features,and uses Support Vector Machine(SVM)algorithm to track and recognize hand movements.In order to verify the recognition accuracy of the platform,a gesture data set containing 12360 video sequences is established.The test results show that the average accuracy of hand action recognition of the system can reach 87.2%,and the recognition accuracy of grasping,grasping and other actions can reach more than 95%.