淮阴师范学院学报(自然科学版）

传统教室灯光等设备控制方式较为粗放，易造成用电浪费.本文基于EasyDL平台设计了九宫格视觉感知模型用于教室人员布局的识别，将实时更新的教室人员布局和灯光状态数据存储在数据库中供前端程序调用.针对教室灯光分区控制需求，收集并标注了教室场景小样本数据集进行训练和优化，然后选择在云端部署.整个系统设计了感知层、传输层、平台层和应用共4层.将感知层获取的教室环境数据以及教室的实时图像预处理后调用云端模型得到教室人员布局编码，通过在教室照明区域与人员布局之间建立区域-设备映射，在应用程序中实现灯光与设备的分区按需控制.测试结果表明系统响应时间小于2 s,能够按照教室不同的人员布局来调整照明策略，达到了省电和智能化控制的目的，为传统教室智慧化改造提供精细化解决方案.

The traditional control methods for classroom lighting and other equipment are relatively extensive, which easily leads to electricity waste. This paper designs a nine-grid visual perception model based on the EasyDL platform to recognize classroom occupancy distribution. Real-time updated occupancy data and lighting status are stored in a database for the front-end application access. To meet the needs of partitioned control of the classroom lighting, a small sample dataset of classroom scenes is collected, and labeled for training and optimization, and then deployed on the cloud. The system is designed with four layers: perception layer, transmission layer, platform layer, and application layer. The classroom environmental data obtained by the perception layer and the real-time images of the classroom are preprocessed, and then the cloud model is called to obtain the classroom personnel layout coding. By establishing a region-equipment mapping between lighting zones and occupancy layout, the partitioned on-demand control of lighting and equipment is realized in the application program. Test results demonstrate that the system response time is less than 2 seconds, and it can adjust the lighting strategy according to different personnel layouts in the classroom, achieving the goals of power saving and intelligent control, and providing a refined solution for the intelligent transformation of traditional classrooms.