基于图像处理和神经网络的光环境优化控制研究

为了精确定位人体目标区域, 以达到照明节能最优化, 提出了一种改进的Canny算子与数学形态学相结合的图像处理技术, 对室内图像进行边缘检测和噪声处理, 得到含有人体目标的二值化图像, 将得到的二值化人体背影模板的七个Hu不变矩作为LVQ神经网络的输入向量, 训练出能够准确识别出人体背影的网络; 再建立出室内图像特征与人体数量和坐标的匹配模型, 从而根据人体坐标和数量来实现对灯具的优化控制。仿真和实验结果表明, 改进的Canny算子与LVQ神经网络的结合对室内光环境的优化控制具有很好的应用价值。

Study of light environment optimization and control based onimage processing and neural network

In order to positioning the target area of human body to achieve the optimization of the lighting energy saving, this paper put forward an image processing technique combined the improved Canny operator and mathematical morphology, detected the edge of the indoor image and noise processing, and then to get the binary image containing human as target. It took seven Hu moment invariants of binary human figure template as the input vector of LVQ neural network, which were calculated to get, and trained network to identify the human figure accurately. After that, building the matching model of indoor image characteristics and the personnel quantity and coordinates was well designed to realize the optimization control to indoor lamps according to personnel quantity and coordinates. Simulation and experimental results show that with the combination of the improved Canny operator and LVQ neural network has quite application value on the optimization control of indoor light environment.