Classroom Brightness Detection Based on Deep Learning

In the classroom energy-saving system, the traditional method of using the illuminance sensor to detect brightness has problems such as wiring difficulties, equipment failure, low detection accuracy, and the need for regular replacement of equipment. Without changing the original wiring layout of the classroom, using the existing cameras in the classroom and adopting the deep learning method based on image processing has the advantages of saving money, high feasibility, and high accuracy. First, the HSV color model is used to obtain the average brightness value of the image, and then the optimizer is used to obtain the best threshold for light and dark classification. The data set is labeled with light and dark through the best threshold. To make up for the lack of data set, use random scaling and cropping to the data After enhancement, the VGG convolutional neural network is used for training, and the accuracy of bright and dark classification reaches 99.6%. At the same time, it is compared with the deep neural network algorithm and the day and night classification algorithm, it is found that using the VGG convolutional neural network for bright and dark classification has the best effect.