AN IMPROVED OBJECT DETECTION METHOD BASED ON DEEP CONVOLUTION NEURAL NETWORK FOR SMOKE DETECTION

The smoke detection plays a very important role in fire alarm. However, the accuracy of smoke detection is low and difficult to detect in open space by traditional methods. In this paper, we introduce an improved object detection method based on deep convolution neural network (CNN) to address this issue. Firstly, we substitute the feature extractor (such as Inception Net and Resnet) in Various neural network object detectors for Faster R-CNN, Single Shot MultiBox Detector (SSD), Region based Fully Convolutional Networks (R-FCN). Secondly, the parameters of the object detection algorithm are optimized on MSCOCO dataset. Finally, the experiments are conducted on the smoke detection dataset. The experiments result demonstrate the mAP reached 56.04% on the smoke detection dataset. Compared with the existing smoke detection methods, the present method has achieved good results both in accuracy and speed.