SBDNN: Straw Burning Detection Based on Improved Faster RCNN Neural Network

In this paper, an improved fast RCNN neural network is bring forward to detect the smoke from straw combustion. The feature graph is extracted from the bottom-up, top-down and horizontal link fusion method, which can better learn and use the smaller features, and increase the accuracy of smoke distinguish. Contrast with the existing convolution neural network model, the experimental results show that this approach has a better detection accuracy. Compared with fast RCNN detection network, the detection precision of this method is superior by 10%, and the false positive rate is reduced by 16.2%.