Fire Scene Information Detection Based On Convolutional Neural Network

In recent years,with the continuous progress of image processing technology,imagebased target recognition and detection technology has developed rapidly.The rise of deep learning technology has made convolutional neural networks shine.At present,it has become a research hotspot in the field of computer vision,providing an effective way to solve practical engineering application problems.In order to effectively detect the fire scene,the image of the fire scene is analyzed by using a target detection algorithm based on a convolutional neural network to provide reliable visual information for fire rescue.Summarize the domestic and foreign research results of object recognition and detection based on computer vision,and introduce the object detection based on convolutional neural network.Aiming at the problem of fire scene fire detection,fire scene pedestrian and vehicle information detection,a fire scene information detection model based on convolutional neural network has proposed.For traditional convolutional neural network target detection algorithms that require a large amount of data to train network parameters,it will take a lot of time,and considering that the fire scene data belongs to small sample data,a convolution based on transfer learning is researched and implemented.The neural network of fire detection method uses a transfer learning method to train a fire detection network model.Experiments were performed on the established fire data set.The results show that using this algorithm for fire detection has the advantages of high accuracy,low false alarm rate,and short detection time.The results obtained by applying it to fire detection are good.It provides model support to the use of drones to detect fire scenes in the future.An improved convolutional neural network detection model based on instance segmentation is proposed for occlusion situations that often occur in pedestrian detection.This model combines object detection and semantic segmentation,which not only solves the problem of pedestrian detection miss detection,but also can be more accurate instance segmentation detection for pedestrians.The experimental results have proved that the convolutional neural network based on instance segmentation can provide accurate results for pedestrian detection problems at the scene of fire,so as to obtain effective information.In order to solve the problem of vehicle detection at the fire scene,the vehicle model based on the instance segmentation was used to optimize the training of the vehicle model based on the case scene to obtain the vehicle detection model at the fire scene.The experimental results prove that the model can accurately detect the number of vehicles in the fire scene and location information is feasible and effective.In addition,in order to further obtain vehicle-related information,the vehicle license plate is detected using a target detection network based on the candidate area,so that when the vehicle hinders rescue,the license plate information can be provided in time to contact the owner and reduce losses.