智慧消防中卷积神经网络算法的研究与实现

Research and implementation of convolutional neural network algorithm in smart fire fighting

消防直接关系到社会发展和公众安全。火灾产生的早期阶段准确快速的烟雾检测对预防火灾至关重要。为了提高火灾防控能力,智能评估火灾隐患,在火灾发生初期,及时掌握火灾现状以便为指挥决策提供参考,本文采用基于卷积神经网络算法的烟雾检测。该方法是由静态和动态卷积神经网络组成,能够有效融合卷积神经网络提取的静态激活特征和视频帧之间的运动特征,提高烟雾特征的表示能力。本文以智慧消防为背景,对卷积神经网络算法在火灾发生初期实现可靠的烟雾识别问题进行研究,试验结果表明复杂场景下算法具有良好的稳定性和高效性。

Fire fighting is directly related to social development and public safety. Accurate and rapid smoke detection in the early stages of a fire is vital to fire prevention. In order to improve fire prevention and control capabilities, intelligently assess fire hazards, and in the early stage of a fire, timely grasp the current situation of the fire to provide reference for command and decision-making, a smoke detection method is proposed based on the convolutional neural network algorithm. This method is composed of static and dynamic multi-layer convolutional neural networks, which can effectively fuse the static activation features extracted by deep convolutional neural networks and motion information between video frames, and improve the feature representation ability of smoke. Based on the background of smart fire protection, the convolutional neural network algorithm is studied to achieve reliable smoke recognition in the early stage of fire. The experimental results show the stability and efficiency of the algorithm in complex scenarios.