摘要
针对屡禁不止的消防车通道占用、堵塞的问题,为保“生命通道”时时畅通,消防部门定时开展消防车通道安全隐患检查工作变得尤为重要,然而人工巡检消防车通道除了投入成本大、效率低等问题外更存在执法力量不足、取证不便、执法告知不易等难点,致使该问题具备隐患动态的特点。对此,该文提出一种基于多尺度特征融合网络的消防车通道占用检测算法和光照补偿模块。首先对图像特别在夜晚场景下进行亮度优化,在图像过亮或过暗的区域实现像素点亮度局部自适应调整,再使用多个分支提取不同尺度图像的特征,并通过注意机制进行优化,最后进行特征融合。通过语义分割得到目标区域的消防车通道和车辆信息后,利用占用判定算法实现消防车通道的占用检测。经验证,该方法在相关数据集上白天和夜晚场景检测准确率分别达到了97%和85.2%,兼具实时性、高效性等优点,能有效降低消防车通道占用检查难度与成本。
In view of the repeated occupation and blockage of fire truck passages,in order to ensure that the"life passage"is always unblocked,it is particularly important for the fire department to regularly carry out the inspection of potential safety hazards of fire truck passages.However,in addition to the problems of high investment cost and low efficiency,manual inspection of fire truck passage also has difficulties such as insufficient law enforcement force,inconvenient evidence collection and difficult law enforcement notification.As a result,the problem has the characteristics of hidden danger dynamics.In this case,a fire truck passages occupancy detection algorithm based on multi-scale feature fusion network and illumination compensation module is proposed.Firstly,the brightness of the image,especially in the night scene,is optimized,and the local adaptive adjustment of pixel illumination is realized in the area where the image is too bright or too dark.Then,multiple branches are used to extract the features of images with different scales and optimized through the attention mechanism.Finally,feature fusion is carried out.After obtaining the fire truck passages and vehicle information in the target area through semantic segmentation,the occupation detection of fire truck passages is realized by using the occupation determination algorithm.It has been verified that the accuracy of scene detection in daytime and night on the relevant data sets has reached 97%and 85.2%respectively.It has the advantages of real-time and high efficiency,which can effectively reduce the difficulty and cost of fire truck access inspection.
作者
张鑫
陈黎
ZHANG Xin;CHEN Li(School of Computer Science and Technology,Wuhan University of Science and Technology,Wuhan 430065,China;Hubei Province Key Laboratory of Intelligent Information Processing and Real-Time Industrial System,Wuhan University of Science and Technology,Wuhan 430065,China)
出处
《计算机技术与发展》
2022年第10期51-57,64,共8页
Computer Technology and Development
基金
国家自然科学基金项目(61773297)。
关键词
语义分割
特征融合
注意力机制
消防车通道
占用检测
semantic segmentation
feature fusion
attention mechanism
fire truck passages
occupancy detection