摘要
国际机场货运区空间广阔、货物种类繁多,加大了监控难度。为此,提出深度学习算法下国际机场货运区全景监控方法。利用灰度级累积概率密度函数对国际机场货运区图像进行划分与校正,通过均衡化映射变换和图像融合技术,获得增强后的清晰图像。建立卷积神经网络模型,通过特征提取和融合,获取货运区目标的特征信息,实现准确的货运目标检测。结合卡尔曼滤波与蒙特卡洛算法,实现对目标的连续、稳定跟踪,完成国际机场货运区的全景监控。仿真结果表明,所提方法灰度直方图分布较均匀,跟踪框与检测框之间的重叠度较高,货运目标检测IOU值高于96.0%,全景监控效果较好,可以在监控过程中可高精度的实现目标检测与跟踪。
The International airport freight area has a wide space and a wide variety of goods,which increases the difficulty of monitoring.Therefore,a panoramic monitoring method of the international airport freight area based on a deep learning algorithm is proposed.The gray level cumulative probability density function is used to divide and correct the image of the international airport freight area,and an enhanced,clear image is obtained through the equalization mapping transformation and image fusion technology.The convolution neural network model is established,and the feature information of the cargo area target is obtained through feature extraction and fusion to achieve accurate cargo target detection.Combining the Kalman filter and the Monte Carlo algorithm,the continuous and stable tracking of the target is realized,and the panoramic monitoring of the international airport freight area is completed.The simulation results show that the gray histogram of the proposed method is evenly distributed,the overlap between the tracking frame and the detection frame is high,the I0U value of cargo target detection is higher than 96.0%,the panoramic monitoring effect is good,and the target detection and tracking can be achieved with high precision in the monitoring process.
作者
李雪晖
贺小辉
LI Xue-huil;HE Xiao-hui(Engineering Construction Command Headquarters of Guangdong Airport Management Group Co.,Ltd,Guangdong Guangzhou 510000,China;Guangdong University of Technology,Guangdong Guangzhou 510006,China)
出处
《计算机仿真》
2025年第10期81-85,共5页
Computer Simulation
基金
广东省科技计划基金资助项目(2022b010101030)。
关键词
深度学习
直方图
国际机场货运区
卷积神经网络模型
全景监控
Deep learning
histogram
International airport cargo area
Convolutional neural network model
Pano-ramic monitoring
