基于多元融合网络道路视频监测技术仿真研究

Simulation research on road video monitoring technology based on multivariate fusion network

为加强高速公路运行智能监测能力,提出了一种基于多元融合网络的不良气象下视频道路监测提升技术,用于提高在雨、雪、雾、霾等不良气象下针对道路车辆运行状态进行视频监测的能力。首先利用不同的气象场景数据建立气象识别网络模型,利用改进的暗通道算法及去雨雪算法识别出的不同气象环境下图像进行增强处理,最后利用改进的Fast r-cnn网络模型对车辆进行识别监测,从而构建了一种基于多元融合网络对不良气象下的道路视频监测系统。通过图像数据24000张及随机未经过图像增强的图片8000张,利用多元融合网络与传统神经网络方法进行实验对比,多元融合网络在不同气象下的视频监测能力有效提升10%到15%左右,从而验证了多元融合网络在不良气象下视频监测提升的有效性。

In order to strengthen the intelligent monitoring ability of expressway operation,a video monitoring promotion technology based on multi fusion network in bad weather is proposed,which is used to improve the ability of video monitoring of road vehicle operation state in bad weather such as rain,snow,fog and haze. Firstly,the network model of weather recognition is established by using different weather scene data,and the images of different weather environments identified by improved dark channel algorithm and snow removal algorithm are enhanced. Finally,the improved fast r-cnn network model is used to identify and monitor vehicles,and a video monitoring system based on multi fusion network for bad weather is constructed. Through 24000 image data and 8000 random images without image enhancement,the multi fusion network is compared with the traditional neural network method,and the video monitoring ability of the multi fusion network in different weather is effectively improved by 10% to 15%,which verifies the effectiveness of the multi fusion network in bad weather video monitoring.