基于视频图像智能解译与地理定位技术的耕地保护实时监测研究

Research on Real-time Monitoring of Cultivated Land Protection Based on Intelligent Interpretation of Video Image and Geographical Positioning Technology

近年来,湖南通过建立卫星监测体系在耕地保护方面取得明显成效。但由于卫星遥感存在容易受到云雨天气影响、获取周期间隔较长、分辨率不够高等缺陷,在耕地保护实时、动态、精准、高效监测尤其是非粮化监测方面,存在一定不足。铁塔视频可作为对卫星监测的有效补充,目前,湖南已经建成铁塔视频监测站点1万座,每天实时产生海量视频与图片信息,依靠人工目视察看视频发现线索问题带来的工作量超出承受能力,基于视频图像的智能解译与定位技术自动识别涉及耕地的疑似违法目标,是开展铁塔视频监测的关键核心技术,亟须开展攻关突破。本文提出一套“基于样本和卷积神经网络算法实现视频图像目标智能解译与定位技术的视频应用”方案,并对其试验结果进行分析阐述,为行业内利用铁塔视频开展耕地实时监测技术研究提供思路和参考。

In recent years,Hunan has achieved remarkable results in farmland protection through the establishment of satellite monitoring system.However,due to the shortcomings of satellite remote sensing,such as being susceptible to cloud and rain weather,long acquisition cycle interval and low-resolution,there are certain shortcomings in real-time,dynamic,accurate and efficient monitoring of cultivated land protection,especially non-grain monitoring.Tower videos can serve as an effective supplement to satellite monitoring.At present,Hunan has built 10,000 tower videos monitoring stations.Massive videos and pictures information in real time were produced every day.Workload Relying on manual visual inspection of the video to find the clues of the problem is beyond the capacity.Intelligent interpretation technology based on video images identifies automatically the suspected illegal targets involving cultivated land,which is the core and key technology to carry out iron tower video monitoring,and it is urgent to make breakthroughs.In this paper,a set of scheme of "intelligent interpretation of video image target based on sample and convolutional neural network algorithm" is proposed,then analyzes and expounds the test results,which provides ideas and reference for the industry to carry out real-time monitoring technology by using tower video.