视频监控中的行人计数是人体目标检测的重要应用,针对该应用对实时性的要求,提出了将混合高斯背景模型和HOG(histograms of oriented gradients)特征结合的方法。该方法首先通过混合高斯背景方法检测出视频中的运动前景,再用HOG+SVM对...视频监控中的行人计数是人体目标检测的重要应用,针对该应用对实时性的要求,提出了将混合高斯背景模型和HOG(histograms of oriented gradients)特征结合的方法。该方法首先通过混合高斯背景方法检测出视频中的运动前景,再用HOG+SVM对提取的感兴趣区域内的行人进行计数,大大降低了计算量,提高了HOG算法的实时性。经试验证明,改进的算法可以将行人计数的准确度提高到95%左右。展开更多
The UAVs (unmanned aerial vehicles) and the use of graphical data files of extension KML (keyhole markup language) have become popular nowadays. The KML files are associated to the Google Earth applications. Devel...The UAVs (unmanned aerial vehicles) and the use of graphical data files of extension KML (keyhole markup language) have become popular nowadays. The KML files are associated to the Google Earth applications. Developed by Microdrone company for planning, overflight simulation and data flight analysis, the program mdCockpit was used in this experiment. The UAV Microdrone's flight path was programmed in two ways: using geo referenced images from Google Earth and entering points coordinates into the program. These points have had their geographical decimal coordinates collected by GPS (global positioning system) (RTK (real time kinematic)) method. The aim here is to describe a way of evaluating the difference between the alternatives of flight planning. A flight path, defined a route with 22 points, was also simulated in the program mdCockpit. The points were collected and saved, while the images captured by the program went automatically forming a georeferenced mosaic. A comparison between the differences of coordinates to each point was made by choosing on the images and also by a RTK positioning. The result was up to 14.20 m to a 100 m flight height. There is a significant change of position and this can result in unwanted over flights in locations or even jeopardizing the safety of air activity.展开更多
文摘视频监控中的行人计数是人体目标检测的重要应用,针对该应用对实时性的要求,提出了将混合高斯背景模型和HOG(histograms of oriented gradients)特征结合的方法。该方法首先通过混合高斯背景方法检测出视频中的运动前景,再用HOG+SVM对提取的感兴趣区域内的行人进行计数,大大降低了计算量,提高了HOG算法的实时性。经试验证明,改进的算法可以将行人计数的准确度提高到95%左右。
文摘The UAVs (unmanned aerial vehicles) and the use of graphical data files of extension KML (keyhole markup language) have become popular nowadays. The KML files are associated to the Google Earth applications. Developed by Microdrone company for planning, overflight simulation and data flight analysis, the program mdCockpit was used in this experiment. The UAV Microdrone's flight path was programmed in two ways: using geo referenced images from Google Earth and entering points coordinates into the program. These points have had their geographical decimal coordinates collected by GPS (global positioning system) (RTK (real time kinematic)) method. The aim here is to describe a way of evaluating the difference between the alternatives of flight planning. A flight path, defined a route with 22 points, was also simulated in the program mdCockpit. The points were collected and saved, while the images captured by the program went automatically forming a georeferenced mosaic. A comparison between the differences of coordinates to each point was made by choosing on the images and also by a RTK positioning. The result was up to 14.20 m to a 100 m flight height. There is a significant change of position and this can result in unwanted over flights in locations or even jeopardizing the safety of air activity.