We are a manufactor of Pipe Industry Videoscope,which are widely used in Power,Petrochemical,Aviation,Police Counter-terrorism,Nuclear industry,Customs,Military,Shipbuilding,Automobile and Locomotive manufacturing,Heavy industry
equipment,and so on.We have rich experience in NDT field,and we do not only provide the best performance pipe Industry Videoscope but also offer the best after-sales service,immediatly repairing (
1 week to 3 weeks ),12 months for warranty freely.And our goals are customers
are the first,to solve problems for our customers is our duty. Pipe Industry Videoscope Pipe Industry Videoscope,Pipe Inspection Videoscope,Portable Industry Videoscope,Flexible Industry Videoscope Zhengzhou Runde Dellonscope Co.,Ltd. , https://www.china-borescope.com
Aerial detection widely uses Dual-NMS algorithm to enhance detection capability
[ Instrument Network Instrument R & D ] Recently, the Shenyang Institute of Automation, Chinese Academy of Sciences has innovatively proposed a method to remove false detection targets from aerial image detection results, and designed a corresponding detection network for aerial image characteristics. Related results were published on Sensors.
Aerial photography is also called aerial photography, aerial photography or aerial photography, which refers to taking pictures of the earth's landform from the air and obtaining a top view. This picture is an aerial photograph. Aerial cameras can be controlled by photographers, and can also be shot automatically or remotely. The platforms used for aerial photography include airplanes, helicopters, hot air balloons, small airships, rockets, kites, parachutes, etc.
Aerial image-based target detection technology has been widely used in civil and military fields, such as city monitoring, border inspection, population density estimation, etc. How to effectively extract specific targets from aerial images has become a current research hotspot. The detection technology of cars and pedestrians in aerial images has gradually matured, but the detection of building targets has always been a difficult problem, and there is no good solution. This is because the shape of the building is complex, and it is easy to be disturbed by the background of roads, cars, and vegetation. Moreover, the aerial image is affected by lighting conditions and shooting angles, which will cause differences in image resolution and quality, which increases the difficulty of detection.
Target detection methods based on deep learning have been studied in aerial image target detection, but due to the inherent characteristics of aerial images, the results of existing detection methods are not good, and there are many cases of false detection and missed detection. Most of the existing detection methods focus on improving the detection capability of the detection network to the target.
Target detection, also called target extraction, is an image segmentation based on the geometric and statistical characteristics of the target. It combines target segmentation and recognition into one. Its accuracy and real-time performance is an important capability of the entire system. Especially in complex scenes, when multiple targets need to be processed in real time, automatic target extraction and recognition becomes particularly important.
With the development of computer technology and the widespread application of computer vision principles, the use of computer image processing technology for real-time tracking of targets has become more and more popular. Dynamic real-time tracking and positioning of targets is in intelligent transportation systems, intelligent monitoring systems, and military target detection. It also has a wide range of applications in surgical instrument positioning in medical navigation surgery.
From the perspective of removing false detection targets from the detection results, this research innovatively proposes the Dual-NMS algorithm. By counting the density of the detection frames generated around each detected target and the corresponding classification confidence, it can automatically remove the detection results. The false detection target in the method greatly improves the accuracy of the detection result. In the research, the judgment result of the false detection target is further used as a constraint for network training, which fundamentally strengthens the detection network's ability to extract the abstract features of the target to be detected.
In addition, the research also introduces the correlation between targets into the detection network, which enhances the method's ability to detect targets. Experimental results show that Dual-NMS can remove nearly 50% of false detection targets, and the optimized detection network performs better than existing algorithms.
Source: Shenyang Institute of Automation, Baidu Academic, Encyclopedia