摘要
针对当前野外搜救效率低下、搜救效果易受环境因素(如夜间)影响等问题,设计制作了一种采用红外视觉的野外搜救无人机。搜救无人机通过机载的红外热像仪FLIR TAU2对搜救区域进行图像采集;将拍摄的红外图像传输给机载的TK1实现对红外图像中人体目标的实时检测;无人机将检测到的人体目标的地理位置发送给地面站,以便搜救人员能够及时完成救援任务。在红外人体目标检测阶段,采用方向梯度直方图(HOG)对红外目标进行特征描述,并通过线性支持向量机(SVM)分类器对HOG特征分类检测。试验结果验证了所设计的搜救无人机可以在多种真实野外场景下完成野外搜救任务。
Aiming at problems that efficiency of current outdour search and rescue is low and the effect of search and rescue is susceptible to environmental factors,such as nighttime,an outdoor search and rescue unmanned aerial vehicle( UAV) using infrared vision is designed and manufactured. Search and rescue UAV uses onboard infrared camera FLIR TAU2 to collect images from the search and rescue area. Captured infrared images are transmitted to onboard TK1 for real-time detection of human targets in infrared image. UAV transmits the detected geographic location of the human target to the ground station,so that the search and rescue personnel can complete the rescue mission in time. In the stage of infrared human target detection,histogram of oriented gradient( HOG) is used to characterize the infrared target,and the HOG feature classification detection is performed by linear support vector machine( SVM) classifier. The test results verify that the designed search and rescue UAV can complete the outdoor search and rescue mission in a variety of real field scenes.
作者
邵延华
张得沛
楚红雨
常志远
展华益
饶云波
SHAO Yan-hua;ZHANG De-pei;CHU Hong-yu;CHANG Zhi-yuan;ZHAN Hua-yi;RAO Yun-bo(School of Information Engineering,Southwest University of Science and Technology,Mianyang 621000,China;Electrical Engineering and Computer Science,Northwestern University,Evanston,IL 60208,USA;Information and Software Engineering,University of Eleeteonic Science&Technology of China,Chengdu 610054,China)
出处
《传感器与微系统》
CSCD
2019年第1期104-106,109,共4页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(61601382)
四川省教育厅项目(17ZB0454)
西南科技大学龙山人才项目(17LZX682
18LZX632)
西南科技大学博士研究基金资助项目(16ZX7148)
关键词
红外视觉
野外搜救
方向梯度直方图
支持向量机
人体目标检测
无人机
infrared vision
outdoor search and rescue
histogram of oriented gradient (HOG)
support vector machine (SVM)
human body target detection
unmanned aerial vehicle(UAV)
