太赫兹科学与电子信息学报
太赫玆科學與電子信息學報
태혁자과학여전자신식학보
Information and Electronic Engineering
2013年
5期
712-717
,共6页
史阳阳%庄丽葵%曹云峰%王彪
史暘暘%莊麗葵%曹雲峰%王彪
사양양%장려규%조운봉%왕표
无人机(UAV)%数字信号处理器(DSP)%目标识别%仿射不变矩%支持向量机(SVM)
無人機(UAV)%數字信號處理器(DSP)%目標識彆%倣射不變矩%支持嚮量機(SVM)
무인궤(UAV)%수자신호처리기(DSP)%목표식별%방사불변구%지지향량궤(SVM)
Unmanned Aerial Vehicle%Digital Signal Processor%target recognition%affine moment invariants%Support Vector Machine
针对复杂海面环境造成的图像干扰变形给无人机着舰过程中人工标识的识别带来的困难，采用了一种基于仿射不变矩和支持向量机(SVM)的识别方法，首先采集各种变形下的适量标识样本，经过适当的预处理之后，提取其仿射不变矩，然后送入 SVM进行训练识别，并且给出了在 DSP 系统内基于上述流程的实现方法。实验表明，该方案对于图像在各种小幅度干扰变形下的识别率均达到了90%以上，平均每帧测试时间为170 ms，对于旋翼型无人机自主着舰具有一定的实用性。
針對複雜海麵環境造成的圖像榦擾變形給無人機著艦過程中人工標識的識彆帶來的睏難，採用瞭一種基于倣射不變矩和支持嚮量機(SVM)的識彆方法，首先採集各種變形下的適量標識樣本，經過適噹的預處理之後，提取其倣射不變矩，然後送入 SVM進行訓練識彆，併且給齣瞭在 DSP 繫統內基于上述流程的實現方法。實驗錶明，該方案對于圖像在各種小幅度榦擾變形下的識彆率均達到瞭90%以上，平均每幀測試時間為170 ms，對于鏇翼型無人機自主著艦具有一定的實用性。
침대복잡해면배경조성적도상간우변형급무인궤착함과정중인공표식적식별대래적곤난，채용료일충기우방사불변구화지지향량궤(SVM)적식별방법，수선채집각충변형하적괄량표식양본，경과괄당적예처리지후，제취기방사불변구，연후송입 SVM진행훈련식별，병차급출료재 DSP 계통내기우상술류정적실현방법。실험표명，해방안대우도상재각충소폭도간우변형하적식별솔균체도료90%이상，평균매정측시시간위170 ms，대우선익형무인궤자주착함구유일정적실용성。
Deformation of the image caused by interference for the complex environment of the sea can bring some difficulty to the UAV’s recognition of the artificial sign in the process of landing on a ship. A recognition method based on affine invariant moments and Support Vector Machine(SVM) is proposed. Firstly, some artificial sign samples under various deformation are collected. After preprocessing, affine invariant moments are extracted, and then are put into SVM classifier to complete artificial sign recognition. Finally,the realization based on the method within Digital Signal Processor(DSP) system is given. Experimental results show that the recognition rate is as high as 90% and the average frame test time is about 170 ms for the images in a variety of small amplitude distortion. So it exhibits a certain practicality for autonomous landing of Rotor UAV.
