激光技术
激光技術
격광기술
Laser Technology
2015年
6期
840-844
,共5页
图像处理%目标检测%脉冲耦合神经网络%高斯混合模型%红外序列图像
圖像處理%目標檢測%脈遲耦閤神經網絡%高斯混閤模型%紅外序列圖像
도상처리%목표검측%맥충우합신경망락%고사혼합모형%홍외서렬도상
image processing%target detection%pulse coupled neural network%Gaussian mixture model%infrared image se-quence
红外图像受随机噪声干扰严重。传统的基于高斯混合模型的检测算法检测得到的红外目标受虚假轮廓影响,不易准确辨识。为了准确识别红外目标,采用了一种基于脉冲耦合神经网络和高斯混合模型的红外目标检测算法。首先利用高斯混合模型定位红外目标区域的位置,然后利用基于空间信息的分水岭算法得到闭合区域,再利用基于脉冲耦合神经网络的分割算法剪切其虚影,最终检测到完整的运动目标。结果表明,该方法能够消除在传统方法中产生的虚影现象,得到精确的红外运动目标。通过比较,实验结果优于传统方法。
紅外圖像受隨機譟聲榦擾嚴重。傳統的基于高斯混閤模型的檢測算法檢測得到的紅外目標受虛假輪廓影響,不易準確辨識。為瞭準確識彆紅外目標,採用瞭一種基于脈遲耦閤神經網絡和高斯混閤模型的紅外目標檢測算法。首先利用高斯混閤模型定位紅外目標區域的位置,然後利用基于空間信息的分水嶺算法得到閉閤區域,再利用基于脈遲耦閤神經網絡的分割算法剪切其虛影,最終檢測到完整的運動目標。結果錶明,該方法能夠消除在傳統方法中產生的虛影現象,得到精確的紅外運動目標。通過比較,實驗結果優于傳統方法。
홍외도상수수궤조성간우엄중。전통적기우고사혼합모형적검측산법검측득도적홍외목표수허가륜곽영향,불역준학변식。위료준학식별홍외목표,채용료일충기우맥충우합신경망락화고사혼합모형적홍외목표검측산법。수선이용고사혼합모형정위홍외목표구역적위치,연후이용기우공간신식적분수령산법득도폐합구역,재이용기우맥충우합신경망락적분할산법전절기허영,최종검측도완정적운동목표。결과표명,해방법능구소제재전통방법중산생적허영현상,득도정학적홍외운동목표。통과비교,실험결과우우전통방법。
Infrared images are usually interfered by random noise seriously .Infrared targets detected by the traditional detection algorithm based on Gaussian mixture model are difficult to be identified because of false contour .In order to identify the infrared target accurately , an infrared target detection algorithm based on pulse coupled neural network ( PCNN) and Gaussian mixture model was proposed .Firstly, Gaussian mixture model was used to locate the approximate location of moving targets .And then, a closed region was obtained by using watershed algorithm based on spatial information .Segmentation algorithm based on PCNN was used to shear the pseudo-target and the complete moving target was detected .The experimental results show that this method can eliminate the pseudo target of the traditional methods and detect the infrared moving targets accurately .The new algorithm is superior to the other conventional algorithms .