舰船科学技术
艦船科學技術
함선과학기술
SHIP SCIENCE AND TECHNOLOGY
2015年
4期
219-222
,共4页
神经网络%BP算法%图像识别%图像跟踪
神經網絡%BP算法%圖像識彆%圖像跟蹤
신경망락%BP산법%도상식별%도상근종
neural network%BP algorithm%image recognition%image tracking
本文利用BP神经网络抗干扰性强,识别精准等优点对船舶进行识别跟踪。首先获取原始图像,然后预处理,以图像的全部灰度值为训练样本,以新不变矩特征向量为样本集输入到3层BP神经网络中,对含不同噪声均值的图像进行识别。实验结果表明,以新不变矩特征向量作为样本集时抗噪能力强,识别率高。最后以新不变矩特征向量作为样本集进行目标跟踪得到跟踪误差。
本文利用BP神經網絡抗榦擾性彊,識彆精準等優點對船舶進行識彆跟蹤。首先穫取原始圖像,然後預處理,以圖像的全部灰度值為訓練樣本,以新不變矩特徵嚮量為樣本集輸入到3層BP神經網絡中,對含不同譟聲均值的圖像進行識彆。實驗結果錶明,以新不變矩特徵嚮量作為樣本集時抗譟能力彊,識彆率高。最後以新不變矩特徵嚮量作為樣本集進行目標跟蹤得到跟蹤誤差。
본문이용BP신경망락항간우성강,식별정준등우점대선박진행식별근종。수선획취원시도상,연후예처리,이도상적전부회도치위훈련양본,이신불변구특정향량위양본집수입도3층BP신경망락중,대함불동조성균치적도상진행식별。실험결과표명,이신불변구특정향량작위양본집시항조능력강,식별솔고。최후이신불변구특정향량작위양본집진행목표근종득도근종오차。
In this paper, because of strong interference and precise identification, etc. , use BP neural network to identification and tracking of ships. First, get the original image, then pretreated. Respectively, all of grayscale images is training samples and a new moment invariant feature vector for the sample set input to the BP neural network. Use BP neural network to identify images with different noise. Experimental results show that when a new moment invariant feature vectors as the sample set has strong noise immunity and high recognition rate. Finally, a new moment invariant feature vectors as the sample set for target tracking to get the tracking error.