摘要
针对实际采集的视频中背景复杂,人物多变,图像处理时间长,训练样本不足的问题,提出了构造虚拟样本,并结合Gabor滤波器及对PCA-LDA算法加以改进的人脸识别算法,以应用于教室点名系统。首先对教室采集到的视频进行裁剪,按帧截取并检测出含有人脸的部分图像并单独保存为测试图像,然后将其与已有人脸库里的训练图像进行对比,最后采用提出的镜像法构造虚拟样本,并结合了Gabor滤波器以及PCNN灰度图像增强处理算法的改进PCA-LDA算法进行人脸识别。仿真实验表明,提出的算法预测了样本可能存在的变化,也在一定程度上降低了计算复杂度,明显地提高了识别率,并在教室点名系统中得到了较好的验证。
Aiming at the problems in the actual collection of video such as the complex background, changeable characters, long image processing time and the lack of training samples, the improved face recognition algorithm is proposed based on constructing virtual samples and the combination of Gabor filter and PCA-LDA algorithm, in order to be applied to the classroom attendance system. First of all, the video collected in the classroom is cut, and part of the images are captured by the frame and saved as test images. Then it compares the collected test images with the training images in the existing face database, finally usies the proposed algorithm which uses the mirror method to construct virtual samples and combines with improved PCA-LDA algorithm based on Gabor filter and PCNN gray image enhancement algorithm for face recognition. Simulation results show that the proposed algorithm predicts the possible changes in samples, and to some extent, the computational complexity is reduced, the recognition rate is improved obviously, and it is well verified in the classroom attendance system.
出处
《计算机工程与应用》
CSCD
北大核心
2017年第23期123-128,共6页
Computer Engineering and Applications
基金
国家自然科学基金(No.61672265)
江苏省产学研联合创新资金项目(No.BY2013015-35)
关键词
构造虚拟样本
教室点名系统
改进主成分分析-线性判别分析(PCA-LDA)算法
人脸识别
constructing virtual samples
classroom attendance system
improved Principal Component Analysis-Linear Discriminant Analysis(PCA_LDA)algorithm
face recognition
