摘要
针对当前卷积神经网络算法在近红外光人脸检测时,普遍存在着检测时间过长、检测性能较低等问题。提出一种基于粒子优化群算法和卷积神经网络的人脸检测方法。通过对近红外光人脸进行分析,在人脸图像预处理中对人脸进行近红外光光照强度补偿,利用小波变换方法提取人脸图像低频子带,引入主特征分析法提取出近红外光人脸图像特征分量,将神经网络参数进行初始化,确定其拓扑结构。将提取的特征向量输入神经网络中,引入粒子群优化算法对神经网络进行优化,完成卷积神经网络算法在近红光中的人脸检测。实验结果表明,本文方法在近红外光人脸检测中检测时间较短,检测性能较高。
In view of the current convolutional neural network algorithm in the near-infrared face detection, there are generally problems such as long detection time and low detection performance. A face detection method based on particle optimization group algorithm is proposed. By analyzing the face of near-infrared light, the face infrared image is compensated for the near-infrared light intensity in the face image preprocessing. The wavelet transform method is used to extract the low-frequency sub-band of the face image, and the main feature analysis method is introduced to extract the near-infrared. The light facial image feature component initializes the neural network parameters and determines its topology. The extracted feature vector is input into the neural network , and the particle swarm optimization algorithm is introduced to optimize the neural network, and the face detection of the convolutional neural network algo rithm in near red light is completed. The experimental results show that the proposed method has shorter detection time and higher detection performance in near-infrared face detection.
作者
王玉晶
莫建麟
WANG Yujing;MO Jianlin(Dean's office of ABA Teachers University, Wenchuan Sichuan 623002, China)
出处
《激光杂志》
北大核心
2019年第4期180-183,共4页
Laser Journal
基金
四川省教育厅课题:(No.ASA18-01
No.17ZA0002)
关键词
神经网络
光照强度
人脸检测
检测性能
convolutional neural network
near-infrared light
face detection
detection performance
