复杂光照场景下基于MTCNN的人脸检测

来源 :南华大学学报(自然科学版) | 被引量 : 0次 | 上传用户:lyh682020
下载到本地 , 更方便阅读
声明 : 本文档内容版权归属内容提供方 , 如果您对本文有版权争议 , 可与客服联系进行内容授权或下架
论文部分内容阅读
为了提高复杂光照条件下的人脸检测识别率,提出了一种基于Retinex图像增强技术应用于多任务卷积神经网络(multi-task cascaded convolutional networks, MTCNN)的人脸测算法。算法用Retinex理论对图像进行增强,能明显提高MTCNN在不同光照场景下的人脸检测精度,同时使面部五个关键点的定位更准确。实验证明,在复杂光照场景下,该方法比原始MTCNN网络的人脸检测具有更好的效果,有利于后期的人脸对齐及分类任务。
其他文献
Hetero-nanostructures of plasmonic metals and semiconductors have attracted increasing attention in the field of photocatalysis.However,most of the hetero-nanos
Full-spectrum underwater optical communication (UOC) is of great significance for major strategic needs including resource development,scientific exploration,an
We employ thermoreflsctance thermal imaging to directly measure the steady-state two-dimensional (2D) temperature field generated by nanostructured heat sources
NiFa layered double hydroxide (NiFe-LDH) nanosheets and metal-nitrogen-carbon materials (M-N-C,M =Ni,Fe,Co,etc.) are supreme catalysts in the oxygen evolution r
How to regulate the supramolecular structures in the assembly of graphene quantum dots (GQDs) is still a great challenge to be overcome.Herein,the GQDs of 1-3 l
Metal-oxide-semiconductor field effect transistors (MOSFET) based on two-dimensional (2D) semiconductors have attracted extensive attention owing to their excel
Since Moore's law in the traditional semiconductor industry is facing shocks,More Moore and More than Moore are proposed as two paths to maintain the developmen
Two-dimensional (2D) tungsten disulfide (WS2) has emerged as a promising ultrathin semiconductor for building high-performance nanoelectronic devices.The contro
Molybdenum disulfide (MoS2) holds great promise as atomically thin two-dimensional (2D) semiconductor for future electronics and opto-alectronics.In this report
All-norganic cesium lead halide based perovskite nanocrystals (PNCs) exhibit promising optoelectronic properties,but their poor stability and anion exchange rea