Multi-Sensor Face Registration Based on Global and Local Structures

Li, Wei; Dong, Mingli; Lü, Naiguang; Lou, Xiaoping; Zhou, Wanyong

doi:10.3390/app9214623

Cited by 5 publications

(4 citation statements)

References 34 publications

(45 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, the structure of FaceNet was not reduced in their work, and its computational burden was still heavy and therefore not suitable for deployment in embedded devices. For most similar studies using multiple spectral images, such as visible and IR images, image registration is the main issue, as discussed in [18,19]. Images from multiple sensors must be fused before applying face recognition.…”

Section: Related Workmentioning

confidence: 99%

Lightweight and Resource-Constrained Learning Network for Face Recognition with Performance Optimization

Deng

Chiang

2020

Sensors

View full text Add to dashboard Cite

Despite considerable progress in face recognition technology in recent years, deep learning (DL) and convolutional neural networks (CNN) have revealed commendable recognition effects with the advent of artificial intelligence and big data. FaceNet was presented in 2015 and is able to significantly improve the accuracy of face recognition, while also being powerfully built to counteract several common issues, such as occlusion, blur, illumination change, and different angles of head pose. However, not all hardware can sustain the heavy computing load in the execution of the FaceNet model. In applications in the security industry, lightweight and efficient face recognition are two key points for facilitating the deployment of DL and CNN models directly in field devices, due to their limited edge computing capability and low equipment cost. To this end, this paper provides a lightweight learning network improved from FaceNet, which is called FN13, to break through the hardware limitation of constrained computational resources. The proposed FN13 takes the advantage of center loss to reduce the variations of the between-class features and enlarge the difference of the within-class features, instead of the triplet loss by using FaceNet. The resulting model reduces the number of parameters and maintains a high degree of accuracy, only requiring few grayscale reference images per subject. The validity of FN13 is demonstrated by conducting experiments on the Labeled Faces in the Wild (LFW) dataset, as well as an analytical discussion regarding specific disguise problems.

show abstract

Section: Related Workmentioning

confidence: 99%

Lightweight and Resource-Constrained Learning Network for Face Recognition with Performance Optimization

Deng

Chiang

2020

Sensors

View full text Add to dashboard Cite

show abstract

“…To recognize such target with multiple categories and rich morphologies accurately, it is also necessary to choose the same type of features which maintain high stability in different growth periods and environments and conform to subjective understanding of human vision. In this study, two types of level-2 features, namely spine and cingulum were chosen [21][22]. The domain specific features were extracted by defining new feature formalized description.…”

Section: Feature Extractionmentioning

confidence: 99%

Research on Imbalanced Microscopic Image Classification of Harmful Algae

Qiao

2020

IEEE Access

View full text Add to dashboard Cite

Image analysis based on biological morphological differences is an important development direction for classification and determination of planktonic algae. However, it has some shortages, such as high degree of sample imbalance and difficult to have formalized description of local physiological features. To overcome these shortages, this study decomposed recognition of harmful algae microscopic images into sample supplementation, accurate segmentation, feature extraction and classification and identification. Firstly, sample imbalance is solved by Kernel-ADASYN method to generate enough samples. Target cells are separated through integration of multi-directional projections. Refined segmentation between the spine and cingulum detail regions is further realized. Later, effective feature extraction and description of global and local features were performed one by one by matching physiological features of algae with machine recognition features. Finally, the SVM model was applied for multi-class recognition. Results demonstrated that the proposed method can reduce imbalance rate of sample size and realize multi-class recognition of microscopic images of 15 categories of algae cells. INDEX TERMS Imbalanced classification, microscopic image recognition, multi-level features extraction LIST OF ABBREVIATIONS Kernel-ADASYN Kernel based adaptive synthetic CCD Charge coupled device GLCM Gray-level co-occurrence matrix ASM Angular second moment SVM Support vector machine PCA Principal component analysis RBF Radial base function

show abstract

“…In the paper entitled 'Multi-Sensor Face Registration Based on Global and Local Structures', Wei Li 1, Mingli Dong, Naiguang Lu, Xiaoping Lou, and Wanyong Zhou [15] introduced a novel multi-sensor face image registration method. This work utilizes global geometrical relationships and local shape features to register visible and infrared (IR) facial images for fusion-based recognition.…”

Section: Registration and Fusionmentioning

confidence: 99%