Showing papers on "Three-dimensional face recognition published in 2019"

Face Alignment in Full Pose Range: A 3D Total Solution

Abstract: Face alignment, which fits a face model to an image and extracts the semantic meanings of facial pixels, has been an important topic in the computer vision community. However, most algorithms are designed for faces in small to medium poses (yaw angle is smaller than 45 degree), which lack the ability to align faces in large poses up to 90 degree. The challenges are three-fold. First, the commonly used landmark face model assumes that all the landmarks are visible and is therefore not suitable for large poses. Second, the face appearance varies more drastically across large poses, from the frontal view to the profile view. Third, labelling landmarks in large poses is extremely challenging since the invisible landmarks have to be guessed. In this paper, we propose to tackle these three challenges in an new alignment framework termed 3D Dense Face Alignment (3DDFA), in which a dense 3D Morphable Model (3DMM) is fitted to the image via Cascaded Convolutional Neural Networks. We also utilize 3D information to synthesize face images in profile views to provide abundant samples for training. Experiments on the challenging AFLW database show that the proposed approach achieves significant improvements over the state-of-the-art methods.

Applications of Pattern Recognition

Abstract: applications of pattern recognition. Book lovers, when you need a new book to read, find the book here. Never worry not to find what you need. Is the applications of pattern recognition your needed book now? That's true; you are really a good reader. This is a perfect book that comes from great author to share with you. The book offers the best experience and lesson to take, not only take, but also learn.

Three-dimensional face recognition using variance-based registration and subject-specific descriptors:

Abstract: Face recognition underpins numerous applications; however, the task is still challenging mainly due to the variability of facial pose appearance. The existing methods show competitive performance b...

5.1 A Stacked Global-Shutter CMOS Imager with SC-Type Hybrid-GS Pixel and Self-Knee Point Calibration Single Frame HDR and On-Chip Binarization Algorithm for Smart Vision Applications

Abstract: Request for smart vision related applications, such as face identification, VR/AR, gesture recognition, 3D imaging, and artificial intelligence (AI), has driven demand for high-performance global-shutter (GS) sensors. Most commercially available GS sensors use a charge-domain storage gate implementation, which suffers from serious light leakage and leads to lower shutter efficiency. This situation worsens when using a BSI fabrication process [1]. In addition, the traditional frame-based or line-based HDR method utilizing multiple exposures adds motion artifact to fast-moving objects, which defeats the purpose of having a global shutter. Moreover, some smart vision applications such as QR 2D barcode scanners and 3D facial recognition with structured light method need image sensors to “read” a certain pattern and “understand” the information within. However, image sensors usually capture a full image that needs to be further transferred to and processed by a companion SoC. Higher resolution and increased complexity of the target pattern pose a growing challenge to transfer and process the entire image at real time, also the required high power consumption lowers handheld device’s battery life.

Various image enhancement and matching techniques used for fingerprint recognition system

Abstract: For biometric identification or verification fingerprint images are most popular due to their uniqueness in nature. Image Enhancement Techniques (IETs) plays a vital role in Fingerprint Recognition (FPR) System and IETs are one of the most important stages in FPR system. Fingerprint images suffer problems related to brightness, poor contrast and blurring due to noise and motion etc. Fingerprint images may be corrupted and degraded due to variation in environmental conditions, skin, pressure on the sensors, and various other impression conditions. To overcome these problems, IETs are used. The main aim of implementing IET to the input image so that the visual quality or information contents are more suitable for a specific application or automated image processing. The performance of FPR system relies on the matching techniques that depend on the input fingerprint image quality and algorithm used. Depending upon the matching process there are various FPR system matching techniques. Enhancing the fingerprint images by IETs provide more reliable feature extraction information for the matching process. This paper presents an overview of various IETs used for FPR system, then types, applications and role in FPR system for researchers. A method is proposed which uses these techniques for better performance of the FPR system by improving the quality of fingerprint images using IETs.

Pose-invariant features and personalized correspondence learning for face recognition

Abstract: In surveillance systems, face recognition plays an important role for human identification. In such systems, human faces are spatially unconstrained, which results in a significant change in pose, and face recognition becomes more challenging when only one frontal image of the face has been registered in the gallery. In this study, we attempt to solve the problem where only one frontal image of the face is registered in the gallery, and the probe faces are captured in unconstrained poses. The face likelihood is measured using pose-invariant features of scale-invariant feature transform (SIFT) and personalized correspondence learning method. A generic correspondence is first learned between the poses, and the pose-invariant SIFT is fulfilled by extracting the keypoints on virtual patches that are generated by a generic correspondence with the pose variation. The generic correspondence is further personalized to fit each subject, and the learning error of the personalized correspondence is combined with pose-invariant SIFT to measure the face likelihood. The experimental results indicated that our proposed algorithm achieved an average performance of 95% across poses within $$40^\circ$$ , which is better than other well-known algorithms.

Undersampled face recognition based on virtual samples and representation classification

Abstract: Classifiers such as sparse representation or collaborative representation can get good performance in face recognition. But these methods require a number of train samples in each class to construct the dictionary. On the condition of undersampled train samples, their performance decreases dramatically. A novel method is proposed in this paper to address undersampled face recognition problem. Firstly, virtual face images are generated by principal component analysis and mirror transform. Secondly, the test sample is collaboratively represented by the augmented train samples and is recognized by classifier based on representation. A number of face recognition experiments on three benchmark face database show that the recognition accuracy of our method is greater than that of a similar method, while time efficiency of our method is competitive to similar method.

Anti-deception three-dimensional face recognition method based on information fusion

Abstract: The invention discloses an anti-deception three-dimensional face recognition method based on information fusion. The anti-deception three-dimensional face recognition method comprises: 1, collecting and processing color images and depth images of a plurality of real faces; 2, establishing Gaussian distribution models of a plurality of real faces according to the depth information of the key pointsin each depth map, and determining a threshold range of Gaussian distribution parameters of the real faces; 3, establishing a Gaussian distribution model of the to-be-recognized face, comparing and judging the Gaussian distribution model parameters of the to-be-recognized face with the threshold range of the Gaussian distribution parameters of the real faces, if the to-be-recognized face is judged to be the real face, executing the step 4, and otherwise, not performing face recognition; 4, constructing a deep convolutional neural network and training the deep convolutional neural network; and5, inputting the to-be-recognized face image into the trained deep convolutional neural network for recognition, and outputting a recognition result. By analyzing and modeling the face depth information and fusing the face depth information at the data end, a lightweight network is constructed, and the performance of the whole face recognition system is improved.

3D facial recognition intelligent lock and 3D face unlock method

Abstract: The invention discloses a 3D facial recognition intelligent lock and 3D face unlock method. The 3D face unlock method comprises: activating a depth camera to capture an image; determined whether thereis an unlocking intention when the image includes face information; identifying the face information when the unlocking intention is determined; opening the lock cylinder when successfully identified. The invention also discloses a 3D facial recognition intelligent lock, comprising a lock cylinder, a depth camera used for capturing images, and a processor. The processor is connected with the depth camera and the lock cylinder, is used for judging the unlocking intention according to the image and face recognition, and controlling the opening of the lock cylinder according to the recognition result.

Gait recognition method based on global and local feature fusion

Abstract: The invention provides a gait recognition method based on global and local feature fusion. The method comprises the following steps of training global gait features from a standardized contour map bya three-dimensional convolution neural network; training the local gait features from the local contour map by three-dimensional convolution neural network; serially fusing the global gait feature andthe local gait feature by a serial combination mode to obtain a combined gait feature, and utilizing the combined gait feature for gait recognition. The method of the invention utilizes the global and local feature fusion to derive the best feature subset, establishes gait feature extraction model, and uniformly solves the problem that the feature set in the existing gait identification method isnot enough valuable, so that better gait identification result can be obtained.

Car anti-theft system based on synchronous facial recognition and iris recognition

Abstract: The invention discloses a car anti-theft system based on synchronous facial recognition and iris recognition. The system has a 3D facial recognition function and an iris recognition function. Structural light is emitted to the face of a driver through a dot array projecting device by a 3D facial recognition function, a 3D model of the face of the driver is obtained through reconstruction after animage is captured through an infrared camera and the position where the light dot array takes place on the face of the driver and impacting by the structure is changed and distorted, the 3D model is compared with a primitively saved 3D model of a car owner, and facial coincidence degree information is obtained. The system has the advantages that an infrared ray is emitted to the eyes of the driverthrough an infrared emitter by the iris recognition function, an iris image is collected by an iris collecting assembly, and the iris coincidence degree information is obtained after the comparison computation with primitively saved iris feature image codes of the car owner is conducted; an unlocking control signal is sent to a vehicle-mounted driving computer only when the iris coincidence information and the facial coincidence information are greater than preset threshold values, and a car is prevented from being stolen.

Three-dimensional Face Recognition based on Feature Points Compression and Expansion

Abstract: Many researchers have attempted to recognize three-dimensional faces using feature points extracted from two-dimensional facial photographs. However, due to the limit of flat photographs, it is very difficult to recognize faces rotated more than 15 degrees from original feature points extracted from the photographs. As such, it is difficult to create an algorithm to recognize faces in multiple angles. In this paper, it is proposed a new algorithm to recognize three-dimensional face recognition based on feature points extracted from a flat photograph. This method divides into six feature point vector zones on the face. Then, the vector value is compressed and expanded according to the rotation angle of the face to recognize the feature points of the face in a three-dimensional form. For this purpose, the average of the compressibility and the expansion rate of the face data of 100 persons by angle and face zone were obtained, and the face angle was estimated by calculating the distance between the middle of the forehead and the tail of the eye. As a result, very improved recognition performance was obtained at 30 degrees of rotated face angle.

Three-dimensional face recognition method, and model training method and device

Abstract: The embodiment of the invention provides a three-dimensional face recognition method, and a model training method and device. The three-dimensional face recognition method comprises the steps: obtaining the face point cloud data of a to-be-recognized face; obtaining a multi-channel image according to the face point cloud data; inputting the multi-channel image into a deep neural network to be trained, and extracting through the deep neural network to obtain face features; and outputting a face category predicted value corresponding to the to-be-identified face according to the face feature.

A three-dimensional face recognition method based on region segmentation

Abstract: The invention provides a three-dimensional face recognition method based on region segmentation, and the method comprises the following steps: 1, carrying out the preprocessing of all test faces and three-dimensional face models in a registration library, and mainly comprises the steps: carrying out the automatic detection of a nose tip point, carrying out the cutting of a three-dimensional face,carrying out the posture correction, and carrying out the point cloud downsampling; 2, dividing the three-dimensional face shape into a rigid region and a non-rigid region according to the degree of influence of expression change; 3, designing different feature description modes for the rigid region and the non-rigid region respectively, and calculating the similarity degree between the test modeland the features of each three-dimensional face model in the registration library; And 4, carrying out weighted fusion on the similarity of the rigid region and the non-rigid region, and finally realizing the recognition of the three-dimensional face. According to the face recognition method provided by the invention, the region segmentation method only uses one feature point of the nose tip point, so that the process of detecting a plurality of face feature points in the previous method is avoided, meanwhile, the dependence on the positioning accuracy of other feature points is reduced, andthe detection success rate is relatively high.

Three-dimensional face recognition method based on plane parameterization

Abstract: The invention discloses a three-dimensional face recognition method based on plane parameterization. The method comprises the following specific steps: firstly, automatically selecting an area containing most significant features of a human face; and then finding a base plane through the nose tip, and calculating a relative depth, namely slope denaturation; then, trangularizing and mapping the ROIin the range data into an isomorphic two-dimensional plane circle, reserving inherent geometric properties, and mapping the relative depth, and finally, applying the feature surface method to the mapping depth image to realize a recognition task. According to the method, the situation that a traditional face recognition method is very sensitive to changes of postures, illumination and expressionsis overcome, meanwhile, due to the fact that three-dimensional geometrical information is stored, more clues expressing the changes are processed, and a face model is rapidly and accurately established.

3D Face Recognition Analysis Using Random Forest

Abstract: Face recognition is an emerging field due to the technological advances in camera hardware and for its application in various fields such as the commercial and security sector. Although the existing works in 3D face recognition perform well, a similar experiment setting across classifiers is hard to find, which includes the Random Forest classifier. The aggregations of the classification from each decision tree are the outcome of Random Forest. This paper presents 3D facial recognition using the Random Forest method using the BU-3DFE database, which consists of basic facial expressions. The work using other classifiers such as Neural Network (NN) and Support Vector Machine (SVM) using a similar experiment setting also presented. As for the results, the Random Forest approach has yield 94.71% of recognition rate, which is an encouraging result compared to NN and SVM. In addition, the experiment also yields that fear expression is unique to each human due to a high confidence rate (82%) of subjects with fear expression. Therefore, a lower chance to be mistakenly recognized someone with a fear expression.

Hyper-parameter Determination of CNN Classifier for Head Pose Estimation of Three Dimensional Degraded Face Images

Abstract: This paper presents the evaluation of parameters for head pose estimation using Convolutional Neural Network (CNN) towards the degraded images. Head pose estimation is one of the important factor for three dimensional face recognition system. Due to its superiority, Convolutional Neural Network (CNN) has been used as a head pose estimator, however, its performance is significantly dropped when the input face images is exposed to noises. As the CNN comes with different choices of pooling layer, two different experimental setups are created with similar architecture and training condition but using a different type of pooling layer. After learning, the CNN are tested with another five different testing datasets to monitor the effects of various particular noises, such as: Gaussian noise, Salt-Pepper, and Speckle. Result of the experiments shows that the usage of max pooling significantly lowering the performance of the CNN, compared to the system with average pooling layer.

Access control system based on 3D facial recognition technology

Abstract: The invention discloses an access control system based on a 3D facial recognition technology, and mainly relates to the field of information security. The access control system is applied to securityaccess control management of the confidential department, solves the problem that in the prior art, the security degree of an access control system is low, and is characterized in that the access control system includes a microprocessor, an access controller, a power supplying module, an encryption storage module, a 3D facial recognition module and a fingerprint sensor; an artificial intelligence(AI) chip serves as the microprocessor; the encryption storage module is connected with the AI chip through an encryption interface; the fingerprint sensor is electrically connected with the AI chip through an A/D sensor; and the storage module is electrically connected with the AI chip. Through the 3D facial recognition technology, facial data of entering personnel are recognized, and compared with a traditional technology, the higher security is achieved; and through dual cooperation of the 3D facial recognition module and the fingerprint sensor, the problem that after the certain module loses efficacy, the personnel cannot enter the department is further effectively avoided.