Ying-Nong Chen

The Application of a Convolution Neural Network on Face and License Plate Detection

In this paper, two detectors, one for face and the other for license plates, are proposed, both based on a modified convolutional neural network (CNN) verifier. In our proposed verifier, a single feature map and a fully connected MLP were trained by examples to classify the possible candidates. Pyramid-based localization techniques were applied to fuse the candidates and to identify the regions of faces or license plates. In addition, geometrical rules filtered out false alarms in license plate detection. Some experimental results are given to show the effectiveness of the approach

Face Recognition Using Nearest Feature Space Embedding

Face recognition algorithms often have to solve problems such as facial pose, illumination, and expression (PIE). To reduce the impacts, many researchers have been trying to find the best discriminant transformation in eigenspaces, either linear or nonlinear, to obtain better recognition results. Various researchers have also designed novel matching algorithms to reduce the PIE effects. In this study, a nearest feature space embedding (called NFS embedding) algorithm is proposed for face recognition. The distance between a point and the nearest feature line (NFL) or the NFS is embedded in the transformation through the discriminant analysis. Three factors, including class separability, neighborhood structure preservation, and NFS measurement, were considered to find the most effective and discriminating transformation in eigenspaces. The proposed method was evaluated by several benchmark databases and compared with several state-of-the-art algorithms. According to the compared results, the proposed method outperformed the other algorithms.

A novel approach for vector quantization using a neural network, mean shift, and principal component analysis-based seed re-initialization

In this paper, a hybrid approach for vector quantization (VQ) is proposed for obtaining the better codebook. It is modified and improved based on the centroid neural network adaptive resonance theory (CNN-ART) and the enhanced Linde-Buzo-Gray (LBG) approaches to obtain the optimal solution. Three modules, a neural net (NN)-based clustering, a mean shift (MS)-based refinement, and a principal component analysis (PCA)-based seed re-initialization, are repeatedly utilized in this study. Basically, the seed re-initialization module generates a new initial codebook to replace the low-utilized codewords during the iteration. The NN-based clustering module clusters the training vectors using a competitive learning approach. The clustered results are refined using the mean shift operation. Some experiments in image compression applications were conducted to show the effectiveness of the proposed approach.

Biased Discriminant Analysis With Feature Line Embedding for Relevance Feedback-Based Image Retrieval

The focus of content-based image retrieval (CBIR) is to narrow down the gap between low-level image features and high-level semantic concepts. In this paper, a biased discriminant analysis with feature line embedding (FLE-BDA) is proposed for performance enhancement in relevance feedback schemes. Maximizing the margin between relevant and irrelevant samples at local neighborhoods was the aim in this study. In reduced subspace, relevant images and query images can be quite close, while irrelevant samples are far away from relevant samples. The results of four benchmark datasets are given to show the performance of the proposed method.

License Plate Detection and Recognition Using a Dual-Camera Module in a Large Space

In this study, a calibrated dual-camera device, a fixed camera and a pan-tilt-zoom camera, is setup to monitor moving vehicles in an open space. This device not only tracks multiple targets but also gets the license plate images with high quality. Next, a convolutional neural network (CNN) is designed to be a detector and a character classifier for efficiently locating the regions of license plates and recognizing the alphabets on them. Two working environments were setup at the entrance of a university and at a pedestrian-only region in campus. Some experimental results are given to show the validity of the proposed approach.

Facial/License Plate Detection Using a Two-level Cascade Classifier and a Single Convolutional Feature Map

In this paper, an object detector is proposed based on a convolution/subsampling feature map and a two-level cascade classifier. First, a convolution/subsampling operation alleviates illumination, ...

Cross-Platform Mobile Personal Health Assistant APP Development for Health Check

Our team proposes a concept allowing patients taking health check anytime everywhere, which can increase patients’ attention to their own health condition. To improve the user experience and convenience, the system must be designed to simply operate and easily connect with the medical devices. Moreover, the system must have the ability to communicate between the patients and doctor or medical personnel. In this paper, we illustrate our system, such as user interface, storage, display and cloud system. The user interface is designed with standards-based Web technologies. We use PhoneGap to build cross-platform mobile apps with HTML, JavaScript, and CSS. Because patients need to keep their record of health check, we use SQLite database for storage. Moreover, for the patients’ health check report which shall be easily understood, we design line charts to display the data. In this paper, we implement the Personal Health Assistant system.

Connectivity Based Human Body Modeling from Monocular Camera

In this paper, we develop a system for automated human body tracking and modeling based on a monocular camera. In this system, ten body parts including head, torso, arms and legs are extracted to build a 2D human body model. One way to decompose human silhouette into different parts is to generate cuts between the negative minimum curvature (NMC) points. However, due to the self-occlusion problem and left-right ambiguity, each individual body part cannot be successfully identified in every frame. Therefore, in addition to utilizing the NMC points, we design a forward and backward tracking mechanism to identify the location of head in each frame. The torso angle and size are determined by integrating multiple-frame information with the modified solution of Poisson equation. Hands and feet can then be identified correctly based on a modified star skeleton approach along with the nearest-neighbor tracking mechanism. The rest of joint points can also be located by making use of the notion “connectivity”. In the experiments, we analyze the performance of the proposed human body modeling mechanism. We also demonstrate a behavior analysis application by employing the proposed method. The experiment results verify the robustness of the proposed approach and the feasibility of the employing the proposed approach to the action recognition application.

A Novel Approach for VQ Using a Neural Network, Mean Shift, and Principal Component Analysis

In this paper, a hybrid approach for vector quantization (VQ) is proposed for obtaining the better codebook. It is modified and improved based on the centroid neural network adaptive resonance theory (CNN-ART) and the enhanced LBG (Linde-Buzo-Gray) approaches. Three modules, a neuronal net (NN) based clustering, a mean shift (MS) based refinement, and a principal component analysis (PCA) based seed assignment, are repeatedly utilized. Basically, the seed assignment module generates a new initial codebook to replace the low utilized codewords during the iteration. The NN-based clustering module clusters the training vectors using a competitive learning approach. The clustered results are refined using the mean shift operation. Some experiments in image compression applications were conducted to show the effectiveness of the proposed approach

A Dimension Reduction Framework for HSI Classification Using Fuzzy and Kernel NFLE Transformation

In this paper, a general nearest feature line (NFL) embedding (NFLE) transformation called fuzzy-kernel NFLE (FKNFLE) is proposed for hyperspectral image (HSI) classification in which kernelization and fuzzification are simultaneously considered. Though NFLE has successfully demonstrated its discriminative capability, the non-linear manifold structure cannot be structured more efficiently by linear scatters using the linear NFLE method. According to the proposed scheme, samples were projected into a kernel space and assigned larger weights based on that of their neighbors. The within-class and between-class scatters were calculated using the fuzzy weights, and the best transformation was obtained by maximizing the Fisher criterion in the kernel space. In that way, the kernelized manifold learning preserved the local manifold structure in a Hilbert space as well as the locality of the manifold structure in the reduced low-dimensional space. The proposed method was compared with various state-of-the-art methods to evaluate the performance using three benchmark data sets. Based on the experimental results: the proposed FKNFLE outperformed the other, more conventional methods.