Search

In this paper, we exploit the multi-modal face recognition capability by a comparative study on 6 fusion methods in the score level, which can be divided into 2 kinds: (1) simple fusion without data training, such as Sum, Product, Max and Min; (2) complex fusion including a predefined data training section, such as Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM). Our experiments are based on the CASIA 3D Face Database and can be divided into two modes: verification and classification. Major conclusions are: (1) 2D modality can achieve similar performance as to 3D modality, and fusion scheme can substantially improve the recognition performance; (2) Product rule gives the best recognition performance in simple fusion methods without training stage; (3) There is no guarantee that the complicated fusion methods will achieve better recognition performance than the simple fusion methods, and it is important to select the most suitable model for fusion according to the tasks.

In this paper, we evaluate the gender classification performance based on 3D faces according to three aspects: image resolution, data fusion and texture descriptor. Our experiments are based on CASIA 3D Face Database, which has 123 individuals in total including different expressions. Main conclusions are as follows: (1) Image resolution has little influence on the gender categorization performance, and there is no guarantee that higher resolution images can obtain better results. (2) Fusion is useful to improve the categorization performance in each single modality. (3) Good local texture descriptors can substantially improve the gender categorization performance, which is even better than that in fusion.

a real-time, interactive video distribution system on Internet for large-scale users faces three bottlenecks: real-time performance, interactive performance and bandwidth pressure. This paper, based on the approach to establish a Real-time Peer-to-Peer network (RTP2P) with the minimum video data delay as the primary goal, and establish a hybrid framework of P2P and centralized networks, argues that the goal of real-time, interactive video distribution system for large-scale users is able to be achieved. The key technologies include: greedy scheduling algorithm, push-pull scheduling model, and the hybrid network model. The test result proves, through an online education system with 2000 simultaneous users, that the end user's video delay is within 10 seconds. Compared with the normal P2P video streaming system with the delay varies between 30 to 120 seconds, Real-time P2P is far more close to C/S networks, the widely adopted network in China, the high bandwidth cost of which restricts its promotion. In short, RTP2P is a more real-time, lower bandwidth-costing network to distribute video stream, saving more than 90% of the bandwidth cost. This program is applicable to the field of large scale users' video real-time performance and interaction, such as online education, chat rooms, multimedia social games, live broadcasting, etc.

An approach of multiple facula targets recognition based on particular distribution is presented. Firstly, a sort of particular distributions is proposed. Secondly, every target's edge and region pixels are searched and recorded after morphology open operation and adaptive thresholding. And then a feature vector is established, in order to segment facula targets from other disturbing regions. Finally, a feature labeling algorithm is proposed to identify each facula target. The approach simplifies the process of matching facula targets observed by two or more cameras in stereo vision measurement by segmenting facula targets and labeling each of them in single image. Experiments on simulated images and real images illustrate validity and robustness of the presented approach.

The application of plastic microfluidic chips can be extended with quartz capillaries connected at the end of their microchannels, e.g. UV absorption detection method can be carried out, which responds to almost 80% chemical compounds in detection. A vision-based experiment system for automatically assembling capillaries to plastic microfluidic chips was set up. UN-curing adhesive is used for the joining procedure. Visual feedback is implemented in the assembly system and the control algorithm is briefly introduced. The methods for obtaining the spatial position deviation between the capillary and the end of the chip's micro channel are described. The deviation information in x-y plane is obtained by performing image processing and converting pixels into actual distance with calibrated data. Two methods for recovering vertical deviation information were explored, and both are feasible for application, one is the depth-in-focus, the other is with the use of microscopic stereovision.

The key techniques for fabrication of polymer microfluidic chips are outlined, including hot-embossing processing technique, fabrication of metal mold-insert with micro structures. The manufacturing process and process control towards the batch production of polymer microfluidic chips are discussed. The developed hot-embossing machines and the prototype automatic fabrication system for plastic microfluidic chips are introduced, some problems for automatic fabrication were solved, and solutions are given