Catálogo de publicaciones - libros

This paper describes a method to know objects in outdoor environment for autonomous robot navigation. The proposition of the method segments and recognizes the object from an image taken by moving robot in outdoor environment. Features are color, straight line, edge, HCM (Hue Co-occurrence Matrix), PCs (Principal Components), vanishing point and geometrical information. We classify the object natural and artificial. We detect tree of natural object and building of artificial object.Then we define their characteristics individually. In the process, we segment regions objects included by preprocessing. Objects can be recognized when we combine predefined multiple features. The correct object recognition of proposed system is over 92% among our test database which consist about 1200 images. We confirm the result of image segmentation using multiple features and object recognition through experiments.

In this paper, selecting line segments as matching features in stereo vision, the orientation difference (O.D.) of the line segments is more deeply evaluated than the previous studies and two new constraints i.e. ordering and collinearity are proposed. The O.D. was supposed to be used for an indoor application, reducing candidates without elimination of the actual matches, especially if the line is horizontal or vertical with respect to the floor. The findings of this paper are as follows: 1) Applying a threshold on O.D. would result in a missing probability for the actual matches and this probability can be calculated for any given threshold, 2) The upper limit of O.D. for horizontal and vertical lines is a function of geometric parameters of the system, 3) An optimal tilt angle, whose results is the minimum upper limit for O.D., can be computed, 4) for disambiguation process, ordering and collinearity constraints are proposed. These are applied in a matching algorithm and their effectiveness is investigated on real stereo images.

In this paper, we proposed image processing technique for automatic real time fire and smoke detection in tunnel environment. To avoid the large scale of damage of fire occurred in the tunnel, it is necessary to have a system to minimize and to discover the incident as fast as possible. However it is impossible to keep the human observation of Closed-Circuit Television (CCTV) in tunnel for 24 hour. So if the fire and smoke detection system through image processing can warn fire state, it will be very convenient, and it can be possible to minimize damage even when people is not in front of monitor. The fire and smoke detection is different from the forest fire detection as there are elements such as car and tunnel lights and others that are different from the forest environment so that an indigenous algorithm has to be developed. The two algorithms proposed in this paper, are able to detect the exact position, at the earlier stage of incident. In addition, by comparing properties of each algorithm throughout experiment, we have proved the validity and efficiency of proposed algorithm.

Natural and efficient HCI technology has been rapidly developing. However, there is always cognitive ’gap’ between HCI and application space. Taking command post of the future as main researching object, the paper explores how to properly combine information modeling with HCI technology and realize the proper abstraction, description and mapping of command space element, according to an order from ’top’ to ’down’: Cognition level, Operation level, Control level and Device level. Starting with conceptual model, command space is modeled on the cognition level. Next, basic methodology are brought forward, which combine command space cognitive concept and map them to operation and control, based on multi-fingers/two-handed touch interaction. This paper focuses on the modeling method of user multi-fingers touch operation model, and the control method from user operation model to multi-fingers touch HCI, to realize conversation between user and device in command space, to improve efficiency of commanding operation and decision-making. The research above can provide theory and method for constructing command and decision-making space of the future and has forerunner effect on multi-modal HCI system.

New proteomic technologies have brought the hope of discovering novel early cancer-specific biomarkers in complex biological samples. Novel mass spectrometry (MS) based technologies in particular, such as surface-enhanced laser desorption/ionisation time of flight (SELDI-TOF-MS), have shown promising results in recent years. To find new potential biomarkers and establish the patterns for detection of cancers, we proposed a novel method to analysis SELDI-TOF-MS using binary cross section imaging and energy curve technology. The proposed method with advantage of visualization is to mining local information adequately so as to discriminate cancer samples from non-cancer ones. Applying the procedure to MS data of proteomic analysis of serum from ovarian cancer patients and serum from cancer-free individuals in the Food and Drug Administration/National Cancer Institute Clinical Proteomics Database, we find that there are outputs of the cancerous when the threshold is above 90 and M/Z is in the range 9362.3296-9747.2723, while outputs of the non-cancerous will appear when the threshold is 60 80 and M/Z is 243.4940-247.8824.

The problem of face feature points detection is an important research topic in many fields such as face image analysis and human-machine interface. In this paper, we propose a robust method of 2D nose detection and tracking system. This system can be valuable for disabled people or for cases where hands are busy with other tasks. The required information is derived from video data captured with an inexpensive web camera. Position of the nose tip is determined with the use of a Gabor wavelet feature based GentleBoost detector. Once the nose tip is initially located, an improved Lucas-Kanade optical flow method is used to track the nose tip feature point. Experiments show that our system is able to process 18 frames per second at a resolution of 320×240 pixels. This method will in future be used in a non-contact interface for disabled users.

The Minimum Bit Error Rate (MBER) detectors outperform the conventional Minimum Mean Squared Error (MMSE) detector by minimizing the Bit Error Rate (BER) directly. In this paper an MBER multiuser detector for Orthogonal Frequency Division Multiplexing-Space Division Multiple Access (OFDM-SDMA) system is proposed employing Particle Swarm Optimization (PSO) for finding the optimum weight vectors. Simulation results show that the proposed system achieves faster convergence with lower complexity as compared to Genetic Algorithms (GA) with same Bit Error Rate (BER) performance.

We have realized an online gesture recognition platform for hand gestures using 2-channel surface EMG signals acquired from the forearm. Several features, such as AMV, AMV ratio and fourth-order AR model coefficients are extracted from the sEMG signal and the gesture segments are recognized with a Weighted Euclidean Distance Classifier. An above 90% recognition rate has been achieved with only a 400 s recognition time. The methods developed in this study are aimed to be applied in a fast-response sEMG control system and be transplanted into an embedded microprocessor system.

This paper suggests 3D co-occurrence texture features by extending the concept of co-occurrence feature to the 3D world. The suggested 3D features are described as a 3D co-occurrence matrix by using a co-occurrence histogram of digital elevations at two contiguous positions. With the addition of 3D co-occurrence features, we encounter the high dimensionality problem in the classification process. Since these ANN (Artificial Neural Networks) clustering algorithms are known as robust in this situation, FCM (Fuzzy C-mean) and GBFCM (Gradient Based Fuzzy C-mean) clustering algorithms are employed to implement the terrain classifier. Experimental results show that the classification accuracy with the addition of 3D co-occurrence features is significantly improved over the conventional classification method only with 2D features.

In this paper, an efficient approach to search for the global threshold of image using Gaussian mixture model is proposed. Firstly, a gray-level histogram of an image is represented as a function of the frequencies of gray-level. Then to fit the Gaussian mixtures to the histogram of image, the Expectation Maximisation (EM) algorithm is developed to estimate the number of Gaussian mixture of such histograms. Finally, the optimal threshold which is the average of these means is chosen. The paper compares the new method with the classical discriminate analysis method of Otsu’s. And the experimental results show that the new algorithm performs better than that of Otsu’s.