Search

To make image segmentation accord with user's inclinations, semi-supervised clustering image segmentation is used. It applies manual guides and pays more attention to user's preferences. Watershed is adopted to segment image into series of small regions, which are basic units for segmentation. In the method, adjacent or nearby regions for labeled regions are assumed to belong to the same cluster. Labeled data and unlabeled data are gotten based on manual guides and assigned different weights during iterative processes. A penalty function is introduced when labeled data are incorrectly segmented. For a complex object to be segmented, its different parts are first segmented independently, and the outputs are merged finally. The experimental results show that region-based semi-supervised clustering image segmentation is fast and precise, and its classification results are more in line with user's requirements.

Different from the conventional lasers, the multiple scattering provides positive feedback in the disordered media. So the research on the relationship between the scatters and the modes exist in two-dimensional disordered media is very significant. The special randomness of scatterers is the main factor to yield the localization of light-waves in the disordered media when the scatterers are isotropic. If the scatterers are made of anisotropic material such as uniaxial crystal, ordered scatterers can also form localization of light-waves if the index of refraction of the uniaxial scatterers distribute randomly in two dimensional media. The randomness of the rotation angle of the optic axis determines the index of refraction of each scattering particle. The threshold property of two-dimensional random media with uniaxial scatterers investigated by simultaneously solving Maxwell's equations and rate equations of electronic population for transverse electric (TE) modes. Results show that the lasing threshold of two-dimensional media direct ratio to the degree of the randomness of the distribution of scatterers including both spatial position and index of refraction. The degree of randomness can be described by a parameter.

Two lasing modes are observed in highly scattering media pumped by 532nm laser pulses. The competition between two modes is investigated by simultaneously solving a new numerical model. Results indicated that the dominating lasing mode is determined by both concentration of laser dye and scattering particle density. And the mode has an obvious shift with the increase of density of scattering particles. The simulations of the new laser model agree very well with the experimental data.

A new discrete particle swarm optimization (DPSO) is proposed to solve the multiple knapsack problems (MKP). In the DPSO, the position and velocity of the particles are redefined and the rules and athletic equations of algorithm are also renewed. The multiple knapsack problems are 0-1 programming. The 0-1 programming will be transformed to the integer programming. Then we use the DPSO to solve several multiple knapsack problems. The experimental results show that the DPSO algorithm is feasible to solve the multiple knapsack problems, especially effectively to large scale multiple knapsack problems.

This paper modified the structure of the original PSO algorithm. It proposes that the particles' position have relationship with the one particle's and the whole swarm's perceive extent in the processing of this time and last time, and presents the inertial weight based on simulated annealing temperature. So a new Particle Swarm Optimization algorithm (NPSO) is proposed. It can not improve the one particle's and the whole swarm's perceivable extent and improve the searching efficient but also increase variety of particles and overcome the defect of sinking the local optimal efficiently. At the same time we give the convergence condition of this new algorithm. The algorithm of PSO and NPSO and LPSO are tested with four well-known benchmark functions. The experiments show that the convergence speed of NPSO is significantly superior to PSO and LPSO. The convergence accuracy is increased.

Detecting incidents on urban freeway or arterials using loop detector data is quite challenging. Considering the ability of fuzzy clustering for data discretization, that of rough sets theory to reduction of decision system, and that of fuzzy neural networks to nonlinear mapping, a novel hybrid neuron-fuzzy inference method that synergies fuzzy c-means (FCM),rough sets theory, and adaptive neuro-fuzzy inference system for incident detection was proposed. Firstly, the continuous attributes from detector loop were discretized with FCM clustering. Then, attribute reduction were performed based on rough sets theory using generic algorithm, and the key conditions for incident were determined. Lastly, according to the chosen attribute data, the ANFIS(Adaptive-Network-Based Fuzzy Inference System) was designed for detection. The major advantage of this approach is to optimize the overall structure of ANFIS and avoid the "dimensional disaster" with rough sets theory to attribute reduction. The efficiency of the new method is also illustrated by means of applying to real traffic data, the result of the experiment demonstrated that the solution was very effective to increase the recognition rate and to reduce the number of false detections.

This paper presents a new adaptive particle swarm optimization algorithm with genetic mutation operator. In the algorithm, we give a new adaptive inertia weight to access to local search quickly at the front of the iteration and use the adaptive variance and immune algorithm new affinity definition of the swarm to judge whether the algorithm sink into local minimum or not, then we use a new genetic mutation operator for some particles to escape from the local minimum's basin of attraction and realize global search. The experiments on six problems show that the modified PSO algorithm can improve the global search ability and greatly enhance the successful rate of search.