Xiong Shenshu

Adaptive filtering of color noise using the Kalman filter algorithm

In the view of the drawbacks of the conventional Kalman filtering, a kind of adaptive Kalman filtering approach of color noise based on the correlative method of the system output is proposed. This approach builds the Kalman filtering model of color noise by adopting the equivalent measurement equation in order to avoid complicated computation and the expansion of the dimension of the filter. It is unnecessary to know the variance of the measurement noise beforehand so that it is closer to the actual situation. The correlative method of the system output is presented to estimate the variance of the measurement noise. The adaptive filtering of color noise using the Kalman filter algorithm is applied to solve the cephalometric images of stomatology. Several experimental results demonstrate the feasibility and good performance of the approach.

Vision-based flight parameter extraction for micro air vehicles

The autonomous flight of MAVs (Micro Air Vehicles) is a very important feature of the vehicles. Due to the limitation of the size. weight and energy of MAVs, the automatic pilot technologies with sensors utilized on larger aircrafts are not currently available for MAVs. As the first step towards MAV autonomic flight, a robust, vision-based flight parameter extraction algorithm is proposed in this paper. Four flight parameters-the pitch angle, the roll angle, the flight height and the relative heading of the MAVS-are extracted in this paper. The algorithm has been demonstrated at 5 H z on PentiumIII750 PC. The height error is below 10m. The parameter extraction algorithm is suitable in different weather conditions and altered circumstances. And the experimental results show the effectiveness of the algorithm. Index Termparameter estimation, vision-based , MAVs, image processing

Approximation to Boolean functions by neural networks with applications to thinning algorithms

In this paper, a theorem on approximation to Boolean functions by neural networks and its proof are proposed. A Boolean function, f:{0,1}n/spl rarr/{0,1} is proved to be approximated by a three layer neural network with 2/sup n/ hidden nodes. With the theorem, a thinning algorithm using the neural network technique is concluded. A hard processor implementing the thinning algorithm is designed to raise the thinning efficiency, which can meet the practical needs better. This makes the algorithm suitable for real-time image processing.

Research on adaptive filtering of colored noise and its application

In the view of the drawbacks of conventional Kalman filtering, a kind of adaptive Kalman filtering approach for colored noise based on the correlative method of the system output is proposed to solve the cephalometric images of stomatology. This approach builds a Kalman filtering model of colored noise by adopting the equivalent measurement equation in order to avoid complicated computation and the expansion of the dimension of the filter. It is also unnecessary to know the variance of measurement noise beforehand so that it is closer to the actual situation. Results of several experiments are presented to demonstrate the feasibility and good performance of this approach.

Complex nonlinear exponential autoregressive model for shape recognition using neural networks

A complex nonlinear exponential autoregressive (CNEAR) process which models the boundary coordinate sequence for invariant feature extraction to recognize arbitrary shapes on a plane is presented. All the CNEAR coefficients can be synchronically calculated by using a neural network which is simple in structure and, therefore, easy in implementation. The coefficients are adopted to constitute the feature set which are proven to be invariant to the transformation of a boundary such as translation, rotation, scale and choice of the starting point in tracing the boundary. Afterwards, the feature set is used as the input to a complex multilayer perceptron (C-MLP) network for learning and classification. Experimental results show that complicated shapes can be recognized in high accuracy, even in the low-order model. It is also seen that the classification method has a good degree of fault tolerance when noise is present.

Dynamic parameter estimation based on spectral analysis and its application in mechanic-electric systems

A dynamic parameter estimation method with a two step algorithm has been developed based on spectral analysis. The spectral density function is first obtained from input-output signals of a system, then the dynamic parameters can be estimated from frequency response data. Some parameter estimation algorithms are derived by minimizing a unified loss function with different weight selections. Some excellent application results have been achieved in the measurement, simulation, compensation and control of sensors, actuators and precision machineries.<<ETX>>