Fu Chengyu

An improved focusing algorithm based on image definition evaluation

In order to overcome the limitations of traditional focusing function, the reasons why traditional approach lower the focusing precision are analyzed, and a theory based on curve fitting and probabilistic model is proposed. This algorithm can effectively improve the limitations that one focusing function is applicable to one type of image only, so that the classical image definition evaluation function are able to determine the focal position. We accomplished some experiments based on the images obtained from the CCD of the lab. Experimental results show that the proposed focusing algorithm based on definition evaluation function has an excellent focusing performance, good noise immunity and practical applications.

Measurement fusion algorithm based on equivalent sensor in sensor networks

In sensor network, several kinds of information with different characteristics are usually sensed by the sensor nodes, and measurement fusion is performed on sink nodes to reduce the communication cost and improve the whole performance. To reduce calculational cost of measurement fusion of sink nodes, the problem is ascribed to the one of measurement fusion for dissimilar sensors. Defect of typical augmented method which causes high calculational cost is analyzed, and then the conception of equivalent sensor and a measurement fusion method based on it is presented. The equivalent sensors with inconsistent elements sequence in original observation vectors are created by inducting invalid elements, so that equivalent sensors with same dimension could be fused and filtered using combined measurement filtering method. In this way, the restriction of combined measurement filtering method was overcome and the calculational cost of measurement fusion of sink nodes is reduced. To be convenient for numerical calculation, suboptimal measurement fusion algorithm is put forward also. Simulation results show that this method has lower calculational cost than augmented method, and the performance doesnpsilat vary with the amount of sensor nodes associated to sink node.

Combined method of uncertain measurement fusion

In multi-sensor system itpsilas difficult to construct observation vector and covariance matrix appropriately when availability of measurement is uncertain. A measurement fusion method for uncertain measurement was presented to resolve the problem. By defining availability function for components of each observation vector, uncertainty of measurement could be expressed, and observation vectors and covariance matrices were generalized to uncertain circumstance. Combining generalized observation vectors based on a minimum-mean-square-error criterion to construct equivalent measurement of Kalman filter, optimal fusion result of uncertain measurement was obtained and current measurement fusion algorithm was generalized. Simulation results show that method presented can deal with the multi-sensor measurement fusion of uncertain availability correctly, and calculational cost is almost as same as one of current algorithm.