Abstracting underwater target azimuth tendency feature is a common method to identify underwater target dimension. The cross-spectrum method of splitting beam is often used to abstracting underwater target azimuth tendency feature. But when the SNR is very low, this method is inefficient. In this thesis, an improved algorithm is presented to abstract underwater target azimuth tendency feature which can be used in the low SNR. In the algorithm, the underwater target highlight structure feature is used, and by least square fitting twice the underwater target azimuth tendency feature will be finally obtained. The simulation experiment results show that the proposed algorithm can be used to abstract underwater target azimuth tendency feature even when the SNR is very low. The improved algorithm has made an important foundation for further studying the feature detection and recognition of an underwater target dimension.

Bridge recognition algorithm based on straight-line characteristic is proposed in order to automatically recognize bridge from aerial images, which includes the steps of edge detection, straight-line extraction, coarse location for bridge, accurate location for bridge. Meanwhile, realize the fast accurate location for bridge area by modified 8-neighborhood connectivity processing. The experiment result shows the reliability and efficiency of the method proposed in this article.

With the distance between sonar system and target enlarging, the underwater target transits from near field to far field. Based on planar element scattering similarity, a novel algorithm is presented in this paper. In this thesis, an underwater target model is set up and meshed with triangular planar element by using the software of Ansys. Due to the sonar system is in far field to every planar element, the matrixes whose elements are the planar elements target strength can be calculated. In the course of an underwater target transits from near field to far field, the matrix whose elements are the planar elements target strength is becoming more and more similar to that in far field. Thus the correlation coefficient between the above two matrixes can be used to analyze the transition of an underwater target. Numerical calculation are presented, and the results show that the underwater target scattering characteristics in critical distance approximate the characteristics in far field, but the target strength which is calculated in critical distance is also should be modified little.

For those underwater vehiclees always stored in underwater vehicles-boxes which are filled in nitrogen annoys,the paper provides a formula of calculating storage reliability and relevant cases.Based on this,according to actual launching frequency of trial mission and storage time.It sets up the formulas of each underwater vehicle on launching frequency、maintenance frequency and the frequency of shore-based ohmic exercises.The results of calculation indicate that they are of high practicability and applicability,and can be applied in theoretical analysis on current storage reliability of underwater vehiclees in commission.

Foot contact detection is critical for legged robot running control using state machine, in which the controller uses different control modules in the leg flight phase and landing phase. This paper presents an online learning framework to improve the rapidity of foot contact detection in legged robot running. In this framework, the Gaussian mixture model with three sub-components is adopted to learn the contact data vectors corresponding to running on flat ground, running upstairs, and running downstairs. An online data stream learning algorithm is used to update the model. To deal with the difficulty in obtaining contact data at landing moment online, a “trace back” module is designed to trace back the contact data in the memory stack until the data meet with the probability contact criterion. To test if the foot is in contact with the ground, a projection method is proposed. The acquiring data vector during the leg flight phase is projected onto an independent random vector space, and the contact event is triggered if all projected random variables fall within 1.5σ of the corresponding Gaussian distribution. Experiments on a legged robot show that the presented algorithm can predict the foot contact 16 ms in advance compared with the prediction using only leg force, which will ease the controller design and enhance the stability of legged robot control.