Detection And Identification Of Simulated Wake Bubble Film

The ship wake is motive bubble film. In recent years,several various parameters and distribution of the wake bubble film(WBF) have been extensively attended by researchers, who's is mainly focused research on acoustic properties, optical properties, electromagnetic properties and thermal properties. The physical properties of ship wake are the basic factors to track and identify underwater moving target. In the target identification process, optical sensors have greater advantages than other sensors, which have high resolution, access large quantities of information and high accuracy, real-time, efficient and intuitive method, used to detect and identify WBF, could be reasonably introduced by combination of computational digital image processing techniques with pattern recognition technology.The paper mainly study for four aspects which contain simulation of the WBF digital image preprocessing, the WBF gray-scale image analysis, the BP neural network design and selection, pattern recognition.The paper is focused on classification of simulated WBF in the case of a larger density bubbles under different pressures.The main contents of the paper consist following areas.Firstly, according to comprehensive analysis of the optical properties and wake detection technology we design the experimental system suitable for laboratory simulation of real wake, and build the image acquisition platform to collect bubble film images under different pressures.Secondly, we preprocess the images by means of digital image acquisition. Using the gray histogram statistical moment method to calculate six characteristic parameters such as mean value, standard deviation, the third moment, uniformity, normalized coefficient, entropy. The statistics under the same pressure are very close to the same value, which indicate that the characteristic parameters have small interfering by random factors; the values of each characteristic parameter under different pressures could be distinguished conveniently, these six characteristics reflect the essence of the image characteristics effectively.Finally, based on the analysis of different optimizing neural network algorithm, we choose the method of BP neural network to classify simulated WBF under different pressures. Through testing, classification and recognition can reached an accuracy level of 91.67%.