Research on Power Transform Technology of High Resolution Range Profile of Measured Ship

Abstract

This paper mainly studies the power transform preprocessing of High Resolution Range Profile (HRRP). Power transform can make the HRRP data tend to be normally distributed, improve the recognition effect of common classifiers such as linear discriminant function and k-nearest neighbor (KNN), increase the role of weak scattered points in recognition, and weaken the shielding effect of strong scattered points on weak scattered points, so as to alleviate the attitude sensitivity problem of HRRP. This paper first introduces the statistical properties of HRRP, then introduces the properties of power transform and parameter estimation methods, including parameter estimation based on the skewness and kurtosis normality test and Jarque-Bera normality test. After analyzing and summarizing the shortcomings of the two methods, the parameter estimation method of adaptive Jarque-Bera normality test was proposed. The power transform parameters obtained from the improved parameter estimation method of Jarque-Bera normality test were used to preprocess the measured HRRP. The results were verified by normplot and the measured ship HRRP, which are more close to the normal distribution; In the meantime, the target classification and recognition experiment is carried out by using the measured ship HRRP data after power transform, and the average recognition accuracy is improved by more than 4.8 percent.