Establishment and analysis of two angular response correction methods for multi-beam backscattering intensity data

The multi-beam sonar backscatter strengths (BS) on seabed sediment are widely applied in many fields, and the high-reliability BS can improve the truth of inversion of seafloor geomorphology. Due to the influence of physical mechanism of beam scattering mode, the BS varies with the beam incidence angle, which is called Angular Response (AR). AR is an important interference factor to obtain actual BS, therefore, weakening the influence of AR is an inevitable mean to improve the reliability of BS. Nevertheless, the low quality of multi-beam sonar image is often due to the poor adaptation of existing AR correction model in complex submarine environment. Two kinds of BS correction methods are proposed to weaken the AR influence, namely BP-wavelet neural network (BPWNN) correction and AR correction. The former is based on the BS variation regulation and the latter is based on the acoustic scattering mechanism. First of all, the regional AR curve was obtained by means of averaging the BS on each angle unit among the region after equivalent angle resampling for each ping. Secondly, the BPWNN fitting and Gaussian fitting were used to smooth the AR curve respectively. Thirdly, the AR correction model parameters (such as boundary angle, curve slope) were calculated from second derivative curves for AR correction method, while nothing was done for BPWNN correction method. Fourthly, the BPWNN correction model and AR correction model were established respectively using model parameters, herein, what they represent for corrected BS and mean BS are different among the two models. Finally, all the raw BS were corrected using BPWNN correction model and AR correction model for each ping. The experimental results demonstrate that the accuracy of the two methods is about 30% higher than that with the default correction method (Kongsberg EM correction model), and the AR correction method is more accurate than the BPWNN fitting correction method, but the computational efficiency is decreased. The above results verify the effectiveness of the two methods, and the correction of BS can significantly improve the quality of sonar images.