**1. Introduction**

Seabed topography and morphology provide fundamental information for marine scientific research and ocean engineering [1–4]. Currently, the most widely used sensors in conducting ocean surveying and mapping are side-scan sonar (SSS) and multibeam echo sounder (MBES) [5,6].

As for the SSS, there are two types of operations: towing operation and hull-mounted [5]. The first type is more popular given that the second type of operation may degrade the SSS backscatter image quality because of the platform movements caused in the surveying process [5,7]. When towing, the SSS is usually installed in a towfish and towed by a cable behind a surveying vessel (Figure 1a). It emits a wide-angle beam and receives thousands of seabed echoes at fixed time intervals to form a high-resolution seabed image [5,8]. This kind of operation can minimize the effects of platform movements on the SSS backscatter images. More importantly, as the towfish is near the seabed, it can be used for deep seafloor observation [5]. Thus, the SSS backscatter image geographic coordinates are not accurate because of towing, potential currents dragging and flat bottom assumption [9–12]. By using the Ultra Short Base Line (USBL) or Short Base Line (SBL) in the surveying process, the accuracy

of towfish positions can be improved. However, these two devices are not used in most cases because of high cost. Moreover, three-dimensional (3D) topography is not available by adopting SSS [5,13]. Although the emergence of interferometric SSS (ISSS) can acquire bathymetry, the measurement quality is very sensitive to underwater noise and seriously degrades for echoes close to nadir [14]. In addition to the ISSS, there also exist some other kinds of sonar systems that can acquire bathymetry data, such as Ping DSP 3DSS, Edgetech 6205 and Klein HydroChart 3500 [15], but they can only be used for hydrographic surveys in restricted areas, such as roadsteads or lakes [15]. The output from 3DSS is in the form of an intensity point cloud, which can be post-processed either as bathymetry, backscatter seabed image or a combination of both. However, this kind of data has been found less capable in the detection of seabed targets [16]. As for the Edgetech 6205 or Klein HydroChart 3500, they are mainly used in shallow waters, less than 35 m or 50 m respectively [17,18]. These new sonar systems have restrictions and their performances in the field of bathymetry measurement are usually inferior to those obtained by MBES [6].

**Figure 1.** SSS (**a**) and MBES (**b**) surveying mechanism. (*xtowfish*, *ytowfish*) and (*xvessel*, *yvessel*) are respectively the coordinates of the towfish and vessel tow point, *L* is the cable length, α is the angle between the cable and horizontal direction, *A* is the vessel heading.

MBES is designed for high-accuracy bathymetric work and modern MBES can receive hundreds of echoes from the seabed for one ping and provide a highly detailed backscatter image simultaneously. The MBES is usually installed on a surveying vessel (Figure 1b). Although it can be used for large-scale bathymetric measurement, the interval of footprints in a ping will be enlarged and the bathymetric resolution will be decreased as the water depth and beam incident angle grow. When producing a MBES image, backscatter data are extracted from the time series traces contained within a beam are recorded [19,20]. The central point of a beam is optimally positioned with minimal geometric distortions and the image pixels are distributed around it. With seabed topography, vessel attitudes and the refraction of acoustic waves considered, the positions of MBES image pixels can be accurately computed [19]. By superimposing the MBES image on the seabed topography, the comprehensive seabed topography and surface features can be presented together [21–23]. However, the resolution, signal to noise ratio (SNR) and intensity contrast of MBES images are lower than those of SSS backscatter images [6]. The complementarity of SSS and MBES data provides a way to obtain detailed seabed features by superposing a two-dimensional (2D) SSS backscatter image onto 3D MBES bathymetric terrain.

However, because SSS backscatter image geographic coordinates are inaccurate, it is a challenge to conduct the superimposition of these two categories of data directly. Much research has been carried out in this field. LeBas et al. [24] created synthetic image from the seabed topography and adopted the hierarchical chamfer matching method to match the SSS backscatter and synthetic images. According to the matched result, the SSS backscatter image geographic coordinates were rectified and the obtained SSS backscatter image was superposed on the seabed topography. Yang et al. [25] used the similarity between MBES topography isobaths and SSS backscatter image contours to carry out image matching and the following superimposition operation. The two methods need sufficient

topological variations on the seafloor for creating distinct edges or feature points in sonar images and may fail when dealing with a flat seabed with various sediments. According to similar seabed topography characteristics, textures, targets and sediment distributions being reflected on both SSS and MBES images, Zhao et al. [26] adopted the improved Speeded-Up Robust Features (SURF) algorithm to extract common feature points and conducted image registration, then rectified the SSS backscatter image geographic coordinates, referring to the MBES image based on the matched results. To obtain more correct matches, they conducted seabed classification for SSS and MBES images and used the classification images for image matching followed by a superimposition operation [27]. This method can improve matching performance but its results depend heavily on the seabed classification accuracy. The above methods ignore the imaging mechanism differences between SSS and MBES images, which often lead to inaccurate descriptions of the feature points and result in inaccurate image matching. Considering the limitations of the existing methods, this paper proposes a new SSS and MBES image matching method for acquiring high-resolution and high-accuracy seabed topography and surface details, which can overcome the limitations of adopting a single MBES or SSS for seabed mapping. The remainder of this paper is organized as follows: Section 2 describes the superimposition method in detail; Section 3 designs the experiments to verify the proposed method; Section 4 analyzes the results; Section 5 discusses the proposed method; and Section 6 presents the conclusions according to the experiments and discussions.
