Real-time Riverbank Line Detection for USV System

Abstract

Accurate and quick riverbank line detection plays an import role in extracting the region of interest (ROI) for an unmanned surface vehicle (USV). Because of the riverbank line can be used for USV obstacle detection, visual navigation, motion state estimation. Different from the sea line detection, the riverbank line detection is usually affected by water waves, reflection and inverted image, so the background of land river riverbank line detection are more complex and diversified. In this paper, a robust morphological riverbank line method to a different environment is presented. This method includes following steps: First, image pre-processing (Resize, Gaussian blue, Map RGB colour to HSV space); Then, the value channel was operated by the morphological gradient to highlight the edge; subsequently, segmentation of watershed algorithm; finally extracting the riverbank line in classical Sobel operator. Experiment result proves that our morphological approach can boost efficiency and performance significantly in multiple and complex environments. Index Terms -Unmanned surface vehicle; HSV color space;Morphology; Riverbank line detection.

Improved UNet-Based Shoreline Detection Method in Real Time for Unmanned Surface Vehicle

“…The processing speed of this method is faster, but it is susceptible to the influence of water reflection or texture. Feng et al [8] computed morphological gradients in HSV color space to highlight edges, then used the watershed algorithm to segment the image area, combined with the use of a filtering operator to detect a river shoreline, which achieved real-time shoreline detection but was still subject to the influence of ambient lighting to some extent. Peng et al [25] analyzed the differences in the characteristics of images in HSV color space under different lighting conditions.…”

“…Depending on their technical means, existing visual-based shoreline detection methods can be classified into traditional image-based methods and deep-learning-based methods. Traditional methods include local binary patterns combined with the gray-level co-occurrence matrix method [5], column-by-column logistic regression combined with the polynomial spline modeling method [6], the calculation of vertical gradients in gray space combined with the random sample consensus (RANSAC) algorithm fitting method [7], and the calculation of morphological gradients on HSV color space combined with the watershed algorithm and edge detection method [8]. The abovementioned methods are subject to limitations in their use and are susceptible to water surface reflections, light changes, waves, and long processing times, making them unable to meet the need for accurate and real-time detection of shorelines.…”

Research Progress of Test Methods for Unmanned Surface Vehicle in Complex Sea Conditions

“…In future military and civilian activities, the USV (Unmanned Surface Vehicle) system is expected to do more dangerous and harder work on behalf of humans. In order to truly realize autonomous navigation, a ship must understand and perceive its environment and its own state information [3]. The existing USV is equipped with sensors such as radars, cameras, and infrared imagers to capture environment and target information within a certain range.…”