基于成像声呐的鱼类三维空间分布

针对海洋牧场中鱼群的三维空间分布问题,本研究提出一种利用成像声呐进行位置计算的方法。将成像声呐固定在船舷外侧的水下,并保证波束发射方向和声呐移动方向一致,通过走航的方式采集水下鱼群信息。首先对采集的原始数据进行图像处理,包括图像构建、背景去除、目标提取等,然后利用基于交互式多模型联合概率数据关联算法对水下目标进行关联处理,得到同一个目标在声呐水平视场中不同帧图像中的对应关系,在此基础上根据连续两帧图像中目标位置关系计算目标的空间坐标,最后结合关联算法获得多目标在三维空间中的运动轨迹以及深度分布情况。研究表明,本方法可以有效获取鱼群在水下的三维运动轨迹及其分布情况,这将为鱼类行为分析以及海洋牧场的资源评估提供技术支持。

Three-dimensional distribution of fish using an imaging sonar

In order to obtain the three-dimensional (3D) distribution of fish in a marine ranching, a method using imaging sonar to calculate the 3D coordinates of fish is proposed in this paper. The imaging sonar used in this research is a Dual-frequency Identification sonar (DIDSON). It is a multi-beam sonar that uses acoustic lens to form individual beams. It constructs a high-definition two-dimensional image for target detection by transmitting ultrasonic beams underwater and receiving the echo signals. It sets a new standard for excellence in underwater vision in black and turbid waters due to obtaining near-video quality dynamic images for the identification of objects underwater. In addition, split-beam echo-sounders have also been used on some occasions to monitor the movements of fishes. Despite the improvements achieved in fish monitoring techniques, the interpretation and classification of the data collected by the traditional acoustic techniques are often challenging and require extensive experience and effort. The DIDSON bridges the gap between existing fisheries-assessment sonar and optical systems. In a DIDSON, 96 transducer elements constitute a linear array and each element both transmits and receives acoustic beams such that echo amplitude is determined by the intensity of the reflected signal. It can obtain the distance and azimuth of the target from sonar images, but it is unable to acquire the elevation of the target from the images. To overcome this difficulty and obtain the fish distribution, a new method is proposed. Firstly, the sonar is fixed on the outside of the ship''s rail and submerged in the water to collect fish''s information through the investigation on navigation. At the same time, the beam emission direction is on the same plane with the sonar''s moving direction. After data collection, image processing is conducted, including image construction, background elimination and target extraction from horizontal field-of-view. Target association based on Interacting Multiple Model Joint Probabilistic Data Association Filtering (IMMJPDAF) is carried out to deal with the extracted targets, thus the relations of one target in different frame images can be obtained. The 3D target coordinates are acquired according to the spatial geometric relationship between the positions in two consecutive frame images. Finally, multiple target trajectories in 3D space and the depth distribution of targets are obtained through the correlation algorithm. The experiment was carried out in Dishui Lake which is located in Shanghai. Experimental results showed that the proposed method can effectively acquire the fish movement tracks in 3D space underwater and the distributions in depth. It also showed that most fish swam in the depth 3–5 meters. It will help to analyze the fish behavior and provide technical support for fishery resource assessment in a marine ranching.