Research On XBT Moving Parameters Measurement Methods Based On Computer Vision

The expendable bathythermograph(XBT) is a free fall instrument for measuring temperature profile of the ocean. Since XBT probe carries no pressure sensor, its depth is not measured directly but inferred from elapsed time and an experiential equation(Fall Rate Equation, FRE). FRE has an important influence to accuracy of XBT’s data. By measuring the pose parameters of XBT probes, we can conduct experiments based on video to explore its fall in the water and improve its FRE. Based on computer vision and image processing, we studies XBT probe’s pose measurement methods deeply. The main work and innovations are summarized as follows:1. Different methods were proposed to detect XBT probes in image sequences automatically. Firstly, this paper proposed a target detection algorithm based on background difference. Using kalman filter to predict target’s detection area in next frame, we used an overall fuzzy evaluating method to recognize XBT probes by calculating every region’s characteristic function. Secondly, a new target detection and recognition algorithm were proposed to realize automatic detection of targets in low-quality images. The template center was determined in the initial frame by artificial guidance method, then, we can determine the template center in the rest frames by tracking and predicting, correlation matching, surface fitting and other algorithms. Finally, simulation and physical experiments were designed to confirm our methods using XBT probes.2. This paper proposed a method based on monocular and stereo vision measurement to measure XBT probe’s moving parameters. Firstly, the falling velocity and acceleration of XBT probe were calculated by using outlier removal algorithm, the polynomial optimal filter and other data post-processing methods. Secondly, a new method was proposed to measure XBT’s velocity in low-quality images based on integral projection and Radon transform. Then, we proposed an iterative optimization method to extract target’s center axis based on Radon transform to measure XBT’s attitude. What’s more, the 3D coordinates of the XBT probe’s edge were calculated by binocular stereo matching based on epipolar line constraint. A new method was proposed based on RANSAC and least square to fit probe’s 3D edge. Lastly, XBT moving parameters measurement systems were constructed to measure XBT‘s moving parameters based on measuring method we proposed. Simulation and physical experiments were designed to confirm our method。XBT moving parameters measurement methods based on computer vision were proposed and experiments were conducted to study the first meters of XBT fall and improve its FRE. Firstly, videos of XBT falling in air and water were recorded by monocular and stereo measurement systems. Secondly, XBT probes were detected and its moving parameters were measured based on image processing and vision measurement methods. Lastly, the methods we proposed were used to measure and analyze XBT probe’s moving parameters.