Research On Method Of Underwater Pipeline Detection Based On Gradient Information For AUV

The ocean is rich in biological resources and mineral resources, and it is becoming the new space for human's resource exploitation. It needs the support of science and technology and equipment during marine resources exploitation. An underwater vehicle with the ability of high efficiency, large working depth, long working term under water has increasingly become an important tool to exploit marine resources. Underwater robot technology is a current research hotspot,and underwater target detection and tracking based on vision is a key technique in the study of underwater robot.Based on the project from the Ministry of Industry and Information Technology (MIIT)named "ocean exploration research with autonomous underwater vehicle for engineering", we have studied the underwater pipe detection and tracking system based on optical vision for an underwater autonomous vehicle (AUV). Main research contents in this paper are as follows:1. Analyzing the characteristics of underwater optical imaging as well as features of underwater images. Firstly, considering the demands of image denoising and algorithm rapidity, Gaussian down sampling was researched. As Gaussian down sampling would smooth the edges in the image, we introduced the concept of image scale space and improved the method of Gaussian down sampling. On the aspect of image enhancement,considering some disadvantanges, such as large amount of calculation and loss of low gray information,the traditional fuzzy enhancement algorithm was improved,and a new image enhancement algorithm based on histogram information was constructed.2. In the process of submarine pipeline detection,traditional method is to use the line detection method based on Hough transform for pipeline detection. We studied the traditional method for underwater pipeline detection based on Hough transform and analyzed the disadvantages of this method. A new segment detection algorithm based on image gradient information was applied to the undersea pipeline detection. The region growing algorithm was improved. According to the characteristics of pipelines, we have set some limits to the detected line segments for finally locating the positions of the pipeline in the image. Finally,we have compared the method put forward in this paper with the traditional method based on Hough transform through a lot of contrast experiments. From the aspects of speed and detection accuracy, the advantages of method in this paper were verified.3. Analyzing the geometric model for camera imaging and deducing the conversion formula between the image coordinate system and the robot coordinate system. Zhang Zhengyou's method is adopted to calculate the internal parameters of the camera model.According to the installation position and angle of the camera, we got the external parameters of the camera. Finally, we have deduced the transformation formula from the image coordinates to the robot coordinates for the pipeline.4. Building the hardware and software structure for the system. Simulation experiments were carried out on the hardware-in-the-loop simulation platform, and the validity of the system for AUV underwater pipeline detection and tracking was verified. Finally, multiple sets of underwater pipeline images were tested using the method put forward in this paper, the result shows that the method can detect underwater pipelines effectively, and can meet the real-time requirements of the system.