Application of Real-Time Image Recognition and Night Vision Techniques to a Small Boat Autopilot Design

• Main Author: 陳俊宏
• Other Authors: Ching-Yaw Tzeng
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/32861135661777522860

碩士 === 國立臺灣海洋大學 === 通訊與導航工程學系 === 101 === This work is concerned with usage of image processing technology to improve the night vision image target recognition performance while employing the visual guidance concept to guide a small boat towards a prescribed straight water channel in the night time. Both the shape feature and color recognition have been considered in the target recognition algorithm. However the low visibility in night time condition easily leads to deformation of the target shape, and only the color recognition method is adopted in the experiments. Specifically the histogram information of HSV(Hue, Saturation, and Value) color space are used to select the appropriate threshold range of HSV to achieve more reliable real-time color image recognition. In the proposed approach, two LED panel targets are arranged in cascade on the quayside to form a leading line. The targets are detected by a CCD camera mounted on the bow and aligned with the center line of the boat, and appropriate camera parameters (Shutter, Exposure and White balance) are set in advance. Use of the histogram information in the processing of the target guarantees more precise thresholds of HSV. The ROI (Region of Interest) scheme is also adopted during the image processing process to enhance identifiability of the leading mark. Calculation of the heading deviation and tracking deviation angles of the boat is carried out by analyzing the displacements of the target centers relative to the CCD center line and these information are fed into the autopilot. This work adopts more sophisticated image processing technique to improve the color recognition performance and outperforms the previous study in which only the red LED can be used in the night time. The experimental results reveal that not only red LED but also blue and green LED can be adopted to guide the small boat into the port in the night. The proposed approach reduces the light disturbance effectively, and the image improvement results in a more identifiable color characteristics that are more suitable for night vision-guided applications.