碩士 / 國立成功大學 / 系統及船舶機電工程學系 / 103 / SUMMARY
This thesis incorporates image processing techniques and fuzzy logic controller for researching on an autonomous underwater vehicle (AUV) that allows static obstacle avoidance in clear and confined water conditions as the AUV moves along a straight line with a constant speed. A webcam is first installed insides a transparent acrylic hemisphere cap located in front of the AUV to capture images of the underwater environment, and an image processing algorithm consisting of grayscale, binary, dilation, erosion, opening, closing and median filter operations is utilized to exclude background and noise of the captured image in order to obtain obstacle information. After obtaining the information of obstacle in the image, the number of pixels in the image obtained through statistical analysis is used as input variables to the fuzzy controller. The fuzzy set for output variable with resulting membership function is the angle of rotation of a servo motor. This research examines two types of controllers, namely Forms I and II. After testing Form I controller in avoidance of obstacle A, it has been improved to enhance reliable obstacle avoidance as called Form II controller since Form I controller does not deliver enough the servo motor angle rotation to perform obstacle avoidance for an obstacle ranging from 5k to 10k pixels. Results of obstacle avoidance with both controllers are compared and obstacle avoidance of obstacle B using the Form II controller is then performed. As for the AUV, it was initially implemented from locally available materials and tested with waterproofing and simple navigation motions, and then was enlarged with upgrading the data processing unit.
Key word: Image processing, fuzzy logic controller, AUV, obstacle avoidance
INTRODUCTION
AUV avoidance usually uses sonar sensors to construct underwater pictures and the obstacle location in the image constructed by the side scan sonar can then be used to plan the best route. However, this thesis proposes to use image processing technology combined with cheap and readily available equipment which is to generate avoidance strategies without making a sophisticated collision system since the image processing technology combined with many algorithms has often been used to recognize obstacles in the international AUV competition. Moreover, an AUV will be able to be given more protection during the AUV’s cruise course if the proposed method can be combined with sonar sensors since they can scan large-scale and construct underwater images. To enhance underwater image quality, this thesis is a preliminary research to use a fuzzy logic controller to overcome underwater visibility. The experimental results are static obstacle avoidance in clear and confined water conditions as the AUV moves along a straight line with a constant speed.
MATERIALS AND METHODS
The experimental step in this thesis is devided into seven steps as described below:
Step1: Use image processing to convert a three-dimension color image into one-dimension grayscale image.
Step2: Use a median filter to filter noises in the background of grayscale image that obtained in step1.
Step3: Use the binary image processing and set a threshold value to separate background and obstacle.
Step4: Perform an erosion to eliminate isolated pixels in the background.
Step5: Follow step4 to use a dilation to repair damaged pixels of an obstacle and restore the initial ratio of the obstacle.
Step6: Follow the above steps and sum the obstacle pixels used as the input value of the fuzzy logic controller.
Step7: Change the obstacle pixels to linguistic variables and input to the fuzzy logic controller along with the membership functions and the fuzzy rules derived in order to obtain an output for controlling the servo motor.
RESULTS AND DISCUSSION
After dodging obstacle A to obtain the distance between AUV and obstacle A, obstacle pixels, servo motor rotation angle with controller Form I, the avoidance probability of dodging obstacle A has been increased from 75% to 90% with improving controller Form I to controller Form II. Then, controller Form II is implemented in AUV to dodge obstacle B for obtaining obstacle’s image information. Since obstacle B is larger than obstacle A (about nine times), servo motor rotation angle with controller Form II approaches the maximum value that a 100% avoidance probability of dodging obstacle B has succeeded in the course of 20 experiments. Figure 1 shows the input-output relation describing an obstacle with controllers of Form I and Form II.
Figure 1 The input-output relation describing an obstacle with controllers of Form I and Form II
CONCLUSION
Since the designed AUV did not have good balance, an external weight was placed outside the AUV. As for avoidance function of the AUV, some errors of calculating obstacle pixels may occur because light is scattered and absorbed by the unstable environment When light is transmitted in water from a subject to an observer. Thus, fuzzy control theory was chosen for the design of obstacle avoidance controllers in this thesis.
| Identifer | oai:union.ndltd.org:TW/103NCKU5345013 |
| Date | January 2015 |
| Creators | Hsuan-FanChen, 陳璿帆 |
| Contributors | Chien-Hsing Lee, 李建興 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 84 |
Page generated in 0.0124 seconds