| Summary: | 碩士 === 中國文化大學 === 數位機電科技研究所 === 98 === In this thesis, the fuzzy theorem and pattern clustering algorithm are studied to provide the image recognition of a mobile robot (CHM-R). The proposed architecture comprises a fuzzy inference engine and a simple moment invariant method. The former is utilized to remedy the feature vector error which is generated from the moment invariant method. However, the simple implementation merit is still held. In the proposed image tracking process, three steps are included: 1) Clustering recognition: Utilize the moment invariant method to calculate the moment invariant and covariance values of the captured image every samplinginterval. Then the similarity function can be obtained through the Bayes classifier by comparing with the built-in sampling pattern. Finally, the turning angle of the robot can be calculated. Repeating this step until the desired pattern appears. 2) Position recognition: Utilize the moment invariant method to calculate the moment invariant and covariance values of the captured image every sampling time. Then the similarity function can be obtained through the Bayes classifier by comparing with the built-in sampling pattern. 3) Enable the fuzzy inference engine to calculate the turning angle of the robot by the information of similarity function and the built-in fuzzy rules. Repeating above steps, until the tracking object appears. This thesis has successfully combined the fuzzy theorem and moment invariant method to provide the image recognition of a mobile robot. The effectiveness of the proposed architecture is verified by Matlab simulation and its merit is easy to be implemented. The self-adaptive topics and other practical issues are worthy of further research.
|
|---|
