Embedded vision-based tracking control for biomimetic robotic fish is a challenging task, but it is still an attractive research direction, which lays a foundation for the development of intelligent underwater vehicles. System design for the biomimetic robotic fish with embedded vision is addressed in this dissertation.Furthermore, stability control of the head of the robotic fish, 3-D positioning based on an artificial landmark and vision guided 3-D tracking control are investigated. Experiments are conducted to verify the effectiveness and feasibility of proposed algorithms. Firstly, the state of the art of visual measurement and control for underwater robots is reviewed, which emphasizes features of underwater vision and existing difficulties. In addition, locomotion performance of the robotic fish is also summarized. The fact that vision-based research of biomimetic robotic fish is still in the primary stage is indicated. This means that there exist both difficulties and potentials in this field. Secondly, the system design of the robotic fish with embedded vision is addressed. Specifically, the mechanical structure with embedded vision and hardware circuits appropriate for visual processing are designed. Specially, the architecture of Codec Engine is presented to handle visual servoing based on large amount of data. At the same time, in order to improve the efficiency of system development, a software simulation platform and a hardware debugging platform are also developed. Further more, 3-D motion gaits, which are verified by experiments on locomotion performance, are designed. Thirdly, the tracking control in the water surface is investigated. An improved CAMSHIFT (Continuous Adaptive Mean Shift) method is proposed to realize automatic recognition and continuous tracking. Specifically, an automatic recognition algorithm based on adaptive color thresholds is presented; An improved real-time positioning method based on the weighed histogram representation of the target model that synthesizes the CAMSHIFT algorithm is putforward. After that, a fuzzy logic controller based on the CPG (Central Pattern Generator) model is employed to generate control signals related to the desired target. The inputs of CPG model are determined by the fuzzy logic, which are from visual positioning. Fourthly, the issue of stability control of the head is explored, which aims at guaranteeing the steady acquisition of image data. Specifically, the traditional model o...
修改评论