Underwater target search with robotic fish based on an improved Camshift algorithm

Biomimetic robotic fish has the distinct advantage of efficient propulsion and high maneuverability over conventional underwater vehicles. This paper addresses the underwater target search issue in a free-swimming robotic fish with embedded vision. In particular, we use the stored original image to acquire the color characteristics of the target and propose an improved Camshift algorithm based on the light intensity distribution to search the target in the captured image. Then the robotic fish is driven towards the identified target smoothly with the aid of bio-inspired Central Pattern Generator (CPG) control. All tracking algorithms are implemented in real time with a hybrid control system combining an embedded microprocessor (TI DM3730) and a microcontroller (ATmega128). Latest aquatic experiments demonstrate that a fairly good tracking effect is resulted and the interference caused by the mirror image effect is largely eliminated. The proposed technical scheme offers an alternative to target search in relatively complex underwater environments.