For the machinery capture of sea organism targets in the offshore aquaculture, a sea organism absorptive underwater vehicle is designed. The underwater vehicle can absorb and capture sea organism targets in manual tele-operation mode and visual servo mode. In order to realize vision based target capture control, the kinematic relationship between the underwater vehicle and the target is established in the camera plane coordinate, and the adaptive recurrent neural network controller is proposed on this basis. The recurrent neural network is designed to estimate and compensate environmental disturbance. The S surface function is utilized to make the underwater vehicle reach the expected position quickly and remain stable. Moreover, a robust function on the basis of recurrent neural network and system dynamic model is designed to improve the reliability and stability of the nonlinear system in visual control. Finally, a visual tracking control experiment for sea organisms is conducted in the offshore aquaculture, and active absorption control of sea organism targets is achieved, which verifies the controller function.