Search

The pressure-holding sampling technology for deep-sea sediments has been identified as a prerequisite for studying the sedimentary environment with respect to the mechanical and biological characteristics of sediments. Thus, in this study, we designed a set of abyssal sediment sampling systems based on image recognition technology. First, a system camera was used to obtain the movement trajectory of the sampling cone in the laboratory simulation scene. After collecting the data of the image including the sampler, we preprocessed the data using an underwater image restoration method based on the dark channel prior. An image recognition unit was used to verify that the position of the sampling cone meets the requirement. The main control unit controlled the actuator of the sampling device to accurately control the position of the sampling cone. Furthermore, we used the Mask R-convolutional neural network architecture to build the network framework for controlling the expansion and braking of the sampler. The experimental results showed that the system achieved a very high detection accuracy for the position of the sampling cone. [ABSTRACT FROM AUTHOR]