In-situ Holothurian Non-contact Measurement based on Parallel Laser Beams and Semantic Segmentation

Holothurian measurement is a basic requirement in aquaculture. In-situ non-contact measurements are basically harmless, which make such techniques as an idea choice for aquaculture monitoring. To develop an in-situ non-contact holothurian measurement system, there are two main challenges. Without contact, the lack of reference would become a serious challenge. Another challenge comes from the detection and localization of the holothurian in an underwater environment. In this paper, we propose an effective method for estimating the length of holothurian in underwater environments using Mask R-CNN model and parallel laser beams. We use two parallel laser beams to provide the distance reference avoiding touching the targets. We include Mask R-CNN model for targets detection and segmentation. The length of the target can thereby be calculated by comparing with the distance between laser beams on the image captured by the camera and the length of the holothurian on the image. Our method can estimate the length of sea cucumbers at a fairly low expense with necessary accuracy.