Underwater benthic organism detection method and system

Abstract

The invention relates to an underwater benthic organism detection method and system. The method comprises the following steps: acquiring a plurality of underwater benthic organism images and corresponding feature information; establishing a benthic organism detection model according to each underwater benthic organism image and the corresponding feature information, including: extracting a multi-dimensional feature map for each underwater benthic organism image; performing feature fusion on the multi-dimensional feature map based on a feature pyramid network to obtain a fused feature map; generating an anchor box according to the fusion feature map; adjusting the fusion feature map through an anchor frame to obtain an adjusted fusion feature map; training an RPN network based on each fusion feature map and the corresponding anchor box to obtain a trained RPN network; determining a region of interest based on the RPN network and each anchor box; generating a benthic organism detection model according to each region of interest and the corresponding feature information; based on the benthos detection model, to-be-detected characteristic information of to-be-detected benthos can be accurately determined, and the detection precision is improved.

Description

technical field [0001] The invention relates to the technical field of robot vision, in particular to a detection method and system for underwater benthic organisms. Background technique [0002] The general object detection problem has been widely studied in recent years, and researchers have maintained a strong interest in its theory, algorithm and application. However, the research on target detection applicable to underwater has made slow progress. The underwater target detection problem is the basic problem of the visual perception of the underwater robot, and it is also the key prerequisite for the autonomous operation of the underwater robot. The accuracy and timeliness of the detection directly affect the success of the autonomous operation task of the robot. [0003] Due to the actual requirements of autonomous operation, underwater robots need to have high requirements for real-time detection in practical applications. This is actually a balance between problem-so...

AI Technical Summary

Problems solved by technology

Although great progress has been made in underwater target detection methods for decades, many problems still exist. For the detection tasks of benthic organisms (sea cucumbers, scallops, sea urchins), the three most critical problems are how to guarantee Under the premise of real-time algorithm, improve the detection accuracy of deformable organisms (such as sea cucumbers and scallops) as much as possible, how to alleviate the problem of unbalanced training samples caused by uneven growth and distribution of benthic organisms, and how to alleviate the problem of limited underwater visibility

Images