Scene analysis method based on binocular vision, robot and storage device

Abstract

The invention discloses a scene analysis method based on binocular vision, and the method comprises the steps: obtaining a first original image and a second original image corresponding to the binocular vision, and extracting the multi-scale features of the first original image and the multi-scale features of the second original image; combining the multi-scale feature of the first original imagewith the multi-scale feature of the second original image, performing depth estimation operation, and outputting primary depth estimation information of the first original image based on the combinedmulti-scale feature; performing semantic segmentation calculation by using the multi-scale features of the first original image, and outputting primary semantic segmentation information correspondingto the first original image; and fusing the primary depth estimation information, the primary semantic segmentation information and the multi-scale features of the first original image, and outputtingfinal depth estimation information and semantic segmentation information. The calculation efficiency can be improved, the calculation speed is increased, and the calculation precision is improved.

Description

technical field [0001] The present application relates to the field of visual recognition technology in the robot industry, in particular to a scene analysis method based on binocular vision, a robot and a storage device. Background technique [0002] Because binocular vision can overcome the interference of active light sources in the external environment, it has become a commonly used vision system solution, and it is used more and more in robots and drones. At the same time, since robots and drones need to operate autonomously in complex, changeable, uncertain or uncontrolled environments, they must have efficient recognition, perception, understanding, judgment and action capabilities for the surrounding environment and things. These factors make Robots and UAVs rely more and more on vision systems. Therefore, there is a need for a vision system that enables robots and drones to efficiently recognize and perceive the surrounding environment and things. At present, scen...

AI Technical Summary

Problems solved by technology

At present, scene analysis methods based on binocular vision often complete depth estimation and semantic segmentation through two independent sub-tasks. , which consumes a lot of computing resources and is cumbersome to use

Images