Binocular camera intelligent sensing method and device capable of adaptively adjusting base line

Abstract

The invention discloses an intelligent sensing method and device for a binocular camera capable of adaptively adjusting a baseline, and the method comprises the following steps: detecting a target object from an image, and obtaining the pixel position of the target object; according to the currently set baseline B1 and the internal and external parameters of the camera, calculating the depth and depth uncertainty of the target object through a binocular depth estimation algorithm; judging whether the current depth uncertainty is too large or not: if the current depth uncertainty is larger than the expected depth uncertainty, calculating an expected baseline B2 through the expected depth uncertainty; automatically adjusting the baseline to an expected baseline B2; extracting feature points from the environment through a binocular camera online calibration technology, and performing feature matching in left and right images to obtain new external parameters; according to the new external parameters and the base line B2, recalculating the depth and depth uncertainty of the target object. According to the invention, the base line can be automatically adjusted according to the distance of the measured object, and it is ensured that the depth estimation error meets the expectation.

Description

technical field [0001] The invention relates to the technical field of visual measurement, in particular to a binocular camera intelligent perception method and device for baseline adaptive adjustment. Background technique [0002] At present, in the fields of robotics, unmanned driving, security monitoring, intelligent transportation, etc., in order to enhance the perception of equipment, a variety of sensors are used, such as monocular cameras, binocular cameras, thermal cameras, lidar, etc. Among many sensors, binocular cameras are widely used because they can simultaneously provide images of the environment and dense depth information, and have low cost. [0003] However, most of the existing binocular cameras use a fixed baseline solution, resulting in a large difference in the depth estimation error of far and near objects. Specifically, as the distance of the measured object increases, the error of depth estimation increases rapidly quadratically. Therefore, the bin...

AI Technical Summary

Problems solved by technology

[0004] Although there are currently some binocular cameras with variable baselines, they cannot automatically determine the appropriate baseline length according to the distance of the measured object, and adjust the baseline to a suitable position

Images