Binocular camera-based real-time step detection method

Abstract

The invention discloses a binocular camera-based real-time step detection method. The method comprises the following steps of: obtaining depth information of two images acquired by a binocular camerathrough a parallax between the two images, and reconstructing the depth information under a camera coordinate system into a three-dimensional point cloud under a world coordinate system by combining anine-axis gyroscope (gravitational acceleration, gyroscope, magnetometer); segmenting a point cloud formed by a road and other point clouds through a plane fitting method; and carrying out plane fitting on the point cloud formed by the road so as to judge whether the road is flat or formed by steps. The method is capable of detecting upper steps, lower steps and distances from the steps to users,so as to help the persons with visual disturbance (blind persons and persons with poor eyesight) to go up and down steps and then ensure the personal safety; the method adopts passive detection, so that the power consumption is low and other persons are not interfered; and through comprehensively considering possible interference factors in various practical environments, the method has the advantages of being high in speed, high in precision and high in cost performance.

Description

technical field [0001] The invention belongs to the technical fields of computer vision and machine learning, and in particular relates to a real-time step detection method based on a binocular camera. Background technique [0002] According to the World Health Organization, there are about 285 million visually impaired people in the world, of which 39 million are totally blind. Due to the lack of normal vision, the visually impaired cannot effectively distinguish whether there are steps on the ground, which poses a threat to the travel of the visually impaired. [0003] At present, the detection of steps is mostly based on radar, but radar is large in size and consumes a lot of power, making it difficult to make it into a wearable device. Image-based step detection is based on the gradient of the grayscale image or the gradient of the depth. The former is affected by light changes and has poor adaptability to the environment. The latter has high requirements on the accura...

AI Technical Summary

Problems solved by technology

Due to the lack of normal vision, the visually impaired cannot effectively distinguish whether there are steps on the ground, which poses a threat to the travel of the visually impaired

[0003] At present, the detection of steps is mostly based on radar, but radar is large in size and consumes a lot of power, making it difficult to make wearable devices

Image-based step detection is based on the gradient of the grayscale image or the gradient of the depth. The former is affected by light changes and has poor adaptability to the environment.

The latter has high requirements on the accuracy of the depth map, and the performance of consumer-level depth cameras cannot meet its accuracy requirements.

Images