Depth extraction method based on monocular vision

Abstract

The invention discloses a depth extraction method based on monocular vision. The method comprises: 1, demarcating the camera of a phone to obtain the inner parameters and image resolution of the camera; 2, establishing a depth extraction model described in the descripton; 3, acquiring the pixel value u, v of the target point through image acquisition of the target to be measured; 4, calculating the depth value of the object to be measured on the image by using the camera internal parameter and the pixel value of the object point obtained in the step 1 and combining with the camera depth extraction model. The depth extraction method based on monocular vision of the invention can be applied to cameras with different parameters such as field angle, focal length, image resolution, etc. to improve ranging accuracy and provide support for object measurement in machine vision and three-dimensional reconstruction of real scene.

Description

technical field [0001] The invention relates to the field of ground close-range photogrammetry, in particular to a method for extracting the depth of a pinhole camera under a monocular vision system. Background technique [0002] Image-based target ranging and positioning, mainly divided into two methods: active ranging and passive ranging [1] . Active ranging is to install a laser ranging device on a machine (such as a camera) for ranging [2-4] . Compared with active ranging, passive ranging does not require the installation of laser ranging devices, so it is favored and widely used. [0003] The most widely used passive ranging is monocular vision ranging [7-9] . The early depth information acquisition methods are mainly binocular stereo vision and camera motion information, which require multiple images to complete the acquisition of image depth information [10-16] , compared with binocular vision distance measurement, monocular distance measurement image acquisitio...

AI Technical Summary

Problems solved by technology

Due to the differences in the internal parameters of different camera devices, for different types of camera devices, this method needs to re-acquire target image information and establish a camera depth extraction model, and different vehicle-mounted cameras have different camera pitch angles due to lens manufacturing and assembly. There are differences, so the method in [21] is less general

In addition, the method of literature [21] uses a vertical target to study the relationship between the imaging angle of the vertical plane image point and the pixel value of the vertical coordinate, and applies this to the measurement of the depth of the object on the horizontal plane, which makes the ranging accuracy relatively low , because the camera’s horizontal and vertical distortion laws are not exactly the same

The invention application with the application number 201710849961.3 discloses an improved camera calibration model and distortion correction model suitable for smart mobile cameras (hereinafter referred to as: improved calibration model with nonlinear distortion items), which can help correct Calibrate the image of the board to obtain higher-precision camera internal and external parameters. The disadvantage is that this method has not been extended to the nonlinear distortion correction of the image to be tested and the measurement of the target object.

Images