Depth image high-resolution reconstruction method based on Kinect camera

Abstract

The invention discloses a depth image high-resolution reconstruction method based on a Kinect camera and relates to the image processing field. The method comprises the steps that a color camera and adepth camera of the Kinect camera are calibrated, and camera parameters of the Kinect camera are acquired; a color image and a depth image of a target object are acquired through the color camera andthe depth camera of the Kinect camera respectively; the depth image is mapped into a color image pixel coordinate system where the color image is located according to the camera parameters of the kinect camera, and an after-alignment depth image is obtained; a non-convex optimization model is constructed according to the color image and the after-alignment depth image; and an alternate directionmultiplier algorithm is utilized to solve the non-convex optimization model to obtain a reconstructed depth image. Through the method, high-resolution reconstruction of the depth image under the guidance of the color image in a natural scene can be realized, and good edges can be obtained and kept smooth and robust while quick convergence is guaranteed.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a high-resolution reconstruction method of a depth image based on a Kinect camera. Background technique [0002] The fruit picking robot can automatically detect and pick the fruit. It is widely used because of its high efficiency and good automation. The picking action of the fruit picking robot depends on the accurate detection and positioning of the fruit by its visual inspection system. For the picking accuracy of robots, the most important thing is to improve the detection and positioning accuracy of fruits in complex scenes. [0003] In complex scenes, fruit detection and recognition based on color images often have the problems of low recognition accuracy and poor robustness. In order to solve this problem, a low-cost Kinect camera can be introduced into the field of fruit detection and recognition. The Kinect camera includes The depth camera collects the depth image of th...

AI Technical Summary

Problems solved by technology

[0003] In complex scenes, fruit detection and recognition based on color images often have the problems of low recognition accuracy and poor robustness. In order to solve this problem, a low-cost Kinect camera can be introduced into the field of fruit detection and recognition. The Kinect camera includes The depth camera collects the depth image of the fruit tree, and detects and locates the fruit based on the depth image

However, due to factors such as illumination changes, occlusion, and the hardware of the Kinect camera itself, the noise of the depth image obtained by the Kinect camera in outdoor natural scenes is more serious than that of the depth image obtained indoors, and the low resolution of the depth image also affects the geometry. The extraction of features brings difficulties, and there are still problems in the depth image that the edge of the depth image does not correspond to the edge of the color image, and depth noise, which leads to a decrease in the accuracy of the information of the fruit tree and affects the identification of subsequent fruits.

Images