Image depth estimation method and system

Abstract

The invention discloses an image depth estimation method and system. The method is implemented by comprising the steps of establishing a depth estimation network, wherein the depth estimation network comprises a coding part, a convolutional connection part and a decoding part, and a deconvolutional layer of the decoding part is connected with a last convolutional layer in a convolutional block with the same scale as the coding part to form a final deconvolutional layer; selecting two continuous images in sample images and a depth image of one image as training samples, and training the depth estimation network by utilizing the training samples to obtain a trained depth estimation network; collecting a test image, and extracting a current frame image of the test image and a previous frame image of the current frame image; and inputting color channels of the current frame image and the previous frame image to the trained depth estimation network to obtain a depth image of the current frame image. The method is high in efficiency; and the obtained depth image is high in accuracy and high in compactness.

Description

technical field [0001] The invention belongs to the field of computer vision, and more particularly relates to an image depth estimation method and system. Background technique [0002] Image depth estimation is widely used in smart car obstacle avoidance, robot control, car assisted driving, augmented reality and other application fields. Vision-based image depth estimation in road scenes uses computer vision technology to obtain guidance information by processing images captured by cameras. Compared with other guidance techniques, the vision-based method does not need to add other sensor facilities, and it is easy to expand the acquisition equipment. With the increase in the number of vehicles in our country and the increasingly complex road conditions and higher requirements for assisted driving functions, vision-based image depth estimation has also been widely used in intelligent assisted driving. [0003] At present, the depth estimation methods based on computer vis...

AI Technical Summary

Problems solved by technology

[0007] In view of the above defects or improvement needs of the prior art, the present invention provides an image depth estimation method and system, the purpose of which is to construct a depth estimation network, use training samples to train the depth estimation network, and obtain a trained depth estimation network; collect and test The image is input to the depth estimation network to obtain a depth map, thereby solving the technical problems of low accuracy and low efficiency in the existing technology, and the final depth map is non-dense

Images