Off-line blind person visual assisting method and device

Abstract

The invention provides an off-line blind person vision assisting method and device, and belongs to the field of computer vision. The method comprises the steps: acquiring an image, wherein the acquired image is an image shot by a blind person in daily life; extracting feature points of the images, and splicing the images with incomplete information by using the extracted feature points; and inputting the spliced image into an image description model processed based on a model compression and acceleration algorithm to obtain image description information. According to the invention, time and energy consumption for image processing can be reduced.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to an off-line visual assistance method and device for the blind. Background technique [0002] The visually impaired group is a huge group that is easily overlooked among the disabled. Defects in the eyes prevent them from perceiving external information through the visual system, which brings great inconvenience to daily life and travel. With the development of deep learning, the research and development of image and language processing have made great progress, bringing new ideas to the topic of how to provide visual assistance to the blind. Using computer vision technology, designing a visual aid system that can be carried on low-cost portable devices has become an urgent need to help the blind. [0003] Due to the continuous development of the neural network model, it has brought problems such as huge amount of calculation and difficulty in practical application of related techn...

AI Technical Summary

Problems solved by technology

Based on model compression, further develop a more efficient compressed image description model to solve the problem of using low-rank decomposition compression method with a large amount of calculation but the cumulative error and the use of the original image description model to provide visual aids for blind people and wearable devices. Expensive, limited accessibility, poor interactivity, unable to go offline, difficult to widely use in daily life scenarios, etc.

Images