Open AccessImage recognition and blind-guiding algorithm based on improved YOLOv3
Author(s) -
Haoyu Lu,
Yunlong Ma
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1865/4/042107
Subject(s) - computer science , upsampling , artificial intelligence , computer vision , software portability , ranging , obstacle , identification (biology) , deep learning , channel (broadcasting) , image (mathematics) , object (grammar) , cognitive neuroscience of visual object recognition , telecommunications , computer network , botany , political science , law , biology , programming language
YOLOv3, a target detection algorithm based on deep learning, is widely applied in object recognition, especially in guiding the blind. The existing products of assisting the blind based on YOLOv3 can already achieve high-precision, high-real-time object recognition. But YOLOv3 also has many limitations, such as the inability to measure distances or it’s hard to recognize objects correctly in fog or haze. For these deficiencies, this paper proposes a road barrier monitoring method based on improved YOLOv3, using an image downsampling algorithm based on the dark channel to defog the image, and then with the binocular distance measurement algorithm to calculate the obstacles from the distance of the camera according to the width and height of the obstacles. The experimental results show that the improved product retains the advantages of high accuracy and fast recognition speed of YOLOv3. At the same time, it also owns the new functions of obstacle ranging and bad weather identification. The improved algorithm can meet the requirements of portability, real-time, and practicality of guide products.