基于YOLOv3算法的盲道识别研究

Research on Blind Track Recognition Based on YOLOv3

世界卫生组织统计的数据显示,全球失明人数超过3600万,且有逐年递增的趋势。对于视障人群而言,其出行十分不便,并且现有的盲道识别算法大多是基于颜色和纹理,检测速度较慢,不能很好地解决盲人出行难的问题。为此,本研究提出一种基于YOLOv3网络模型的盲道识别算法。笔者使用LabelImg工具对收集到的数据进行标注,再将标注后的图片送入模型中进行训练,并调整参数,得到最佳的检测模型。试验结果表明,YOLOv3算法的识别准确率达到98%,为优化盲道识别算法提供了新思路。

According to the statistics of the WHO,the number of blind people worldwide is more than 36million,and the trend is increasing year by year.For the visually impaired people,their travel is very inconvenient,and most of the existing blind track recognition algorithms are based on color and texture,and the detection speed is slow,which cannot well solve the problem of difficult travel for blind people.To this end,this paper proposes a blind track recognition algorithm based on YOLOv3 network model.The author use the LabelImg tool to label the collected data,and then feed the labeled images into the model training and adjust the parameters to get the best detection model.The experimental results show that the accuracy of YOLOv3 reaches 98%,which provides a new idea for blind track recognition algorithm.