摘要
由于盲人缺乏视觉感知能力,因此在户外独立出行时具有较大的风险。为了增强盲人户外场景下的环境感知能力,本文针对导盲系统的实际应用,提出一种基于YOLOv5s改进的导盲系统障碍物检测算法。首先,为了降低整体模型的计算量,使用MobileNetV3代替原网络的主干特征提取网络;然后,引入CA注意力机制使模型更好地关注训练过程中的有效特征;最后,采用EIoU边界框损失函数替换原模型的CIoU,优化了预测框的回归速度与精度。在服务器上进行模型验证,实验结果表明本文所提算法相较原模型计算量降低了59%,参数量降低了49.3%,同时mAP提高了2.3%,具有一定的实用价值。
Because blind people lack visual perception,they are at greater risk when traveling outdoors independently.In order to enhance the environment perception ability of blind people in outdoor scenes,this paper proposes an obstacle detection algorithm for guide system based on YOLOv5s for the practical application of guide system.Firstly,in order to reduce the calculation amount of the overall model,MobileNetV3 is used instead of the backbone feature extraction network of the original network.Then,the CA attention mechanism is introduced to make the model pay better attention to the effective features in the training process.Finally,the EIoU bounding box loss function is used to replace the CIoU of the original model,and the regression speed and accuracy of the prediction box are optimized.Compared with the original model,the experimental results show that compared with the original model,the algorithm proposed in this paper reduces the calculation amount by 59%,the number of parameters is reduced by 49.3%,and the mAP is increased by 2.3%,which has certain practical value.
作者
刘昕斐
张荣芬
刘宇红
刘源
程娜娜
杨双
LIU Xinfei;ZHANG Rongfen;LIU Yuhong;LIU Yuan;CHENG Nana;YANG Shuang(College of Big Data and Information Engineering,Guizhou University,Guiyang 550025,China)
出处
《智能计算机与应用》
2023年第11期220-226,共7页
Intelligent Computer and Applications
基金
贵州省基础研究(自然科学)项目黔科合基础-ZK[2021]重点001资助。
