Abstract: As a very important passive protection measure, safety belt can effectively reduce the death rate of accidents. Therefore, by detecting whether the driver wears a seat belt, the death caused by not wearing a seat belt can be reduced and the safety awareness of the driver wearing a seat belt can be improved. This paper introduces several common safety belt detection methods, including detection methods based on image classification, detection methods based on target detection and semantic segmentation, and detection methods based on traditional target detection combined with SVM support vector machine. The results show that the traditional detection method needs a lot of preprocessing of images, which reduces the detection speed. The detection method based on deep learning is superior to the traditional detection method in speed and accuracy.

Key words: Seat belt detection; Convolutional neural network; Target detection; Support vector machine

摘要： 安全带作为一种十分重要的被动保护措施，可有效降低事故发生时的驾乘人员死亡率。因 此，通过识别驾驶员是否佩戴安全带，可以减少由没有系安全带而带来的交通事故，并提高司机 系安全带的安全意识。文章介绍了常见的几种识别方法，包括基于图像分类的识别方法，基于目 标检测和语义分割的识别方法和传统目标检测结合 SVM 支持向量机的识别方法。结果表明，传统 的识别方法需要对图片进行大量的预处理，降低了检测速度，基于深度学习的识别方法在速度和 精度方面都优于传统的识别方法。

关键词: 安全带识别；卷积神经网络；目标检测；支持向量机