The invention provides a road condition detection method based on an ultrasonic array and deep learning. According to the method, firstly, an ultrasonic array facing different directions is arranged to obtain the undulation condition of a road surface; and then acquiring a large amount of ultrasonic array data under various road conditions in an actual operation environment. And data of a period of time window is manually marked, and a corresponding road condition is marked. And training a deep artificial neural network on the marked data. And quantizing the trained artificial neural network and then loading the quantized artificial neural network into a single-chip microcomputer system. The single-chip microcomputer system obtains data in real time through the ultrasonic array and inputs the data into the artificial neural network, and the corresponding road condition is calculated through a deep learning algorithm. And prompting is carried out through a voice or display module.
本发明提出一种基于超声阵列和深度学习的路况探测方法。该方法首先设置一个朝向不同方向的超声阵列用于获取路面的起伏情况。然后在实际运行环境中获取大量各种路面情况下超声波阵列的数据。并对一段时间窗口的数据进行人工标注,标明对应的路况。在这些有标记数据上训练深度人工神经网络。将训练好的人工神经网络量化后装入单片机系统。单片机系统通过上述的超声阵列实时获取数据并输入人工神经网络,通过深度学习算法计算得出对应的路况。通过语音或者显示模块进行提示。
Road condition detection method based on ultrasonic array and deep learning
一种基于超声阵列和深度学习的路况探测方法
14.05.2024
Patent
Elektronische Ressource
Chinesisch
Deep learning-based extended road condition detection method
Europäisches Patentamt | 2024
|