The invention discloses a parking space detection method based on deep learning. The method comprises a data preprocessing step, a convolutional neural network model construction step based on semi-supervised classification, a parking space detection optimization function design step, a step of inputting a training image into a classification model, a step of performing pre-training to obtain an optimization function initial parameter, a step of inputting the training image into the classification model, and a training step. According to the invention, the automatic driving automobile system can accurately detect whether the target parking space is available; further, the vehicle can be accurately parked on the parking space; the sub-problem of parking space detection in automatic drivingautomobile research is effectively solved; the objective optimization function for PU learning in the semi-supervised classification problem is designed, whether the parking space in the image is available or not can be normally classified even under the condition that the marks of the parking space detection data set are not comprehensive, and the parking space detection problem under various conditions can be solved.
本发明公开了一种基于深度学习的车位检测方法,其中方法包括:数据预处理步骤,基于半监督分类的卷积神经网络模型构建步骤,停车位检测优化函数设计步骤,将训练图像输入至分类模型,进行预训练获得优化函数初始参数步骤,将训练图像输入至分类模型,进行训练步骤;本发明能够让自动驾驶汽车系统准确的检测目标车位是否可用,进而精准的在车位上停车,对自动驾驶汽车研究中的停车位检测的子问题进行了有效的解决,本发明设计了一种针对于半监督分类问题中的PU学习的目标优化函数,即使在停车位检测数据集的标记不全面的情况下,也能够正常的分类出图像中的停车位是否可用,可以处理多种情况的车位检测问题。
Parking space detection method based on deep learning
一种基于深度学习的车位检测方法
2020-06-12
Patent
Electronic Resource
Chinese
Outdoor parking lot unoccupied parking space detection method based on deep learning
European Patent Office | 2021
|Deep learning-based empty parking space batch detection method and system
European Patent Office | 2024
|Smart park parking lot parking space prediction method based on deep learning
European Patent Office | 2024
|