Forward vehicle ranging is an important function in ADAS technology such as adaptive cruise control and forward collision warning system, which can help drivers make accurate judgments on impending danger in time to avoid accidents. At the same time, traffic scenarios in bad weather lead to frequent traffic accidents due to situations such as obstructed vision. In order to adapt the front vehicle ranging to multiple scenarios and increase the generalization performance of the ranging algorithm, a deep estimation network based on generative adversarial network is designed for weather scenarios such as rain, snow and fog. First, images of different weather scenes are input in the generator network, using the attention mechanism and the masking operation of the images. Then, feature extraction is performed on the images, and the encoder results in the normal weather depth estimation network are used as the network true values for training to ensure that the generator network can generate image features consistent with normal weather. Finally, the discriminator is used to output the depth map. The results show that the algorithm is highly adaptable to extreme weather such as rain, snow and fog.
Vehicle Depth Estimation Based on Monocular Camera in Extreme Weather
28.10.2022
4528591 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Monocular camera system for depth estimation of objects around vehicle
Europäisches Patentamt | 2024
|Local Scene Depth Estimation Using Rotating Monocular Camera
SAE Technical Papers | 2015
|Local scene depth estimation using rotating monocular camera
Kraftfahrwesen | 2015
|Local Scene Depth Estimation Using Rotating Monocular Camera
British Library Conference Proceedings | 2015
|