All-in-one network for nlos mm-wave radar object detection based on transformer

No-line-of-sight perception is a fundamental and challenging problem due to the complex environment and high demands in detection tool. In recent years, the upsurge of autonomous driving has made the mm-wave radar hardware more mature and convinent. Contemporary NLOS detection approaches using signal processing technique mainly focus on eliminating interference from extranous signals and recover the detailed images of NLOS scenes. In this paper, a unified and all-in-one network is proposed which directly deals with mm-wave radar received signal and extrats the effective object's state information in the signal. We construct a mm-wave dataset with Ti's 77GHz mm-wave radar to train and evaluate our network. Experiments on the dataset show that our network has a good performance on the NLOS objects. And it can easily expanded for NLOS moving objects tracking task.