Towards Pedestrian Detection in Radar Point Clouds with Pointnets

Cennamo, Alessandro; Kaestner, Florian; Kummert, Anton

doi:10.1145/3459066.3459067

Cited by 4 publications

(2 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…x p,c of the observed person is available. The coarse position estimate only necessitates an accuracy as can be expected from the initial human detection, a task that has been shown to be feasible with radar data only [25], [26], [27], [28], [29]. Since radarbased human detection is out of the scope of this paper, a stereo video system that records synchronously with the radar sensor network is used instead to obtain the measured humans' positions.…”

Section: B Spatial Filteringmentioning

confidence: 99%

“…While ideas such as voxelization tend to be computationally expensive due to the sparse nature of high-resolution multi-dimensional radar data, point-wise processing can offer an efficient yet powerful way to handle point cloud data. Therefore, the PointNet model proposed by the authors in [21] is adapted, which originally was intended for object classification from 3D scans and which has been successfully applied to radar point clouds for car [31], pedestrian [29], and gesture detection [18], [17]. The key idea behind PointNet is to approximate a function f that operates on a point set by a point-wise function h followed by a symmetric function g,…”

Section: Target List-based Classification Algorithm a Pointnet+lstm A...mentioning

confidence: 99%

See 1 more Smart Citation

PointNet+LSTM for Target List-Based Gesture Recognition With Incoherent Radar Networks

Kern¹,

Grebner²,

Waldschmidt³

2022

IEEE Trans. Aerosp. Electron. Syst.

View full text Add to dashboard Cite

Radar-based gesture recognition can provide autonomous electronic systems with a reliable way to infer a human's intention, e.g. in traffic environments involving vulnerable road users (VRUs). Particularly in complex scenarios, algorithms operating on radar target lists derived from constant false-alarm rate (CFAR) outputs present an attractive solution, as they not only enable the filtering of relevant targets but can also make full use of the diverse, high-resolution target parameters provided by modern radar sensors. Therefore, this paper proposes PointNet+LSTM for the enhanced target list-based recognition of challenging traffic gestures, combining per-frame feature extraction with PointNet and learning from sequences with a long short-term memory (LSTM). The approach is generalized to facilitate the use of multistatic radar data from sensor networks to exploit slightly different viewing angles, which is particularly helpful for motions with low radial velocity. The proposed method is validated on a comprehensive dataset comprising eight traffic gestures and data recorded from 35 participants. Measurements are conducted both indoors and outdoors with an incoherent radar sensor network comprising three chirp sequence (CS)-multiple-input multiple-output (MIMO) sensors. On this challenging dataset, our approach clearly outperforms a reference convolutional neural network (CNN), reaching up to 92.2 % cross-validation accuracy.

show abstract