1
2

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?

More than 3 years have passed since last update.

【3D OD LiDAR編】BirdNet : BirdEyeViewで2D Object Detection

Last updated at Posted at 2020-10-03

BirdNet: a 3D Object Detection Framework from LiDAR information

![image.png](https://qiita-image-store.s3.ap-northeast-1.amazonaws.com/0/482094/ebea1d6e-94bc-fb01-ddc1-af066056eecf.png)

Point CloudをBEV(Bird Eye Biew)の2Dとして見立てて、2D Object Detectecitonを行うモデル

  1. BEV generation =>3DのPoint Cloudを2Dとして見立てる為の処理
  2. Feature Extraction =>VGG-16を使用して特徴量マップを抽出
  3. 2D Object Detector =>Faster-RCNNを使用して2D(BEV)の位置とyawを推定する
  4. Post Processing =>2D Detectionの結果と地面からの一番高いpointから3DのBounding Boxを作る

新規性

Bird eye’s view generation

  1. PointをVoxelに区切る
  2. Voxelの特徴量を以下の3つの特徴とする

height => cell内の一番高いポイント(地面から3mが以上は無視する)
intensity => intensityの平均値
density => cell内のpointの数 / cell内に取り得る最大のpoint数

結果

![image.png](https://qiita-image-store.s3.ap-northeast-1.amazonaws.com/0/482094/466ffe58-e857-121d-a97f-6bf680b6a6fe.png) 実際に使う情報ってBEVのx,yだから精度も速度もいいかも

image.png
FOV(Field Of View)110°で前方35mまでしか取ってないことに注意!

結論

・x,y,zの情報を失ったのにかなり精度良いな〜。Gridが0.05mで小さいからかな

参考文献

BirdNet: a 3D Object Detection Framework from LiDAR information https://arxiv.org/pdf/1805.01195.pdf
1
2
0

Register as a new user and use Qiita more conveniently

  1. You get articles that match your needs
  2. You can efficiently read back useful information
  3. You can use dark theme
What you can do with signing up
1
2

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?