Small object detection based on GM-APD lidar data fusion

Table 1. Details of data set
View table
View in Article
Table 1. Details of data set
Number of images Image size Average number ofobjects per image Proportion of small objects Minimum number of pixels occupied by object
1 600 64×64 11 99.1% 6

Table 2. Software and hardware environment
View table
View in Article
Table 2. Software and hardware environment
Configuration item Value
System Ubuntu18.04
GPU GeForce RTX 2080Ti
CPU Intel Core i7-8700 CPU @3.20 GHz

Table 3. Precision of different detection networks on lidar intensity image data set
View table
View in Article
Table 3. Precision of different detection networks on lidar intensity image data set
Method Parms AP_50：95 AP₅₀ AP₇₅ FPS
Faster RCNN 108M 47.5 92.4 46.2 26
YOLO 102M 39.8 84.1 31.9 45
YOLOv3 235M 46.3 94.9 45.7 53
YOLOv5 21.2M 54.1 96.0 55.3 55
SSD 100M 53.0 95.5 53.3 41
Our method 163M 54.7 96.5 56.4 20

Table 4. Precision of different detection networks based on intensity and range information

View table

View in Article

Table 4. Precision of different detection networks based on intensity and range information

Method	Parms	AP_50：95	AP₅₀	AP₇₅	FPS
Faster RCNN	108M	47.6	92.7	47.3	26
YOLO	102M	39.9	84.7	32.5	45
YOLOv3	235M	46.7	95.2	45.9	53
YOLOv5	21.2M	54.5	96.3	55.3	55
SSD	100M	53.1	95.7	53.3	41
Our method （Stage1）	163M	54.8	96.7	56.2	20
Our method （Stage1，2）	181M	56.7	98.8	57.3	17

Table 5. Precision of different methods in ablation experiment

View table

View in Article

Table 5. Precision of different methods in ablation experiment

Method	RFB				CBAM	AP_50∶95	AP₅₀	AP₇₅
Method	Mod_1	Mod_2	Mod_3	Mod_4	Mod_4	AP_50∶95	AP₅₀	AP₇₅
FPN						53.0	95.5	52.2
Im_FPN1	√	√				54.3	96.3	55.4
Im_FPN2	√	√			√	54.7	96.5	56.4
Im_FPN3			√	√		53.8	96.1	54.9
Im_FPN4			√	√	√	54.7	96.2	54.8
Im_FPN5	√	√	√	√		53.8	95.8	54.3
Im_FPN6	√	√	√	√	√	53.9	96.0	53.2

Table 6. Comparison of detection accuracy of 3D point clouds and 4D point clouds
View table
View in Article
Table 6. Comparison of detection accuracy of 3D point clouds and 4D point clouds
Input Parms Epochs AP_50∶95 AP₅₀ AP₇₅
3D point clouds 13.7M ~80 54.3 96.7 54.2
4D point clouds 13.8M ~30 56.7 98.8 57.3

Tools

Get Citation

Copy Citation Text

Dakuan DU, Jianfeng SUN, Yuanxue DING, Peng JIANG, Hailong ZHANG. Small object detection based on GM-APD lidar data fusion[J]. Optics and Precision Engineering, 2023, 31(3): 393

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Information Sciences

Received: Jun. 2, 2022

Accepted: --

Published Online: Mar. 7, 2023

The Author Email: JIANG Peng (jphit2000@126.com)

DOI:10.37188/OPE.20233103.0393

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Details of data set

Table 1. Details of data set

Table 2. Software and hardware environment

Table 2. Software and hardware environment

Table 3. Precision of different detection networks on lidar intensity image data set

Table 3. Precision of different detection networks on lidar intensity image data set

Table 4. Precision of different detection networks based on intensity and range information

Table 4. Precision of different detection networks based on intensity and range information

Table 5. Precision of different methods in ablation experiment

Table 5. Precision of different methods in ablation experiment

Table 6. Comparison of detection accuracy of 3D point clouds and 4D point clouds

Table 6. Comparison of detection accuracy of 3D point clouds and 4D point clouds

微信扫一扫：分享