Dual-LiDAR Detection Method for Nearshore Situational Awareness

Feng Sun; Yufeng Xiao; Haiyang Wang; Hongsen He; Ran Liu; Liqiong Yang

doi:10.3788/LOP242094

Laser & Optoelectronics Progress, Volume. 62, Issue 12, 1228003(2025)

Dual-LiDAR Detection Method for Nearshore Situational Awareness

Feng Sun¹, Yufeng Xiao^1、*, Haiyang Wang¹, Hongsen He¹, Ran Liu¹, and Liqiong Yang²

¹School of Information and Control Engineering, Southwest University of Science and Technology, Mianyang 621010, Sichuan , China

²School of Civil Engineering and Architecture, Southwest University of Science and Technology, Mianyang 621010, Sichuan , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(16)

Fig. 1. System framework for point cloud target detection of multiple solid-state LiDARs in unmanned surface vehicles

Download full size

Fig. 2. Calibration model for dual-LiDAR

Download full size

Fig. 3. Division results for voxel regions of point clouds obtained by LiDAR scanning at different postures. (a) Calm water surface; (b) dynamic water surface

Download full size

Fig. 4. Scanning model of LiDAR on the water surface

Download full size

Fig. 5. 3D feature extraction module of network. (a) Original 3D convolution framework and 3D convolution framework with selective convolution kernel; (b) Sparse SK Conv

Download full size

Fig. 6. Detection head framework with layer attention

Download full size

Fig. 7. Data collection platform

Download full size

Fig. 8. Partial typical scenes

Download full size

Fig. 9. Distribution of annotated objects at different distances in the data set

Download full size

Fig. 10. Loss curves during training process

Download full size

Fig. 11. Test results visualization in a real-world environment. (a) Scene overview; (b) visualization view 1; (c) visualization view 2

Download full size

Fig. 12. Comparison of detection results with and without IMU-corrected during hull tilt. (a) Camera view; (b) original point cloud view; (c) IMU-corrected point cloud view

Download full size

Table 1. AP of baseline and proposed algorithms on the developed dataset
View table
Table 1. AP of baseline and proposed algorithms on the developed dataset
Algorithm BEV AP /% 3D AP /% FPS /（frame·s^-1）
boat buoy boat buoy
Improvement 2.30 9.99 4.00 5.32 9.41
Baseline 59.45 57.36 50.76 36.63 20.20
Proposed 61.75 67.35 54.76 41.95 29.61

Table 2. Impact of IMU on detection accuracy
View table
Table 2. Impact of IMU on detection accuracy
Algorithm IMU BEV AP 3D AP
Baseline — 59.45 50.76
√ 64.20 52.19
Proposed — 61.75 54.76
√ 64.85 58.84

Table 3. Impact of different modules on network

View table

Table 3. Impact of different modules on network

Baseline	DV	RSK	CL	BEV AP /%		3D AP /%		mAP /%	FPS /（frame·s^-1）
Baseline	DV	RSK	CL	boat	buoy	boat	buoy	mAP /%	FPS /（frame·s^-1）
√				59.45	57.36	50.76	36.63	51.05	20.20
√	√			59.94	64.53	53.03	36.09	53.40	32.70
√	√	√		59.63	66.48	52.44	36.73	53.82	30.82
√	√	√	√	61.75	67.35	54.76	41.95	56.45	29.61

Table 4. Performance comparison between proposed and other algorithms

View table

Table 4. Performance comparison between proposed and other algorithms

Algorithm	BEV AP /%			3D AP /%			FPS /（frame·s^-1）
Algorithm	Overall	boat	buoy	Overall	boat	buoy	FPS /（frame·s^-1）
SECOND^［19］	49.29	62.40	36.18	40.87	59.28	22.45	25.90
PointPillars^［9］	46.86	59.70	34.02	38.21	55.90	20.52	22.10
CenterPoint^［17］	58.41	59.45	57.36	43.70	50.76	36.63	20.20
Proposed	64.55	61.75	67.35	48.36	54.76	41.95	29.61

Tools

Get Citation

Copy Citation Text

Feng Sun, Yufeng Xiao, Haiyang Wang, Hongsen He, Ran Liu, Liqiong Yang. Dual-LiDAR Detection Method for Nearshore Situational Awareness[J]. Laser & Optoelectronics Progress, 2025, 62(12): 1228003

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites