Joint Optimization Modeling of Downsampling and Quantization Parameters for LiDAR Point-Cloud Compression

Xianfeng Yang; Chen Liao; Chang Duan; Hui Shu; Mengjun Lai; Chao Zhang

doi:10.3788/LOP232530

Laser & Optoelectronics Progress, Volume. 61, Issue 14, 1415006(2024)

Joint Optimization Modeling of Downsampling and Quantization Parameters for LiDAR Point-Cloud Compression

Xianfeng Yang¹, Chen Liao¹, Chang Duan², Hui Shu¹, Mengjun Lai¹, and Chao Zhang^3、*

Author Affiliations

¹College of Computer Science, Southwest Petroleum University, Chengdu 610500, Sichuan, China

²School of Electrical Engineering and Information, Southwest Petroleum University, Chengdu 610500, Sichuan, China

³Intelligent Policing Key Laboratory of Sichuan Province, Sichuan Police College, Chengdu 610206, Sichuan, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(14)

Fig. 1. Flow chart of modeling for the DQPP-RD

Download full size

Fig. 2. Relationship between distortion and parameters in the reconstructed point cloud compared to the original point cloud

Download full size

Fig. 3. Curves of the relationship between distortion and parameters for various rates projected onto different planes. (a) Projection on the $Q - D$ plane; (b) projection on the $Q - W$ plane

Download full size

Fig. 4. Relationship between rate and corresponding minimum distortion. (a) Illustration of the relationship between rate and $D_{m i n}$ ; (b) illustration of the curve fitting

Download full size

Fig. 5. Relationship between estimated minimum distortion and optimal downsampling parameters at different rates. (a) Illustration of the relationship between $D_{m i n}$ and $W_{o p t}$ ; (b) illustration of the curve fitting

Download full size

Fig. 6. Relationship among optimal parameters at different rates. (a) Illustration of the relationship between $l n W_{o p t}$ and $Q_{o p t}$ ; (b) illustration of the curve fitting

Download full size

Fig. 7. Accuracy of the DQPP-RD

Download full size

Table 1. Accuracy and predictive results of the relationship in fitting between $D$ and $W$ , $Q$

View table

Table 1. Accuracy and predictive results of the relationship in fitting between $D$ and $W$ , $Q$

$R_{t}$	$D = a_{0} e x p (b_{0} Q) + c_{0} e x p (d_{0} Q)$		$W = m_{0} Q^{n_{0}} + s_{0}$	$W = p_{0} Q + q_{0}$	$W_{o p t}$	$Q_{o p t}$
$R_{t}$	$D - S C C$	$D_{m i n}$	$W - S C C$	$W - S C C$	$W_{o p t}$	$Q_{o p t}$
80000	0.96	10744.55	0.99	0.97	780	0.36
110000	0.91	7516.92	0.97	1.00	1151	0.24
140000	0.99	5997.95	0.97	0.80	1605	0.21
150000	1.00	5524.39	0.98	0.85	1761	0.20
210000	1.00	3068.62	0.94	1.00	2192	0.19
235000	0.99	2461.59	0.92	0.99	2337	0.16
260000	1.00	1960.53	0.93	0.99	2476	0.14

Table 2. Accuracy of the relationship between $R$ and $D_{m i n}$ fitting with various curve models
View table
Table 2. Accuracy of the relationship between $R$ and $D_{m i n}$ fitting with various curve models
Curve model SCC RMSE
$l n D_{m i n} = a_{1} {(l n R)}^{b_{1}} + c_{1}$ 0.997 170.109
$D_{m i n} = a_{1} {(l n R)}^{2} + b_{1} l n R + c_{1}$ 0.998 136.148
$D_{m i n} = a_{1} {(l n R)}^{3} + b_{1} {(l n R)}^{2} + c_{1} l n R + d_{1}$ 0.998 123.284

Table 3. Accuracy of the relationship between $D_{m i n}$ and $W_{o p t}$ fitting with various curve models
View table
Table 3. Accuracy of the relationship between $D_{m i n}$ and $W_{o p t}$ fitting with various curve models
Curve model SCC RMSE
$l n W_{o p t} = a_{2} (l n D_{m i n})^{3} + b_{2} (l n D_{m i n})^{2} + c_{2} l n D_{m i n} + d_{2}$ 0.990 59.517
$W_{o p t} = a_{2} e x p (b_{2} D_{m i n}) + c_{2} e x p (d_{2} D_{m i n})$ 0.988 64.424
$W_{o p t} = a_{2} D_{m i n} + b_{2}$ 0.982 79.385

Table 4. Accuracy of the relationship between $l n W_{o p t}$ and $Q_{o p t}$ fitting with various curve models
View table
Table 4. Accuracy of the relationship between $l n W_{o p t}$ and $Q_{o p t}$ fitting with various curve models
Curve model SCC RMSE
$Q_{o p t} = m_{1} (l n W_{o p t})^{2} + n_{1} l n W_{o p t} + q_{1}$ 0.954 0.014
$Q_{o p t} = m_{1} (l n W_{o p t})^{3} + n_{1} (l n W_{o p t})^{2} + s_{1} l n W_{o p t} + q_{1}$ 0.992 0.006
$Q_{o p t} = m_{1} (l n W_{o p t})^{4} + n_{1} (l n W_{o p t})^{3} + s_{1} (l n W_{o p t})^{2} + p_{1} l n W_{o p t} + q_{1}$ 0.997 0.004

Table 5. Accuracy of the DQPP-RD
View table
Table 5. Accuracy of the DQPP-RD
SCC between $D_{m i n}$ and $D_{p r e d}$ SCC between $W_{o p t}$ and $W_{r}$ SCC between $Q_{o p t}$ and $Q_{r}$
0.998 0.996 0.991

Table 6. Comparison of BD-rate performance between R-PCC and R-PCC+DQPP-RD on fit datasets and test datasets
View table
Table 6. Comparison of BD-rate performance between R-PCC and R-PCC+DQPP-RD on fit datasets and test datasets
Dataset BD-rate（fit datasets）/% BD-rate（test datasets）/%
Avg -10.43 -16.39
Odometry -21.41 -24.22
Tracking -10.35 -16.02
Object -4.55 -17.52
Road -5.41 -7.80

Table 7. Generalization of the DQPP-RD on different datasets
View table
Table 7. Generalization of the DQPP-RD on different datasets
Dataset BD-rate /%
Avg -8.23
Odometry -21.41
Tracking -5.93
Object -3.64
Road -1.93

Tools

Get Citation

Copy Citation Text

Xianfeng Yang, Chen Liao, Chang Duan, Hui Shu, Mengjun Lai, Chao Zhang. Joint Optimization Modeling of Downsampling and Quantization Parameters for LiDAR Point-Cloud Compression[J]. Laser & Optoelectronics Progress, 2024, 61(14): 1415006

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Machine Vision

Received: Nov. 20, 2023

Accepted: Dec. 21, 2023

Published Online: Jul. 4, 2024

The Author Email: Chao Zhang (galoiszhang@163.com)

DOI:10.3788/LOP232530

CSTR:32186.14.LOP232530

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Accuracy and predictive results of the relationship in fitting between D and W, Q

Table 1. Accuracy and predictive results of the relationship in fitting between D and W, Q

Table 2. Accuracy of the relationship between R and Dmin fitting with various curve models

Table 2. Accuracy of the relationship between R and Dmin fitting with various curve models

Table 3. Accuracy of the relationship between Dmin and Wopt fitting with various curve models

Table 3. Accuracy of the relationship between Dmin and Wopt fitting with various curve models

Table 4. Accuracy of the relationship between lnWopt and Qopt fitting with various curve models

Table 4. Accuracy of the relationship between lnWopt and Qopt fitting with various curve models

Table 5. Accuracy of the DQPP-RD

Table 5. Accuracy of the DQPP-RD

Table 6. Comparison of BD-rate performance between R-PCC and R-PCC+DQPP-RD on fit datasets and test datasets

Table 6. Comparison of BD-rate performance between R-PCC and R-PCC+DQPP-RD on fit datasets and test datasets

Table 7. Generalization of the DQPP-RD on different datasets

Table 7. Generalization of the DQPP-RD on different datasets

Table 1. Accuracy and predictive results of the relationship in fitting between $D$ and $W$ , $Q$

Table 1. Accuracy and predictive results of the relationship in fitting between $D$ and $W$ , $Q$

Table 2. Accuracy of the relationship between $R$ and $D_{m i n}$ fitting with various curve models

Table 2. Accuracy of the relationship between $R$ and $D_{m i n}$ fitting with various curve models

Table 3. Accuracy of the relationship between $D_{m i n}$ and $W_{o p t}$ fitting with various curve models

Table 3. Accuracy of the relationship between $D_{m i n}$ and $W_{o p t}$ fitting with various curve models

Table 4. Accuracy of the relationship between $l n W_{o p t}$ and $Q_{o p t}$ fitting with various curve models

Table 4. Accuracy of the relationship between $l n W_{o p t}$ and $Q_{o p t}$ fitting with various curve models