# US 2025/0095334 A1 · Method and system for improving point cloud classification accuracy based on graph spectral domain

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2025/0095334 A1
- Document ID: US20250095334A1
- Status: Published application
- Kind: A1
- Application number: 18/289,159
- Filed: 2022-11-25
- Published: 2025-03-20
- Earliest claimed: 2022-11-22
- Assignees: Peking University
- Applicants: Peking University
- Inventors: Wei HU; Daizong LIU
- Web page: https://patentgrep.com/patent/US20250095334A1

## CPC

- G06V 10/764: using classification, e.g. of video objects
- G06N 3/04: Architecture, e.g. interconnection topology
- G06N 3/0475: Generative networks
- G06N 3/08: Learning methods
- G06T 5/10: using non-spatial domain filtering
- G06T 5/20: using local operators
- G06T 17/00: Three-dimensional [3D] modelling for computer graphics
- G06V 10/82: using neural networks
- Y02P 90/30: Computing systems specially adapted for manufacturing
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- Y02P: CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS

## Abstract

The present invention discloses a method and system for enhancing the accuracy of point cloud classification in the spectral domain, relating to the field of point cloud classification. The method comprises the following steps: acquiring original point cloud data of 3D objects; constructing a KNN graph on the original point cloud to represent the geometric structural information, wherein the KNN graph transforms the original point cloud data from the data domain to the spectral domain using GFT; constructing spectral filters to filter the spectral features of the data transformed to the spectral domain, generating perturbed spectral signals; reverting the perturbed spectral signals back to the data domain through GFT, obtaining adversarial point cloud data; generating samples based on the original point cloud data and adversarial samples based on the adversarial point cloud data, serving as training data, and inputting them into the point cloud classification model for classification training; using the trained point cloud classification model to classify the original point cloud data of the target 3D object and producing a classification result. This invention can enhance the accuracy of point cloud classification and recognition by the model.

## Claims

### Claim 1

A method for improving point cloud classification accuracy based on map domain, comprising:
acquiring original point cloud data of 3D objects;
constructing a KNN graph on the original point cloud to represent the geometric structural information, wherein the said KNN graph is transformed from the data domain to the spectral domain through GFT;
creating spectral filters to filter the spectral features of the data transformed to the spectral domain, resulting in perturbed spectral signals;
using IGFT, converting the perturbed spectral signals back to the data domain, generating adversarial point cloud data;
generating samples based on the original point cloud data and adversarial samples based on the adversarial point cloud data, serving as training data, and inputting them into the point cloud classification model for classification training;
employing the trained point cloud classification model to classify the original point cloud data of the target 3D object, and providing the classification result as output.

### Claim 2 (depends on 1)

The method according to claim 1, wherein the graph-spectral filter is a polynomial function of spectral eigenvalues, and the data transformed into the graph-spectral domain is perturbed in a learnable manner, and a perturbed spectrum is generated by minimizing the adversarial loss Signal.

### Claim 3 (depends on 2)

The method according to claim 2, wherein the expression for minimizing the adversarial loss is as follows:
min
Δ
⁢
L
a
⁢
d
⁢
v
⁢
(
P
′
,
P
,
y
)
,

ϕ
GFT
(
P
′
)
-
ϕ
GFT
(
P
)

p
<
ϵ
,
where
P
′
=
ϕ
IGFT
(
Δ
⁡
(
ϕ
GFT
(
P
)
)
,
Δ
=
[
Δ
w
,
1
·
Σ
k
=
0
K
-
1
⁢
Δ
h
,
k
⁢
λ
1
k
⋱
Δ
w
,
n
·
Σ
k
=
0
K
-
1
⁢
Δ
h
,
k
⁢
λ
n
k
]
wherein Ladv(P′, P, y) is the adversarial loss, P denotes the original point loud, P′ denotes the adversarial point cloud, ϕGFT( ) denotes the graph Fourier transform, ϕIGFT( ) denotes the inverse graph Fourier transform, y denotes the object class of point cloud, Δ denotes the learnable perturbation in the spectral domain, ∈ denotes the threshold restricting the perturbation, {Δw,i}i=1n is utilized to learn the contribution of each frequency component, {Δh,k}k=0K highlights the contributed frequency components, i and k are the index, n and K are the total number.

### Claim 4 (depends on 3)

The method according to claim 3, wherein the adversarial learning loss Ladv(P′, P, y) is composed of a cross-entropy loss function Lclass(P′,y) to facilitate combating point cloud misclassification, a regularization term Lreg(P′, P) and a low-frequency constraint Lconstrain({tilde over (P)}′, {tilde over (P)}), which is formulated as:
L
a
⁢
d
⁢
v
(
P
′
,
P
,
y
)
=
L
c
⁢
l
⁢
a
⁢
s
⁢
s
(
P
′
,
y
)
+
L
r
⁢
e
⁢
g
(
P
′
,
P
)
+
L
c
⁢
o
⁢
n
⁢
s
⁢
t
⁢
r
⁢
a
⁢
i
⁢
n
(
P
′
~
,
P
~
)
wherein, {tilde over (P)}′, {tilde over (P)} are the reconstructed point cloud by only utilizing the low-frequency component of P′, P.

### Claim 5 (depends on 4)

The method according to claim 4, wherein,
L
c
⁢
l
⁢
a
⁢
s
⁢
s
(
P
′
,
y
)
=
{
-
log
e
(
p
y
′
(
P
′
)
)
,
for
⁢
targeted
⁢
attack
log
e
(
p
y
⁢
(
P
′
)
)
,
for
⁢
untargeted
⁢
attack
;
wherein p(⋅) denotes the softmax functioned on the output of the target model, y denotes the object class of point cloud y′ denotes the adversarial class.

### Claim 6 (depends on 4)

The method according to claim 4, wherein, Lreg(P′, P) is a regularization term that minimizes the distance between P′, P to guide the perturbation at appropriate frequencies.

### Claim 7 (depends on 4)

The method according to claim 4, wherein,
L
c
⁢
o
⁢
n
⁢
s
⁢
t
⁢
r
⁢
a
⁢
i
⁢
n
(
P
′
~
,
P
~
)
=

P
~
-
P
′
~

2
;
wherein
P
~
=
U
[
h
⁡
(
λ
1
)
⋱
h
⁢
(
λ
n
)
]
⁢
U
T
⁢
P
P
′
~
=
U
[
h
⁡
(
λ
1
)
⋱
h
⁢
(
λ
n
)
]
⁢
U
T
⁢
P
′
h(λi) is a low-pass graph filter:
h
⁡
(
λ
i
)
=
{
0
,
i
>
b
1
,
i
≤
b
b is the upper bound of the low-frequency band.

### Claim 8 (depends on 1)

A system for improving the accuracy of point cloud classification based on a map domain, comprising a memory and a processor, wherein the memory stores a computer program executing the steps of the method according to claim 1.

### Claim 9 (depends on 2)

A system for improving the accuracy of point cloud classification based on a map domain, comprising a memory and a processor, wherein the memory stores a computer program executing the steps of the method according to claim 2.

### Claim 10 (depends on 3)

A system for improving the accuracy of point cloud classification based on a map domain, comprising a memory and a processor, wherein the memory stores a computer program executing the steps of the method according to claim 3.

### Claim 11 (depends on 4)

A system for improving the accuracy of point cloud classification based on a map domain, comprising a memory and a processor, wherein the memory stores a computer program executing the steps of the method according to claim 4.

### Claim 12 (depends on 5)

A system for improving the accuracy of point cloud classification based on a map domain, comprising a memory and a processor, wherein the memory stores a computer program executing the steps of the method according to claim 5.

### Claim 13 (depends on 6)

A system for improving the accuracy of point cloud classification based on a map domain, comprising a memory and a processor, wherein the memory stores a computer program executing the steps of the method according to claim 6.

### Claim 14 (depends on 7)

A system for improving the accuracy of point cloud classification based on a map domain, comprising a memory and a processor, wherein the memory stores a computer program executing the steps of the method according to claim 7.
