Machine Vision Indoor Positioning Algorithm Based on Improved

Convolutional Neural Network Structure

Wangping Zou

ChiZhou Vocational and Technical College, Chizhou Anhui, 247100, China

Keywords: Convolution Theory, Improvement, Convolution, Manual Positioning Methods, Method.

Abstract: The role of machine vision in indoor positioning is very important, but there is a problem of inaccurate visual

positioning. The manual positioning method cannot solve the machine vision problem in visual positioning,

and the positioning is unreasonable. Therefore, this paper proposes an improved convolutional neural network

method for machine vision indoor localization analysis. Firstly, the convolution theory is used to evaluate the

indoor situation, and the indicators are divided according to the machine vision indoor positioning standards

to reduce the interference factors in the indoor positioning of machine vision. Then, the convolution theory

forms an indoor positioning scheme for machine vision and comprehensively analyzes the positioning results.

MATLAB simulation shows that under certain conditions of the indoor environment, improving the

convolutional neural network method can improve the accuracy of indoor positioning Shorten the positioning

time, and the results are better than manual positioning methods.

1 INTRODUCTION

Machine vision is one of the important contents of

indoor positioning and is of great significance to

indoor positioning research (Safinatul, Dadang, et al.

2022). However, in the process of machine vision

indoor positioning, there is a problem of poor

accuracy (Sharma, and Hota, 2022). Some scholars

believe that the application of the improved

convolutional neural network method to visual

positioning analysis can effectively improve the

indoor positioning effect of machine vision and

provide corresponding theoretical support for indoor

positioning (Arunglabi, and Taliang, 2022).On this

basis, this paper proposes to improve the

convolutional neural network method to optimize the

machine vision indoor positioning scheme and verify

the effectiveness of the model.

2 RELATED CONCEPTS

2.1 Improve the Mathematical

Description of the Convolutional

Neural Network Method

The method of improving the convolutional neural

network is to optimize the machine vision indoor

positioning scheme by using convolution theory

(Arunglabi, and Taliang, 2022), and find the outliers

in visual positioning according to various indicators

in machine vision indoor positioning (Mubeen,

Kulkarni, et al. 2022).At the same time, the machine

vision indoor positioning scheme is integrated to

judge the feasibility of visual positioning finally. The

improved convolutional neural network method gives

full play to the advantages of convolution theory and

uses visual positioning for analysis, which can

improve the accuracy of machine vision indoor

positioning.

Hypothesis 1: The indoor positioning standard is

𝑘(𝑥



), the machine vision indoor positioning scheme

is 𝑎



, the satisfaction of the machine vision indoor

positioning scheme is 𝑥



, and the judgment function

of the machine vision indoor positioning scheme is 𝑦



as shown in equation (1).

146

Zou, W.

Machine Vision Indoor Positioning Algorithm Based on Improved Convolutional Neural Network Structure.

DOI: 10.5220/0013537000004664

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 3rd International Conference on Futuristic Technology (INCOFT 2025) - Volume 1, pages 146-149

ISBN: 978-989-758-763-4

()

ii i

Fd x

 

)

2.2 Selection of Accuracy Schemes

Hypothesis 2: The visual localization function is

()

gx and the position coefficient is

w , then the

positioning scheme is shown in equation (2).

()= (, )

ii ii i

gx z Fd y w

⋅+

∏



(2)

2.3 Comprehensive Judgment of

Machine Vision Indoor Positioning

Scheme

Before the analysis of the improved convolutional

neural network method, the indoor positioning

scheme should be analyzed in multiple dimensions,

and the indoor positioning criteria should be mapped

to the visual positioning library to eliminate the

unqualified Machine vision indoor positioning

scheme. First, the visual positioning is

comprehensively analyzed, and the threshold and

indicator position of the machine vision indoor

positioning scheme are set to ensure the improvement

of the convolutional neural network method

Accuracy. Visual positioning is a system to test the

indoor positioning scheme of machine vision, and

positioning analysis is required. If the visual

positioning is in a nonnormal distribution, the

machine vision indoor positioning scheme will be

affected, reducing the accuracy of the overall

machine vision indoor positioning. In order to

improve the accuracy of the convolutional neural

network method and improve the level of machine

vision indoor positioning, the machine vision indoor

positioning scheme should be selected, and the

specific scheme selection is shown in Figure 1.

Interior space

Posi tion

Displacement

Space

Spatia l dat a

integration

Outpu t positioning results

Positi on in fo rmation

intervention

Figure 1: Results of the selection of an accuracy scheme

The survey of machine vision indoor positioning

scheme shows that the accuracy scheme presents a

multi-dimensional distribution, which is in line with

objective facts. Visual localization is not directional,

indicating that the accuracy scheme has strong

randomness, so it is regarded as a high analytical

study. Visual positioning conforms to the normal

standard, mainly convolution theory adjusts visual

positioning, removes duplicate and irrelevant

schemes, and supplements the default scheme to

make the whole The dynamic correlation of machine

vision indoor positioning scheme is strong.

Comprehensive

3 OPTIMIZATION STRATEGY

OF VISUAL POSITIONING

The improved convolutional neural network method

adopts the stochastic optimization strategy and

adjusts the indoor situation parameters to realize the

scheme optimization of visual positioning. The

convolutional neural network method is improved to

divide visual positioning into different machine

vision indoor positioning levels, and different

schemes are randomly selected. In the positioning

process, the machine vision indoor positioning

scheme with different machine vision indoor

positioning levels is optimized and analyzed. After

the optimization analysis is completed, the indoor

positioning level of machine vision of different

schemes is compared to record the best visual

positioning.

4 PRACTICAL EXAMPLES OF

VISUAL POSITIONING

4.1 Introduction to Machine Vision

Indoor Positioning

In order to facilitate the indoor positioning of

machine vision, the visual positioning in complex

cases is taken as the research object, with 12 paths and

a test time of 12h Table 1 shows the scheme.

Table 1: Machine vision indoor positioning standards

Scope of

lication

Grade Targeting

Effects

Accuracy

Indoors Transverse 83.65 83.07

Lon

itudinal 84.28 84.65

Space Transverse 83.23 84.40

Lon

itudinal 83.87 85.10

Range Transverse 85.60 83.67

Lon

itudinal 83.82 82.79

The machine vision indoor positioning process in

Table 1 is shown in Figure 2.

Machine Vision Indoor Positioning Algorithm Based on Improved Convolutional Neural Network Structure

147

Figure 2: Analysis process of visual positioning

Compared with the manual positioning method,

the improved convolutional neural network method is

closer to the actual standard for the indoor positioning

scheme. In terms of the rationality and fluctuation

amplitude of visual positioning, the manual

positioning method of convolutional neural network

method is improved. Through the change of machine

vision indoor positioning scheme in Figure 4, it can

be seen that the result stability of the improved

convolutional neural network method is better and the

judgment speed is faster. Therefore, the speed and

accuracy scheme of machine vision indoor

positioning scheme of convolutional neural network

method is improved.

4.2 Visual Positioning

The machine vision indoor positioning scheme

includes position, displacement, etc., and the

preliminary visual positioning is obtained after the

preselection of the improved convolutional neural

network method

[61]

. Machine vision indoor

positioning scheme, and analyze the feasibility of

machine vision indoor positioning scheme for visual

positioning. In order to verify the visual positioning

effect more accurately, select the visual positioning of

different machine vision indoor positioning levels,

and the machine vision indoor positioning scheme is

shown in Table 2.

Table 2: Overall picture of the accuracy

Categor

Satisfaction Analysis rate

Indoors 87.49 92.16

ace 87.05 90.38

Location 88.22 89.26

Mean 87.92 90.04

89.06 87.49

P=3.074

4.3 Accuracy and Stability of Machine

Vision Indoor Positioning

In order to verify the accuracy of the improved

convolutional neural network method, the machine

vision indoor positioning scheme is compared with

the manual positioning method, and the machine

vision indoor positioning scheme is shown in Figure

Figure 3: Accuracy of different algorithms

It can be seen from Figure 3 that the accuracy of

the improved convolutional neural network method is

higher than that of the manual positioning method.

However, the error rate is lower, indicating that the

improved convolutional neural network method is

improved the indoor positioning of machine vision is

relatively stable, while the indoor positioning of

machine vision of manual positioning method is

uneven. The average machine vision indoor

positioning scheme of the above three algorithms is

shown in Table 3.

Table 3: Comparison of indoor positioning accuracy of

machine vision with different methods

Algorithm Accuracy Magnitude

Of Change

Error

Improved

convolutional

neural network

methods

86.83 89.63 2.52

Manual

positioning

methods

88.80 90.68 9.27

P 6.30 1.33 0.56

It can be seen from Table 3 that the manual

positioning method has deficiencies in accuracy and

stability in visual positioning, and the visual

positioning changes greatly, and the error rate is high.

The general results of the improved convolutional

neural network method have higher accuracy and are

better than the manual localization method. At the

INCOFT 2025 - International Conference on Futuristic Technology

148

same time, the accuracy of the improved

convolutional neural network method is greater than

90%, and the accuracy does not change significantly.

In order to further verify the superiority of improving

the convolutional neural network method.

5 CONCLUSIONS

Aiming at the problem that the accuracy of visual

localization is not ideal, this paper proposes an

improved convolutional neural network method and

combines convolutional theory to optimize visual

positioning.At the same time, the indoor positioning

threshold of machine vision is analyzed in depth to

construct an indoor situation collection. Research

shows that improving the convolutional neural

network method can improve the accuracy and

stability of visual positioning and can generalize

visual positioning Machine vision indoor positioning.

However, in the process of improving the

convolutional neural network method, too much

attention is paid to the analysis of positioning criteria,

resulting in irrationality in the selection of machine

vision indoor positioning indicators.

ACKNOWLEDGEMENTS

Key project of Natural Science research in

Universities of AnhuiProvince（KJ2021A1417）；

Academic Grant Project for Top Subject (Major)

Talents in Anhui Province （gxbjZD2021119）；

Anhui Provincial Quality Engineering Project

(2021jyxm1027，2021xnfzxm071，2020mooc354).

REFERENCES

Safinatul Hasanah Harahap, Dadang Sunendar, Sumiyadi,

& Vismaia S.Damayanti.(2022).Requirements

Analysis: Drama Education in High

School.Educational Administration: Theory and

Practice, 28(02), 66–73.

Sharma, A.S.., & Hota, D.H..(2022).ECG Analysis-Based

Cardiac Disease Prediction Using Signal Feature

Selection with Extraction Based on AI

Techniques.International Journal of Communication

Networks and Information Security (IJCNIS), 14(3),

73–85.

Rismawaty Arunglabi, A.T.I.R., & Askar Taliang,

M.R..(2022).5G Technology in Smart Healthcare and

Smart City Development Integration with Deep

Learning Architectures.International Journal of

Communication Networks and Information Security

(IJCNIS), 14(3), 99–109.

Mubeen, .S.., Kulkarni, D.N.., Tanpoco, M.R., Kumar,

D.R.., M, .L.N.., & Dhope, T..(2022).Linguistic Based

Emotion Detection from Live Social Media Data

Classification Using Metaheuristic Deep Learning

Techniques.International Journal of Communication

Networks and Information Security (IJCNIS), 14(3),

176–186

Machine Vision Indoor Positioning Algorithm Based on Improved Convolutional Neural Network Structure

149