Journal of Electronic Research and Application

Infrared Image Feature Enhancement Under Complex Backgrounds Using an Improved UNetFPN

Qiuyu Wang — 2026-05-08

Infrared images acquired under complex background conditions are often affected by background clutter, local high-response interference, and non-uniform fluctuations, which may reduce target saliency and local discriminability. To address this issue, this paper proposes an improved UNetFPN-based feature enhancement network, termed CBAM-UNetFPN. Built on an encoder-decoder architecture, the proposed method introduces a feature pyramid fusion mechanism to combine shallow spatial details with deep semantic information, and incorporates an attention enhancement strategy to enhance target-related responses while suppressing redundant background activations. Experiments were conducted on three public infrared image datasets, namely NUDT-SIRST, IRSTD-1k, and WideIRSTD-Weak, and the enhancement performance was evaluated using the signal-to-clutter ratio, background suppression factor, and contrast gain. The results show that the proposed method achieves stable enhancement performance across scenes with different levels of complexity. Comparative experiments further indicate that CBAM-UNetFPN can better balance target response enhancement and background suppression under complex background conditions, thereby improving the local discriminability between target regions and the surrounding background.

Research Progress on Resource Recovery in Industrial Wastewater Treatment

Shanshan Cao — 2026-05-21

The implementation of the strategy of circular economy and green development has changed people’s understanding of industrial wastewater, which is no longer regarded as a burden, but has unlimited potential in resource utilization, and can bring good benefits to enterprises. The research focuses on the development of industrial wastewater resource recovery from three aspects: technological progress, policy support and economic pressure, and introduces the research results of scholars in recent years from the aspects of chemical energy conversion, water resource circulation and reuse, and the selective recovery of high value substances. It is suggested that in order to improve the efficiency of resource recovery, we should not only promote the development of industrial wastewater treatment technology, but also strengthen the governance of multiple stakeholders, and provide theoretical support for the sustainable development of industrial wastewater resource recovery.

Unconstrained Latent Factorization-Based Improved Relief-F

Wenting Chen — 2026-05-21

Feature selection is essential for dimensionality reduction on big data, but it faces considerable challenges when applied to high-dimensional and sparse datasets. To address these challenges, this paper proposes Unconstrained Latent Factorization-based Improved Relief-F (ULF-IR), a novel feature selection method tailored for such complex scenarios. The method integrates two main components: (1) a double factorization (DF)-based unconstrained latent factor model is employed to accurately reconstruct missing data without relying on pre-imputation or strict non-negativity constraints; (2) an improved Relief-F (IRelief-F) algorithm assigns reliable importance weights to features, effectively differentiating among highly similar features even in the presence of noise introduced during imputation. Comprehensive experiments on three real-world datasets show that ULF-IR consistently surpasses state-of-the-art methods in both classification accuracy and robustness, demonstrating its effectiveness as a dependable solution for feature selection on high-dimensional, incomplete data.

Research on Signal Processing Architecture Design and Compatibility of Multimodal Perception System for Emergency Rescue Unmanned Aerial Vehicles

Jiahang Zhu — 2026-05-21

In response to the core pain points in emergency rescue scenarios, such as the strong heterogeneity of multi-modal sensing signals, low efficiency of collaborative processing, and insufficient hardware adaptability, this paper reviews the application status of multi-modal sensing technology in emergency rescue fields, analyzes the key issues of the existing signal processing system, and constructs a three-level modular signal processing architecture of “sensing–transmission–pre-processing”. The sensing layer inputs the raw signals and outputs signals in a unified data format; the transmission layer inputs multiple signals, outputs these signals with priority, different paths, and strategies; the pre-processing layer processes the input signals and transmits the information to the unmanned aerial vehicle (UAV) terminal. This research clearly defines the core principles and modular logic of the architecture design, and explores the transmission optimization, resource allocation, and hardware compatibility schemes of different signal types from a technical perspective, demonstrating the balance between the general and specific paths of the architecture. The research results can provide architecture-level design references for the multi-modal sensing module of the airborne-ground integrated emergency rescue UAV, solve the problem of heterogeneous signal collaborative processing, and improve the overall response efficiency and reliability of the emergency rescue system.

Design of UWB-AOA Circularly Polarized Antenna for Automotive Digital Key Systems

Binghao Zeng — 2026-05-21

This paper designs an ultra-wideband (UWB) circularly polarized antenna based on Angle of Arrival (AOA) positioning technology. The antenna element adopts a rectangular dual-feed microstrip patch structure, and realizes right-hand circular polarization (RHCP) radiation through a 90° feeding network, which effectively suppresses vehicle multipath interference and polarization mismatch problems; a compact L-shaped three-antenna array is built based on the elements, and the array layout and feeding network are optimized to ensure phase and amplitude consistency. The simulation analysis results show that the operating frequency band of the antenna element can completely cover 6.24~8.24 GHz, with a return loss of ≤ -10 dB and an axial ratio of ≤ 3.0 dB within this band; the isolation between ports is ≤ -21dB, the gain is stable at 3.55–5.05 dBi, and the AOA positioning error is ≤ 1.4° (positioning distance 0.5–5 m). The overall size is controlled at 42mm × 30mm × 2mm, which fully meets the application requirements of automotive digital keys for miniaturization, high precision, and anti-interference. It can be directly integrated into digital key terminals, improving the reliability of contactless unlocking and the security of anti-theft authentication.

Characteristic Frequency Injection Technology for Secondary Cables in Substation Renovation Projects

Zhengjun Li — 2026-05-21

This paper investigates key technologies for reliable electrical monitoring in complex power environments. Core techniques include anti-electromagnetic interference sensing, multi-source signal isolation and filtering, characteristic signal injection with non-contact identification, multi-channel real-time data acquisition, and rapid fault recognition with protection interruption. Experimental analysis demonstrates stable monitoring performance and effective fault detection under electromagnetic disturbance conditions.

An Improved YOLOv8-Based Algorithm for Industrial Metal Surface Defect Detection

Yinghao Zhu — 2026-05-22

To address challenges in industrial metal surface defect detection, including tiny defects, significant scale variation, and complex backgrounds, this study proposes an enhanced YOLOv8s-based model, termed MEAF-YOLOv8s. Based on the original YOLOv8s architecture, the model introduces several improvements to enhance feature extraction and multi-scale representation. First, a CSP-MSEE module is incorporated in the feature extraction stage to strengthen the capture of edge and detail information of tiny defects, thereby effectively alleviating the problem of insufficient feature representation for small targets. Second, an AFRBN module is introduced to establish long-range spatial dependencies. By leveraging global contextual information, the module suppresses texture background interference, while a re-parameterization strategy is adopted to maintain the lightweight nature of the model and ensure that inference efficiency is not compromised. In addition, a CA-HFPN feature fusion structure is employed, which incorporates a direction-aware coordinate attention mechanism and a hierarchical pyramid architecture to promote precise cross-scale feature alignment and adaptive fusion, thereby improving the model’s adaptability and localization accuracy for defects of different sizes. To evaluate the proposed method, experiments are carried out on seven common defect types collected from real industrial environments. The results indicate that MEAF-YOLOv8s improves mAP50 by 4.72% and mAP50-95 by 1.28%, while decreasing the number of parameters by approximately 5M. These findings confirm that the proposed model can effectively enhance defect detection performance under complex background conditions.

Deep Learning Prediction Model for Dynamic Response of Bridge Cranes

Guanlan Li — 2026-05-21

The transient dynamic response of a bridge crane’s lifting mechanism is critical for operational safety and structural fatigue life. While traditional multi-body dynamics simulations offer high fidelity, their substantial computational cost hinders real-time analysis in digital twin applications. To overcome this bottleneck, this paper proposes a deep learning surrogate model based on a Long Short-Term Memory (LSTM) network for rapid prediction of the transient dynamics in a double-girder bridge crane’s lifting system. First, a high-fidelity dynamic benchmark model incorporating wire rope flexibility and contact friction is developed in ADAMS. Second, a high-quality dataset of 400 samples is constructed via Latin Hypercube Sampling, covering variations in load, lifting height, speed, and acceleration. Third, a three-layer encoder-LSTM-decoder network is designed and trained using a cosine annealing learning rate schedule and the AdamW optimizer. Experimental results demonstrate that the proposed model achieves excellent prediction accuracy, with a normalized mean absolute error (NMAE) of 0.0431, a normalized root mean square error (NRMSE) of 0.0681, and an average peak relative error of 4.72%, meeting engineering requirements. Most notably, the prediction time is reduced from approximately 30 minutes per simulation to 300 milliseconds, representing a computational efficiency improvement by a factor of about 6000 compared to conventional dynamic simulation.

New Iterative Methods for Solving Exponentially General Regularized Nonconvex Variational Inequality

Siyu Zhang — 2026-05-21

In this paper, we introduce and study a new class of extended exponentially general regularized nonconvex variational inequalities. We showed that the inequalities are equivalent to fixed-point problems through the use of the projection properties. Based on this equivalence, we discuss the existence and uniqueness of solutions to the extended exponentially general regularized nonconvex variational inequalities. We present a new self-adaptive finite p-step iterative projection scheme that uses multiple updates to obtain common solutions of the extended exponentially general regularized nonconvex variational inequality. Furthermore, we analyze the convergence of this algorithm under various suitable conditions.

Simulation Study on Corrosion Behavior of High Entropy Oxide Modified Epoxy Coating

Zhongli Dong — 2026-05-21

The widespread application of aluminum alloys in automotive lightweighting is frequently compromised by corrosion initiated at coating defects, yet the specific failure mechanisms governed by defect morphology remain difficult to isolate via traditional experimentation. This study employs a two-dimensional multiphysics coupling model, constructed on the COMSOL platform, to quantitatively investigate the corrosion evolution of epoxy-coated aluminum alloys containing two representative defect types: interfacial delamination and through-thickness bubbles. By integrating electrochemical kinetic parameters, the simulation elucidates how defect geometry modulates potential fields, current density distributions, and ionic transport within the “coating–electrolyte–substrate” system. Results indicate that delamination thickness positively correlates with corrosion rates; thicker cavities (50μm) significantly reduce ohmic resistance, thereby accelerating micro-galvanic coupling between the α-Al matrix and β-phase and expanding the corrosion domain. Furthermore, a distinct size effect was observed for through-thickness bubbles: large-scale defects (50μm) facilitate unimpeded mass transport, leading to severe lateral propagation and extensive substrate loss due to stable charge loops. Conversely, small-scale defects (20μm) induce a pronounced occluded cell effect, generating high local current densities that drive rapid vertical penetration, identifying them as the critical factor for deep-base pitting. These findings provide a theoretical basis for optimizing coating integrity and predicting localized corrosion behaviors in aluminum alloy structures.

From Lithium Plating Onset to Severe Deposition: A Three-Electrode-Based Grading of Lithium Plating

Yu Fan — 2026-05-21

Lithium (Li) plating is a critical barrier to fast charging in lithium-ion batteries as it accelerates degradation and compromises safety. In practice, however, identifying plating onset alone is insufficient; real-time evaluation of plating severity is more valuable for charging control. Here, a lithium plating grading framework is established using the evolution of negative-electrode potential in a three-electrode pouch cell. By operando tracking of the graphite anode potential during charging, the transition from stable Li intercalation to plating-prone conditions is identified and correlated with different degrees of metallic lithium deposition. Dynamic electrochemical impedance spectroscopy (DEIS) is employed as an auxiliary tool to probe the evolution of the overall interfacial response. Combined with post-mortem SEM, EDS, and XPS analyses, lithium plating is classified into four levels: plating-free, mild, moderate, and severe. The results show that increasingly negative anode potential corresponds to intensified Li deposition and progressive changes in surface morphology and interfacial chemistry. This work provides a severity-resolved approach to lithium plating diagnosis and offers a practical basis for fast-charging risk assessment and charging protocol optimization.

3D Object Detection Algorithm based on Improved PV-RCNN

Zhicheng Zhao — 2026-05-21

In autonomous driving perception, point cloud-based 3D object detection plays an important role. This task still faces two challenges in long-range and small-object detection: loss of fine details and weak context modeling. To solve these problems, this paper proposes HFA-RCNN based on PV-RCNN. The method adds an encoder-decoder structure to the 3D sparse convolution backbone. This design improves multi-scale context modeling and preserves more detailed features. In the BEV feature generation stage, the method also designs a spatial-frequency aggregation network. This network combines complementary information from the spatial domain and the frequency domain. This design improves feature representation. Results on the KITTI dataset show that the proposed method preserves strong detection performance for the Car category and further improves detection accuracy for the Pedestrian and Cyclist categories. These results confirm the effectiveness of the method in long-range and small-object detection.

Design of a Programming Trainer Based on Narrative Theory

Huitong Zeng — 2026-05-21

Addressing the high cognitive barriers and abstract nature of early programming for children aged 5–8, this study integrates narrative theory into the design of Tangible User Interfaces (TUIs). We developed a five-dimensional narrative model encompassing themes, characters, actions, scenes, and props to mitigate learners’ cognitive load through contextualized representation. A one-week comparative experiment demonstrated that children in the narrative tangible programming group significantly outperformed those in traditional computer programming and abstract tangible programming groups in terms of core concept comprehension, task efficiency, and self-correction proficiency. The findings suggest that narrative design achieves the “de-abstraction” of programming logic by embedding it into concrete storylines, fostering deep logical understanding in autonomous learning environments. This research provides valuable insights and design pathways for the development of early programming educational tools.

Improved Exam Paper Score Detection Algorithm Based on YOLO11n

Ruilin Mu — 2026-05-21

To address the challenges in detecting scores and related information on test papers, such as complex backgrounds and diverse handwriting styles, this paper proposes an improved algorithm based on YOLO11n. A Difference-of-Gaussians downsampling module, DOG-Stem, is designed to enhance edge feature extraction. Moreover, a lightweight grouped detection head, EfficientHead, is constructed, reducing parameters and computational complexity by 10.5% and 15.9%, respectively, while maintaining high performance. Finally, the WIoU loss function is introduced to accelerate model convergence. Experimental results demonstrate that the improved model achieves an mAP50 of 96.3% and an mAP50-95 of 68.6% on the test set, representing increases of 1.3% and 1.6% over the original YOLO11n. The proposed model exhibits superior precision and robustness.

Optimization and Implementation of a Lightweight Neural Network Architectures for Edge Computing

Xianke Meng — 2026-05-21

For the resource constraints and real-time requirements of edge computing scenarios, this paper systematically studies the optimization and implementation method of a lightweight neural network architecture. From the constraints of edge device computing power, power consumption constraints and the diversity of deployment environment, this paper discusses the architecture design strategy based on convolution decomposition, efficient embedding of attention mechanism and width depth collaborative optimization, and then proposes a comprehensive optimization scheme of training perception quantization, pruning and neural architecture search combined optimization, reasoning engine collaborative adaptation and end-to-end dynamic optimization. Through the deep integration of algorithm innovation and hardware characteristics, a complete technical framework from model design to edge deployment is constructed, aiming to achieve a good balance between accuracy and efficiency under resource constraints, and provide theoretical support and practical guidance for the model selection and deployment optimization of edge intelligent system.

SpecBEV-IR: Illumination-Robust Front-End Enhancement for Multi-View BEV 3D Object Detection

Yu Lin — 2026-05-21

Multi-view visual BEV 3D object detection projects image information from different camera views into a unified bird’s-eye-view space and has become an important paradigm for autonomous driving perception due to its low cost, flexible deployment, and rich semantic information. However, under complex lighting conditions such as nighttime, backlighting, local overexposure, and uneven illumination, multi-view input images often suffer from degraded brightness distribution, local contrast, and structural details, which further affects image feature extraction, view transformation, and unified spatial modeling. To address this issue, this paper proposes SpecBEV-IR, an illumination-robust multi-view BEV 3D object detection method. Built upon the SpecBEV framework, the proposed method introduces an illumination-robust image front-end enhancement module, termed ICF, between the multi-view input images and the shared 2D encoder. The ICF module consists of an invariant cue extraction unit (ICE) and a fusion convolution unit (Fuse Conv). ICE extracts more stable illumination-invariant cues from raw images, while Fuse Conv integrates these cues with the original image content to generate enhanced input representations for subsequent feature encoding and view transformation. Different from conventional enhancement methods that mainly improve visual appearance, SpecBEV-IR emphasizes structural stability and cross-view consistency for downstream 3D detection. Experiments on the nuScenes dataset show that SpecBEV-IR achieves 0.4121 mAP and 0.5174 NDS on the validation set, while also obtaining better or more balanced performance on multiple error metrics, including mATE, mASE, mAOE, and mAAE. The results demonstrate that the proposed method effectively improves the overall robustness and detection performance of multi-view visual 3D object detection under complex lighting conditions.

Robust Multi-Source Odometry Based on Cascaded Filtering and Hierarchical Optimization

Xiang Fu — 2026-05-21

Aiming at the trajectory drift and long-term computing power bottleneck of urban service robots in large-scale scenarios, this paper proposes a practical LIO-RTK-PGO multi-source fusion odometry. The front-end adopts a cascaded tightly coupled architecture, introducing RTK observation to correct the state in the filtering prediction stage to fundamentally suppress elevation and heading divergence; the back-end proposes hierarchical pose graph optimization (PGO), combining local high-frequency sliding window and global keyframe sparsification to control the computational complexity at O(1). Verified by real-vehicle tests and standard computing power platforms, the system eliminates long-range cumulative errors, providing a low-computing-power and highly reliable state estimation scheme for large-scale engineering implementation.

Tree Sap Flow Prediction Based on the Fusion of CEEMDAN-Copula Entropy-LSTM

Zixiang Wang — 2026-05-21

Copula entropyTree trunk sap flow is jointly affected by environmental factors and physiological mechanisms, showing nonlinear and random characteristics, which makes it difficult for traditional methods to achieve high-precision prediction. To address this problem, this paper introduces CEEMDAN to decompose the sap flow sequence at multiple scales, combines Copula entropy and signal energy to construct a modal component reconstruction strategy, and further uses LSTM to realize prediction. Experimental results show that the proposed model achieves 0.6759 and 0.9755 in MAPE and R² indicators respectively, which is superior to the comparison models, providing a new idea for sap flow prediction and transpiration flux estimation.

A Multimodal Sensor Fusion Approach for Real- Time Detection of Electric Vehicles in Elevators

Fenglin Zhang — 2026-05-21

This paper presents a multimodal sensing-based system designed to improve the accuracy and real-time performance of detecting electric vehicles (EVs) attempting to enter elevators. The system architecture integrates a data acquisition layer, an analysis and processing layer, an intelligent decision layer, and an application layer. By employing sensors such as flexible tactile arrays and visual cameras, the system captures multimodal data. Detection results are generated through a global scene context information extraction module and a local spatio-temporal feature extraction module. Experimental outcomes indicate that the proposed system achieves a mean average precision (mAP) of 95.60%, operates at 31.5 frames per second (FPS), and maintains a model size of 26 MB. Under occluded conditions, it attains average mAP@.5 and mAP@.5:.95 scores of 91.2% and 80.9%, respectively, surpassing other existing detection methods. The results demonstrate that this multimodal sensing system can effectively mitigate safety risks associated with EVs entering elevators, offering high practical utility and reliability.

Resilience Investment for Distribution Networks Facing Urban Waterlogging: A Two-Stage Method Combining AHP Zoning and Mixed-Integer Optimization

Weiwei Wu — 2026-05-21

Against the background of frequent urban waterlogging disasters caused by global climate change, the vulnerability of distribution systems has become increasingly prominent. To address the blindness and inefficiency of traditional protective equipment layout, this paper proposes a two-stage optimization method based on risk zoning, aiming to scientifically guide the layout of protective equipment represented by waterproof transformers. First, the method integrates six core indicators including rainfall, elevation, and slope, and uses the Analytic Hierarchy Process (AHP) to conduct refined waterlogging risk zoning of the distribution network, dividing nodes into three risk levels: high, medium, and low. Taking this risk assessment result as a key input, a mixed-integer programming model with the goal of maximizing comprehensive benefits is constructed. Finally, the model is verified through a 25-node numerical example. The results show that through strategic layout, the model protects most loads including high-risk nodes with limited costs, and makes a strategic abandonment of low-value nodes in line with the optimization goal. This study has important practical significance for improving investment efficiency and urban power grid resilience.

Research on the Application of Computer Vision in Equipment Fault Diagnosis

Xiaoquan Zhu — 2026-05-21

With the continuous improvement of industrial automation, rapid and accurate diagnosis of equipment faults is the key to ensuring production safety and efficiency. With the advantages of non-contact sensing, real-time processing and high-precision recognition, computer vision has broad application prospects in fault diagnosis. This technology integrates image acquisition, feature extraction and deep learning models to automatically identify and classify equipment faults such as appearance damage, motion abnormalities and thermal state changes. Multi-modal image fusion further improves fault positioning accuracy under complex working conditions. In scenarios such as mine electrical equipment, construction engineering inspection cold-chain storage and unmanned aerial vehicle (UAV) inspection, its detection performance is superior to traditional methods, providing strong technical support for building an intelligent equipment operation and maintenance system and promoting the in-depth integration of industrial Internet and intelligent manufacturing.

Research on Fault Diagnosis and Intelligent Maintenance Technology of Airborne Electronic Equipment

Zhaopeng Jin — 2026-05-21

With the increasing integration and complexity of avionic systems, fault diagnosis and intelligent maintenance technologies for airborne electronic equipment have become critical supports for ensuring flight safety and improving equipment integrity. This paper systematically reviews the research status and development context of fault diagnosis technologies for airborne electronic equipment. It summarizes major research achievements and technological advances in the field from the perspectives of traditional fault diagnosis methods, integrated intelligent diagnosis strategies, artificial intelligence-driven technologies, data-driven methods, and intelligent maintenance assistance systems. On this basis, core bottlenecks in current research are analyzed, including data dependency, poor model interpretability, and insufficient generalization ability. Future development directions are prospected, such as few-shot learning, explainable AI, digital twin, and edge intelligence.

Research on Coherence Optimization in E-commerce Video Advertising Generation Using AIGC

Xiaoyao Xiong — 2026-05-21

E-commerce video advertising is one of the primary means of e-commerce marketing, and the coherence of e-commerce video advertising directly affects the efficiency of information transmission. Current mainstream AIGC models tend to encounter issues such as visual stuttering, logical inversion, and excessive stiffness when generating long-sequence e-commerce advertisements, leading to narrative confusion and unclear content expression in video advertisements. This paper explores optimization strategies for enhancing the coherence of e-commerce advertising generation through literature review and case analysis. The results indicate that the optimization methods improve advertisement coherence and completion rates. This study provides an operational optimization path for the controllable generation of AIGC in e-commerce video advertising.

Exploration of Automated Operation Technology for Power Transmission, Distribution, and Utilization Engineering

Jiahe Yan — 2026-05-21

The rapid development of the socio-economy has driven a continuous increase in electricity demand, placing higher requirements on the operation of the power system. Power transmission, distribution, and utilization engineering play a crucial role in the entire power system, with their operational efficiency directly affecting the overall stability of the power supply. This paper analyzes the application advantages of automated operation technology in power transmission, distribution, and utilization engineering, and examines specific applications of relevant technologies, such as holographic perception and real-time monitoring technology, edge computing and cloud computing collaboration technology, multi-source data integration technology, adaptive control technology, and network security defense technology. Based on this, strategies for the efficient automated operation of power transmission, distribution, and utilization engineering are proposed, providing valuable references for the long-term development of power engineering.

Advances in Cement Bond Logging Technology and Issues in Cement Job Quality Evaluation

Fei Li — 2026-05-21

The Cement Bond Log (CBL) is currently the primary method used in domestic oilfields to evaluate cement job quality. However, its interpretation results are prone to ambiguity, leading to misjudgment and erroneous conclusions. This results in artificially high reported pass rates for cement jobs, which hinders the development of cementing technology. This paper systematically analyzes the limitations of CBL in evaluating cement job quality, discusses in detail ten influencing factors including tool eccentricity, fast formations, micro-annulus, cement sheath thickness, gas-cut drilling fluid, casing parameters, localized channeling, cement slurry properties, logging timing, and acoustic frequency. It also points out inherent shortcomings of CBL in evaluating the second interface, identifying thin beds, locating channels, and providing quantitative interpretation. Research indicates that the consistency rate of CBL in reflecting actual cement job quality is only about 33%. In contrast, the Sector Bond Tool (SBT) offers technical advantages through multi-parameter visual display and cement sheath imaging. It is recommended that SBT be used as a means for detailed evaluation of cement job quality in key or problematic wells.

Design and Optimization of Visual Algorithms for Inspection Robots

Yong Zhang — 2026-05-21

As an important equipment for intelligent operation and maintenance, inspection robots have been widely used in high-risk and complex scenarios such as power, mining, and chemical industries. The visual system, as the “eyes” of inspection robots, undertakes tasks including image enhancement, navigation and positioning, target recognition, and error correction, and its performance directly affects the robots’ autonomous operation capabilities. Currently, the visual algorithms of inspection robots still face several problems, such as poor adaptability to complex environments, insufficient navigation accuracy, difficulty in balancing target recognition accuracy and real-time performance, and weak adaptability of error correction. Combined with the current application status of inspection robots, this paper elaborates on the design ideas of the four major modules of visual algorithms and proposes optimization strategies for existing problems, providing references for improving the autonomous inspection capabilities of inspection robots and promoting the upgrading of intelligent inspection technology.