Feature Information Research Articles

Weak saliency ensemble network for person Re-identification using infrared light images

In recent years, person re-identification (re-id) has primarily been studied using visible light (VL) images. However, the challenges of employing VL images in nighttime environments have prompted research into using infrared light (IR) images. Yet, the utilization of both VL and IR images in person re-id has resulted in increased computational cost and processing time in multi-modality systems, leading to studies focusing solely on IR images. Nevertheless, IR images, lacking color and texture information, generally yield lower recognition performance in existing person re-id studies. In addition, previous studies have shown that person re-id performance suffers in the presence of complex background noise. To tackle these challenges, this study proposes a new weak saliency ensemble network (WSE-Net) for person re-id using IR images. WSE-Net incorporates a channel reduction of feature (CRF) method to reduce computational cost in the ensemble network, a technique for converting input images into group of patch images and feeding them into the ensemble model to enhance the reduced feature information, and a grouped convolution ensemble network (GCE-Net) that enables the fusion of features extracted from original and attention-guided ensemble models.The performance of person re-id using WSE-Net was evaluated on the Dongguk body-based person recognition database version 1 (DBPerson-Recog-DB1) and the Sun Yat-sen university multiple modality re-identification version 1 (SYSU-MM01). Experimental results demonstrated that on DBPerson-Recog-DB1, WSE-Net achieved 93.65% in rank 1, 95.28% in mean average precision (mAP), and 93.52% in the harmonic mean of precision and recall. Additionally, on SYSU-MM01, WSE-Net achieved 86.85% in rank 1, 44.58% in mAP, and 40.06% in the harmonic mean of precision and recall. Furthermore, the accuracy of WSE-Net on both datasets surpassed that of state-of-the-art methods.

Digital signal detection and recognition in the communication field combining DAEN and CNN

In communication, the detection and recognition of digital signals have always faced problems such as low adaptability and high misclassification rate. To address these issues, this study innovatively fused the deep auto-encoder network with convolutional neural networks to construct a novel digital signal detection and recognition model. Firstly, this study utilized the powerful feature extraction capabilities of the deep auto-encoder network to extract key feature information from massive amounts of data. Then these features were combined with convolutional neural networks to construct a detection and recognition model. These results confirmed that the constructed digital signal detection and recognition model had values of 93.75% and 94.18% in data detection recognition rate and average classification accuracy, respectively. Meanwhile, this model also performed well in terms of data processing accuracy and recall. In the comparison of data processing, the accuracy and recall rates were 93.59% and 94.67%, respectively, and the performance of data detection and recognition was better than that of the comparison methods. This indicates that the constructed digital signal detection and recognition model can significantly improve the reliability and robustness of signal detection and recognition. This paper brings new breakthroughs to the development of digital signal detection and recognition technology in communication.

Optimizing flotation froth image segmentation via parallel branch network and hybrid loss supervision

Flotation is a crucial technology for fine coal separation, and accurately acquiring bubble size information during the flotation process is essential for monitoring flotation conditions and achieving intelligent control. However, existing semantic segmentation models encountered issues with boundary disconnection when segmenting flotation bubbles, resulting in deviations between the extracted bubble sizes and their true values. To address the aforementioned challenges, a semantic segmentation model was proposed to maintain high-resolution feature maps throughout the network by designing a parallel branch network structure. Additionally, a ConvTranspose module was proposed to preserve the detailed feature information of images while gradually enhancing the resolution of feature maps. In the model training phase, a hybrid loss function combining pixel classification loss with shape similarity loss was proposed to alleviate the sample imbalance problem caused by the substantial difference in the number of pixels between bubble boundaries and the interior of bubbles. Moreover, since traditional semantic segmentation evaluation metrics, such as MIoU, lack a mechanism for measuring bubble boundary continuity and cannot effectively penalize the problem of boundary disconnection, this paper proposed a new evaluation method for assessing the segmentation performance of flotation froth images. To comprehensively evaluate the effectiveness of the proposed method, this paper conducted tests using flotation froth images collected from actual production processes. Compared with existing methods, the segmentation model proposed in this paper exhibited clear superiority in mitigating the problem of bubble boundary disconnection. The prediction error for the number of bubbles was 6.38 %, which is significantly better than other methods.

IMCMK-CNN: A lightweight convolutional neural network with Multi-scale Kernels for Image-based Malware Classification

Rapid and accurate identification of unknown malware and its variants is the premise and basis for the effective prevention of malicious attacks. However, with the explosive growth of malware variants, the efficiency of manual updating of the sample database is getting worse and worse. It is difficult for the traditional identification methods to effectively capture the sample feature information operated by the confusion method only based on the delayed database information. The research into the direction of malware detection is dedicated to surmounting the limitations of conventional detection methodologies, and delves deeply into the application of cutting-edge technologies such as data visualization, machine learning, and hybrid detection within the realm of malware detection. Through these investigations, our goal is to construct a detection system that is both more precise and efficient, capable of addressing the ever-evolving threats to cybersecurity. Pursuing research in this direction is not only vital for enhancing network security defenses and safeguarding user data, but it will also foster the advancement of related state-of-the-art technologies and further mitigate the economic and societal repercussions of malware attacks. In light of this issue, this paper proposes the Image-based Malware Classification with Multi-scale Kernels (IMCMK), a Convolutional Neural Network (CNN) architecture using multi-scale convolution kernels mixing action to improve malware variants detection capabilities. First, we propose the Multi-scale Kernels (MK) block combining deep large kernel convolution and standard small kernel convolution with shortcuts to improve the accuracy. Furthermore, we propose Multi-scale Kernel Fusion (MKF) to reduce the number of parameters that come with the large kernels. The improved Squeeze-and-Excitation (SE) block can obtain the correlation between different channels to further increase the model performance. Experimental results show that IMCMK outperforms the state-of-the-art methods in malware family classification accuracy, which has achieved 99.25 %.

YOLO-GE: An Attention Fusion Enhanced Underwater Object Detection Algorithm

Underwater object detection is a challenging task with profound implications for fields such as aquaculture, marine ecological protection, and maritime rescue operations. The presence of numerous small aquatic organisms in the underwater environment often leads to issues of missed detections and false positives. Additionally, factors such as the water quality result in weak target features, which adversely affect the extraction of target feature information. Furthermore, the lack of illumination underwater causes image blur and low contrast, thereby increasing the difficulty of the detection task. To address these issues, we propose a novel underwater object detection algorithm called YOLO-GE (GCNet-EMA). First, we introduce an image enhancement module to mitigate the impact of underwater image quality issues on the detection task. Second, a high-resolution feature layer is added into the network to improve the problems of missed detections and false positives for small targets. Third, we propose GEBlock, an attention-based fusion module that captures long-range contextual information and suppresses noise from lower-level feature layers. Finally, we combine an adaptive spatial fusion module with the detection head to filter out conflicting feature information from different feature layers. Experiments on the UTDAC2020, DUO and RUOD datasets show that the proposed method achieves an optimal detection accuracy.

Database Construction and Application of Large-scale Geospatial Data

Abstract. Geospatial data refers to data that related to spatial location. It is the abstract representation of objects, events, or other characteristic information located on or near the surface of the Earth. It characterizes attribute features, spatial features, and associated information of geospatial features. Geospatial data commonly include various data types such as point, polyline, polygon, and so on. Geospatial database is the basic method to store and manage geospatial data. It is also the important component of geographic information systems (GIS). Geospatial database generally stores data in a specific structure in a medium, which can achieve efficient management of geospatial data. Users can add, modify, and delete data efficiently. Furthermore, query retrieval and statistical analysis can be quickly performed based on geospatial database. In this paper, distributed geospatial database is constructed by using database partition tables storage technology. Meanwhile, database construction software is developed to achieve the construction, management, and application of large-scale geospatial data. It can improve data import, query and retrieval efficiency significantly. The database application service system can directly connect to the database for displaying statistical analysis results. At the same time, geospatial data are processed to vector tiles, which stored in the tile database. Furthermore, standard online service is then published, which can be used for database management and application service system. It can be used for natural resource management work such as basic surveying, survey and monitoring, and so on.

ILMCNet: A Deep Neural Network Model That Uses PLM to Process Features and Employs CRF to Predict Protein Secondary Structure.

Protein secondary structure prediction (PSSP) is a critical task in computational biology, pivotal for understanding protein function and advancing medical diagnostics. Recently, approaches that integrate multiple amino acid sequence features have gained significant attention in PSSP research. We aim to automatically extract additional features represented by evolutionary information from a large number of sequences while simultaneously incorporating positional information for more comprehensive sequence features. Additionally, we consider the interdependence between secondary structures during the prediction stage. To this end, we propose a deep neural network model, ILMCNet, which utilizes a language model and Conditional Random Field (CRF). Protein language models (PLMs) pre-trained on sequences from multiple large databases can provide sequence features that incorporate evolutionary information. ILMCNet uses positional encoding to ensure that the input features include positional information. To better utilize these features, we propose a hybrid network architecture that employs a Transformer Encoder to enhance features and integrates a feature extraction module combining a Convolutional Neural Network (CNN) with a Bidirectional Long Short-Term Memory Network (BiLSTM). This design enables deep extraction of localized features while capturing global bidirectional information. In the prediction stage, ILMCNet employs CRF to capture the interdependencies between secondary structures. Experimental results on benchmark datasets such as CB513, TS115, NEW364, CASP11, and CASP12 demonstrate that the prediction performance of our method surpasses that of comparable approaches. This study proposes a new approach to PSSP research and is expected to play an important role in other protein-related research fields, such as protein tertiary structure prediction.

Explorative Binary Gray Wolf Optimizer with Quadratic Interpolation for Feature Selection.

The high dimensionality of large datasets can severely impact the data mining process. Therefore, feature selection becomes an essential preprocessing stage, aimed at reducing the dimensionality of the dataset by selecting the most informative features while improving classification accuracy. This paper proposes a novel binary Gray Wolf Optimization algorithm to address the feature selection problem in classification tasks. Firstly, the historical optimal position of the search agent helps explore more promising areas. Therefore, by linearly combining the best positions of the search agents, the algorithm's exploration capability is increased, thus enhancing its global development ability. Secondly, the novel quadratic interpolation technique, which integrates population diversity with local exploitation, helps improve both the diversity of the population and the convergence accuracy. Thirdly, chaotic perturbations (small random fluctuations) applied to the convergence factor during the exploration phase further help avoid premature convergence and promote exploration of the search space. Finally, a novel transfer function processes feature information differently at various stages, enabling the algorithm to search and optimize effectively in the binary space, thereby selecting the optimal feature subset. The proposed method employs a k-nearest neighbor classifier and evaluates performance through 10-fold cross-validation across 32 datasets. Experimental results, compared with other advanced algorithms, demonstrate the effectiveness of the proposed algorithm.

Channel Interaction and Transformer Depth Estimation Network: Robust Self-Supervised Depth Estimation Under Varied Weather Conditions

Monocular depth estimation provides low-cost environmental information for intelligent systems such as autonomous vehicles and robots, supporting sustainable development by reducing reliance on expensive, energy-intensive sensors and making technology more accessible and efficient. However, in practical applications, monocular vision is highly susceptible to adverse weather conditions, significantly reducing depth perception accuracy and limiting its ability to deliver reliable environmental information. To improve the robustness of monocular depth estimation in challenging weather, this paper first utilizes generative models to adjust image exposure and generate synthetic images of rainy, foggy, and nighttime scenes, enriching the diversity of the training data. Next, a channel interaction module and Multi-Scale Fusion Module are introduced. The former enhances information exchange between channels, while the latter effectively integrates multi-level feature information. Finally, an enhanced consistency loss is added to the loss function to prevent the depth estimation bias caused by data augmentation. Experiments on datasets such as DrivingStereo, Foggy CityScapes, and NuScenes-Night demonstrate that our method, CIT-Depth, exhibits superior generalization across various complex conditions.

Non-intrusive identification of building loads using EDCA-ShuffleNetV2 with fused feature visualization

Abstract Non-intrusive load monitoring (NILM) enables monitoring of appliance operation status and energy consumption without additional meters, providing innovative data support for energy management. However, current NILM systems still face challenges such as insufficient load feature information and low load identification accuracy. To utilize the load feature information and improve the accuracy of load identification more effectively, this paper first selects loads with high energy consumption and high usage frequency as the research object, makes full use of the load feature information through feature fusion, and converts the fused typical load data into an image through the Gramian summation angular field, and further carries out the identification of loads that still have similarity after feature fusion. Secondly, this paper improves the ShuffleNetV2 model structure by embedding an efficient dual-channel attention (EDCA) module in the basic feature extraction module of ShuffleNetV2 to enhance it to extract adequate load feature information. A residual structure is also introduced to mitigate the information loss and gradient disappearance issues of the model. Finally, the superiority and effectiveness of the fused load features and the proposed model in load identification and typical scenario applications are validated using the iAWE and AMPDS datasets.

A LiDAR SLAM Algorithm Considering Dynamic Extraction of Feature Points in Underground Coal Mine

Abstract. To address the absence of GNSS signals in underground coal mines and the susceptibility of mainstream LiDAR SLAM to degradation due to insufficient feature constraints, this paper proposes a tightly-coupled SLAM algorithm incorporating LiDAR and IMU for use in such environments. The paper initially introduces a dynamic feature point extraction strategy, where the number of corner feature points can be dynamically adjusted based on the occurrence of degradation in the underground coal mine environment. This approach constructs a constraint matrix with rich and robust feature information, enhancing pose estimation accuracy in environments with inadequate feature constraints, mitigating degradation effects, and reducing global cumulative error. Subsequently, the consistent building of the underground coal mine environment is achieved through back-end factor graph optimization. Finally, to validate the effectiveness of the method, experimental analysis is conducted in an underground coal mine. The results demonstrate that LeGO_LOAM exhibits poor robustness in underground coal mines and fails to construct a globally consistent pose estimation and map. Conversely, the pose estimation error of the proposed method in this paper is 50.93% lower in the plane direction and 42.13% lower in the elevation direction compared to LIO_SAM. This underscores the method's significance as crucial technical support for the intelligent perception and positioning of robots in underground coal mines.

Driver Distraction Detection Based on Fusion Enhancement and Global Saliency Optimization

Driver distraction detection not only effectively prevents traffic accidents but also promotes the development of intelligent transportation systems. In recent years, thanks to the powerful feature learning capabilities of deep learning algorithms, driver distraction detection methods based on deep learning have increased significantly. However, for resource-constrained onboard devices, real-time lightweight models are crucial. Most existing methods tend to focus solely on lightweight model design, neglecting the loss in detection performance for small targets. To achieve a balance between detection accuracy and network lightweighting, this paper proposes a driver distraction detection method that combines enhancement and global saliency optimization. The method mainly consists of three modules: context fusion enhancement module (CFEM), channel optimization feedback module (COFM), and channel saliency distillation module (CSDM). In the CFEM module, one-dimensional convolution is used to capture information between distant pixels, and an injection mechanism is adopted to further integrate high-level semantic information with low-level detail information, enhancing feature fusion capabilities. The COFM module incorporates a feedback mechanism to consider the impact of inter-layer and intra-layer channel relationships on model compression performance, achieving joint pruning of global channels. The CSDM module guides the student network to learn the salient feature information from the teacher network, effectively balancing the model’s real-time performance and accuracy. Experimental results show that this method outperforms the state-of-the-art methods in driver distraction detection tasks, demonstrating good performance and potential application prospects.

EIDU-Net: edge-preserved inception DenseGCN U-Net for LiDAR point cloud segmentation

With the development of laser scanners and machine learning, point cloud semantic segmentation plays a significant role in autonomous driving, scene reconstruction, human-computer interaction, and other fields. In recent years, point cloud semantic segmentation based on deep learning has become one of the key research directions in point cloud processing. Due to the limited ability to exploit geometric details and contextual information in point clouds, most methods that adopt encoder-decoder architecture lose local structural information easily, especially detailed features, and extract features insufficiently. To address this issue, the edge-preserving inception DenseGCN U-Net (entitled as EIDU-Net) is proposed. EIDU-Net makes full use of the complementation between geometric details in the original point cloud and high-level features. The edge-preserved graph pooling (EGP) layer, the key module of the EIDU-Net, is designed to retain additional edge feature information from the original point cloud during pooling operations. Accordingly, the edge-preserved graph unpooling (EGU) layer can restore the feature graph more efficiently based on the additionally retained edge features. Extensive experiments demonstrate that our proposed EIDU-Net has remarkable improvements on semantic segmentation tasks under whatever S3DIS or Terracotta Warrior fragments. Our code is publicly available at https://github.com/caoxin918/EIDU-Net.

Algorithm for detecting surface defects in wind turbines based on a lightweight YOLO model

Improving wind power generation efficiency and lowering maintenance and operational costs are possible through the early and efficient diagnosis and repair of surface defects in wind turbines. To solve the lightweight deployment difficulty and insufficient accuracy issues of the traditional detection methods, this paper proposes a high-precision PC-EMA block based on YOLOv8 using partial convolution (PConv) combined with an efficient multiscale attention (EMA) channel attention mechanism, which replaces the bottleneck layer of the YOLOv8 backbone network to improve the extraction of target feature information from each layer of the network. In the feature fusion phase, GSConv, which can retain more channel information, is introduced to balance the model’s complexity and accuracy. Finally, by merging two branches and designing the PConv head with a low-latency PConv rather than a regular convolution, we are able to effectively reduce the complexity of the model while maintaining accuracy in the detection head. We use the WIoUv3 as the regression loss for the improved model, which improves the average accuracy by 5.07% and compresses the model size by 32.5% compared to the original YOLOv8 model. Deployed on Jetson Nano, the FPS increased by 11 frames/s after a TensorRT acceleration.

RTINet: A Lightweight and High-Performance Railway Turnout Identification Network Based on Semantic Segmentation.

To lighten the workload of train drivers and enhance railway transportation safety, a novel and intelligent method for railway turnout identification is investigated based on semantic segmentation. More specifically, a railway turnout scene perception (RTSP) dataset is constructed and annotated manually in this paper, wherein the innovative concept of side rails is introduced as part of the labeling process. After that, based on the work of Deeplabv3+, combined with a lightweight design and an attention mechanism, a railway turnout identification network (RTINet) is proposed. Firstly, in consideration of the need for rapid response in the deployment of the identification model on high-speed trains, this paper selects the MobileNetV2 network, renowned for its suitability for lightweight deployment, as the backbone of the RTINet model. Secondly, to reduce the computational load of the model while ensuring accuracy, depth-separable convolutions are employed to replace the standard convolutions within the network architecture. Thirdly, the bottleneck attention module (BAM) is integrated into the model to enhance position and feature information perception, bolster the robustness and quality of the segmentation masks generated, and ensure that the outcomes are characterized by precision and reliability. Finally, to address the issue of foreground and background imbalance in turnout recognition, the Dice loss function is incorporated into the network training procedure. Both the quantitative and qualitative experimental results demonstrate that the proposed method is feasible for railway turnout identification, and it outperformed the compared baseline models. In particular, the RTINet was able to achieve a remarkable mIoU of 85.94%, coupled with an inference speed of 78 fps on the customized dataset. Furthermore, the effectiveness of each optimized component of the proposed RTINet is verified by an additional ablation study.

Triboelectric signal enhancement via interface structural design and integrated with deep learning for real-time online material recognition

Triboelectric nanogenerator (TENG) is a promising technology for material recognition due to a large amount of distinguishable feature information in triboelectric signal when the TENG device touches different materials. Nevertheless, the identification precision has only improved from the perspective of sensor construction and operation of design, rather than from the electrodes and the triboelectric interface structure. To prevent such issues, a design strategy of electrode/triboelectric material interface structure is applied to enhance the output performance and recognition accuracy. Electroless nickel was plated on a sponge to fabricate a TENG device electrode with a roughness of 179.1 µm. A coarse triboelectric material/electrode interface can improve the output performance of corresponding device. Hence, the nickel layer surface generates more the triboelectric charges, further leading to characteristic triboelectric signals, which can be distinguished by deep machine learning. As expected, the collected TENG signal has the unique characteristic of being able to identify the touched material with a recognition accuracy of 96 %. Based on this, combined with deep machine learning and the triboelectric effect, a material awareness system integrating TENG-based sensors, data collation, and display components was also designed and built for real-time material type identification with superior accuracy. The potential application advantages of the manufactured TENG device in nonvisual intelligent recognition can be foreseen.

Rolling bearing fault diagnosis method based on MSDCNN in strong noise environment

Aiming at the problems of poor noise resistance, high computational complexity and insufficient generalization performance of traditional bearing fault diagnosis methods based on deep learning, a rolling bearing fault diagnosis method based on multi-scale dynamic convolutional neural network (MSDCNN) is proposed. Firstly, the one-dimensional vibration signal of the rolling bearing is converted to the frequency domain by Fourier transform, and the features are further extracted by wide convolution kernel. Secondly, a multi-scale dynamic convolution structure is proposed, and the feature information extracted by convolution kernels of different sizes is given different weights by using an improved channel attention mechanism. Then, a self-calibrating spatial attention mechanism (SCSAM) is designed, and the extracted feature information is input into the spatial attention mechanism to capture the importance of different regions. Finally, the features are further extracted by small convolution kernels, and the fault category is classified by using Softmax classifier. The fault diagnosis performance of the proposed model is verified by three different data sets. The experimental results show that compared with other intelligent models, the proposed model has higher classification accuracy, better generalization ability and stronger robustness under strong noise background.

The Fusiongram: a periodic weak fault feature extraction strategy and its application in bearing fault diagnosis

Abstract The weak periodic transient impact responses caused by localized defects in rolling bearings are often obscured by complex interferences, such as white noise, random transient impact responses, and periodic responses from system operations. Meanwhile, the fault feature information contributing to damage detection may be distributed across different frequency bands in the vibration signal. Therefore, under the influence of complex interference, it is a challenging problem regarding how to 1) accurately select the frequency bands containing rich fault feature information, 2) effectively extract damage information from different frequency bands, and thereby, 3) successfully achieve fault diagnosis for machinery condition monitoring. To address these issues, this research introduces a novel signal processing strategy, termed as Fusiongram, for extracting weak periodic fault features amidst the influence of complex interferences. Firstly, the method of complementary hierarchical decomposition is proposed, in which the signal is decomposed into multiple components with overlapping frequency contents. Then, an index with interference resistance is constructed to select the components carrying rich damage feature information. Finally, the adaptive threshold denoising and multicomponent normalized averaging techniques are employed to fuse the information from the squared envelope spectra of the selected components, thus obtaining the reconstructed squared envelope spectra for fault diagnosis. The Fusiongram is able to achieve the goal of weak fault feature extraction from signals with complex interference. The analysis results of numerical simulation and experimental testing verify the effectiveness and advantages of the proposed strategy.

YOLOv8-CDD: an improved concrete defect detection method combined CNN with transformer

Abstract Concrete defect detection is a critical task to ensure the safe and stable operation of concrete infrastructure. In order to effectively solve the problems of low efficiency, high cost and poor performance of existing methods, a high-precision concrete defect detection method YOLOv8-CDD (Concrete Defect Detection) combining convolutional neural network and transformer is proposed in this paper. Firstly, based on the features with a large span of concrete defect features, a bot-transformer module that can effectively extract the global information of defect features is proposed to improve the network’s ability to extract global features. Secondly, in order to further strengthen the interaction between defect feature channels and spatial information, a convolutional triplet attention module is introduced into the feature enhancement network to effectively integrate the information of different dimensions of defect features and improve the model detection accuracy. Additionally, in order to enhance the learning of samples with different degrees of difficulty, the introduction of Focaler-CIoU instead of the original boundary regression loss function can optimize the model training process. Finally, the dataset was collected and organized in concrete scenarios from bridge towers, dams, and tunnel corridors, and our method achieved 0.898 average precision, 0.893 average recall, 0.031 average FPR, 0.895 average F1 score, 0.929 mAP50, and 0.731 mAP50:95 on the dataset. The experimental results show that the proposed method achieves the best performance in concrete defect detection.

Lightweight Indoor Positioning System Based on Multiple Self-Learning Features and Key Frame Classification

Abstract. Traditional indoor positioning technologies mostly require advanced installation of hardware devices, resulting in high costs and long-term maintenance. With advancements in image recognition and deep learning technologies, indoor visual positioning based on image recognition has become increasingly mature. This method offers the benefits of low cost and does not require additional hardware installation. However, it still has inherent defects, such as cumbersome data collection, complex algorithms, and universality. To minimize indoor information pre-collection cost, improve versatility, and enable rapid deployment in low-performance mobile devices, this paper proposes a lightweight indoor positioning system based on multiple self-learning features and key frame classification. The system is divided into two stages: preprocessing and real-time positioning. In the preprocessing stage, image information is collected for the entire indoor environment, and a key-frame recognizer is trained based on the image information. Simultaneously, an environmental feature information database is established. In the real-time positioning stage, the system first uses mobile devices such as smartphones to obtain real-time video streams. A key frame recognizer based on convolutional neural networks identifies key frames in each video stream frame, thereby obtaining approximate positions for rough positioning. Second, feature points are identified in each frame of the video stream and matched with feature points with location information in the built environmental feature information database to calculate precise positions for fine positioning. It has significant optimizations compared with conventional visual solutions in terms of preprocessing data collection, algorithm performance consumption, and versatility.