Convolution Module Research Articles

PII-GCNet: Lightweight Multi-Modal CNN Network for Efficient Crowd Counting and Localization in UAV RGB-T Images

With the increasing need for real-time crowd evaluation in military surveillance, public safety, and event crowd management, crowd counting using unmanned aerial vehicle (UAV) captured images has emerged as an essential research topic. While conventional RGB-based methods have achieved significant success, their performance is severely hampered in low-light environments due to poor visibility. Integrating thermal infrared (TIR) images can address this issue, but existing RGB-T crowd counting networks, which employ multi-stream architectures, tend to introduce computational redundancy and excessive parameters, rendering them impractical for UAV applications constrained by limited onboard resources. To overcome these challenges, this research introduces an innovative, compact RGB-T framework designed to minimize redundant feature processing and improve multi-modal representation. The proposed approach introduces a Partial Information Interaction Convolution (PIIConv) module to selectively minimize redundant feature computations and a Global Collaborative Fusion (GCFusion) module to improve multi-modal feature representation through spatial attention mechanisms. Empirical findings indicate that the introduced network attains competitive results on the DroneRGBT dataset while significantly reducing floating-point operations (FLOPs) and improving inference speed across various computing platforms. This study’s significance is in providing a computationally efficient framework for RGB-T crowd counting that balances accuracy and resource efficiency, making it ideal for real-time UAV deployment.

User emotion recognition and indoor space interaction design: a CNN model optimized by multimodal weighted networks

In interior interaction design, achieving intelligent user-interior interaction is contingent upon understanding the user’s emotional responses. Precise identification of the user’s visual emotions holds paramount importance. Current visual emotion recognition methods rely solely on singular features, predominantly facial expressions, resulting in inadequate coverage of visual characteristics and low recognition rates. This study introduces a deep learning-based multimodal weighting network model to address this challenge. The model initiates with a convolutional attention module, employing a self-attention mechanism within a convolutional neural network (CNN). As a result, the multimodal weighting network model is integrated to optimize weights during training. Finally, a weight network classifier is derived from these optimized weights to facilitate visual emotion recognition. Experimental outcomes reveal a 77.057% correctness rate and a 74.75% accuracy rate in visual emotion recognition. Comparative analysis against existing models demonstrates the superiority of the multimodal weight network model, showcasing its potential to enhance human-centric and intelligent indoor interaction design.

LCGSC-YOLO: a lightweight apple leaf diseases detection method based on LCNet and GSConv module under YOLO framework.

In response to the current mainstream deep learning detection methods with a large number of learned parameters and the complexity of apple leaf disease scenarios, the paper proposes a lightweight method and names it LCGSC-YOLO. This method is based on the LCNet(A Lightweight CPU Convolutional Neural Network) and GSConv(Group Shuffle Convolution) module modified YOLO(You Only Look Once) framework. Firstly, the lightweight LCNet is utilized to reconstruct the backbone network, with the purpose of reducing the number of parameters and computations of the model. Secondly, the GSConv module and the VOVGSCSP (Slim-neck by GSConv) module are introduced in the neck network, which makes it possible to minimize the number of model parameters and computations while guaranteeing the fusion capability among the different feature layers. Finally, coordinate attention is embedded in the tail of the backbone and after each VOVGSCSP module to improve the problem of detection accuracy degradation issue caused by model lightweighting. The experimental results show the LCGSC-YOLO can achieve an excellent detection performance with mean average precision of 95.5% and detection speed of 53 frames per second (FPS) on the mixed datasets of Plant Pathology 2021 (FGVC8) and AppleLeaf9. The number of parameters and Floating Point Operations (FLOPs) of the LCGSC-YOLO are much less thanother related comparative experimental algorithms.

Automatic Apple Detection and Counting with AD-YOLO and MR-SORT.

In the production management of agriculture, accurate fruit counting plays a vital role in the orchard yield estimation and appropriate production decisions. Although recent tracking-by-detection algorithms have emerged as a promising fruit-counting method, they still cannot completely avoid fruit occlusion and light variations in complex orchard environments, and it is difficult to realize automatic and accurate apple counting. In this paper, a video-based multiple-object tracking method, MR-SORT (Multiple Rematching SORT), is proposed based on the improved YOLOv8 and BoT-SORT. First, we propose the AD-YOLO model, which aims to reduce the number of incorrect detections during object tracking. In the YOLOv8s backbone network, an Omni-dimensional Dynamic Convolution (ODConv) module is used to extract local feature information and enhance the model's ability better; a Global Attention Mechanism (GAM) is introduced to improve the detection ability of a foreground object (apple) in the whole image; a Soft Spatial Pyramid Pooling Layer (SSPPL) is designed to reduce the feature information dispersion and increase the sensory field of the network. Then, the improved BoT-SORT algorithm is proposed by fusing the verification mechanism, SURF feature descriptors, and the Vector of Local Aggregate Descriptors (VLAD) algorithm, which can match apples more accurately in adjacent video frames and reduce the probability of ID switching in the tracking process. The results show that the mAP metrics of the proposed AD-YOLO model are 3.1% higher than those of the YOLOv8 model, reaching 96.4%. The improved tracking algorithm has 297 fewer ID switches, which is 35.6% less than the original algorithm. The multiple-object tracking accuracy of the improved algorithm reached 85.6%, and the average counting error was reduced to 0.07. The coefficient of determination R2 between the ground truth and the predicted value reached 0.98. The above metrics show that our method can give more accurate counting results for apples and even other types of fruit.

FCSwinU: Fourier Convolutions and Swin Transformer UNet for Hyperspectral and Multispectral Image Fusion.

The fusion of low-resolution hyperspectral images (LR-HSI) with high-resolution multispectral images (HR-MSI) provides a cost-effective approach to obtaining high-resolution hyperspectral images (HR-HSI). Existing methods primarily based on convolutional neural networks (CNNs) struggle to capture global features and do not adequately address the significant scale and spectral resolution differences between LR-HSI and HR-MSI. To tackle these challenges, our novel FCSwinU network leverages the spectral fast Fourier convolution (SFFC) module for spectral feature extraction and utilizes the Swin Transformer's self-attention mechanism for multi-scale global feature fusion. FCSwinU employs a UNet-like encoder-decoder framework to effectively merge spatiospectral features. The encoder integrates the Swin Transformer feature abstraction module (SwinTFAM) to encode pixel correlations and perform multi-scale transformations, facilitating the adaptive fusion of hyperspectral and multispectral data. The decoder then employs the Swin Transformer feature reconstruction module (SwinTFRM) to reconstruct the fused features, restoring the original image dimensions and ensuring the precise recovery of spatial and spectral details. Experimental results from three benchmark datasets and a real-world dataset robustly validate the superior performance of our method in both visual representation and quantitative assessment compared to existing fusion methods.

A lightweight and explainable model for driver abnormal behavior recognition

With the advancement of intelligent transportation systems, accurate identification of driver abnormal behavior is crucial for enhancing road safety. However, the limited computing power of vehicular systems poses a challenge for running efficient and explainable behavior recognition models. This paper proposes a lightweight and explainable driver abnormal behavior recognition model based on an improved You Only Look Once version 8 (YOLOv8). Firstly, a Spatial and Channel Reconstruction Convolution (SCConv) module is introduced to optimize the Convolution to Feature (C2f) structure, enhancing the model's feature extraction capabilities while reducing parameter redundancy. Secondly, a Spatial Pyramid Pooling with Fast Large Separable Kernel Attention (SPPF-LSKA) module is designed to better capture image context and integrate global information. Additionally, a Dynamic upsample (Dysample) module is introduced to improve the model's ability to capture subtle driver movements. Lastly, a Lightweight Shared Group Normalization Convolution Detection Head (LSGCDH) is designed to enhance the model's generalization ability, significantly reducing the model's computational load, parameter count, and size. Experimental results demonstrate that our approach has significant advantages for edge device deployment compared to mainstream algorithms. The visualization results effectively corroborate the role of each improved structure, enhancing the explainability of the abnormal behavior recognition model, which is beneficial for deployment in vehicular systems and contributes to improving road traffic safety.

DCS-YOLOv5s: A Lightweight Algorithm for Multi-Target Recognition of Potato Seed Potatoes Based on YOLOv5s

The quality inspection of potato seed tubers is pivotal for their effective segregation and a critical step in the cultivation process of potatoes. Given the dearth of research on intelligent tuber-cutting machinery in China, particularly concerning the identification of bud eyes and defect detection, this study has developed a multi-target recognition approach for potato seed tubers utilizing deep learning techniques. By refining the YOLOv5s algorithm, a novel, lightweight model termed DCS-YOLOv5s has been introduced for the simultaneous identification of tuber buds and defects. This study initiates with data augmentation of the seed tuber images obtained via the image acquisition system, employing strategies such as translation, noise injection, luminance modulation, cropping, mirroring, and the Cutout technique to amplify the dataset and fortify the model’s resilience. Subsequently, the original YOLOv5s model undergoes a series of enhancements, including the substitution of the conventional convolutional modules in the backbone network with the depth-wise separable convolution DP_Conv module to curtail the model’s parameter count and computational load; the replacement of the original C3 module’s Bottleneck with the GhostBottleneck to render the model more compact; and the integration of the SimAM attention mechanism module to augment the model’s proficiency in capturing features of potato tuber buds and defects, culminating in the DCS-YOLOv5s lightweight model. The research findings indicate that the DCS-YOLOv5s model outperforms the YOLOv5s model in detection precision and velocity, exhibiting superior detection efficacy and model compactness. The model’s detection metrics, including Precision, Recall, and mean Average Precision at Intersection over Union thresholds of 0.5 (mAP1) and 0.75 (mAP2), have improved to 95.8%, 93.2%, 97.1%, and 66.2%, respectively, signifying increments of 4.2%, 5.7%, 5.4%, and 9.8%. The detection velocity has also been augmented by 12.07%, achieving a rate of 65 FPS. The DCS-YOLOv5s target detection model, by attaining model compactness, has substantially heightened the detection precision, presenting a beneficial reference for dynamic sample target detection in the context of potato-cutting machinery.

SGW-YOLOv8n: An Improved YOLOv8n-Based Model for Apple Detection and Segmentation in Complex Orchard Environments

This study addresses the problem of detecting occluded apples in complex unstructured environments in orchards and proposes an apple detection and segmentation model based on improved YOLOv8n-SGW-YOLOv8n. The model improves apple detection and segmentation by combining the SPD-Conv convolution module, the GAM global attention mechanism, and the Wise-IoU loss function, which enhances the accuracy and robustness. The SPD-Conv module preserves fine-grained features in the image by converting spatial information into channel information, which is particularly suitable for small target detection. The GAM global attention mechanism enhances the recognition of occluded targets by strengthening the feature representation of channel and spatial dimensions. The Wise-IoU loss function further optimises the regression accuracy of the target frame. Finally, the pre-prepared dataset is used for model training and validation. The results show that the SGW-YOLOv8n model significantly improves relative to the original YOLOv8n in target detection and instance segmentation tasks, especially in occlusion scenes. The model improves the detection mAP to 75.9% and the segmentation mAP to 75.7% and maintains a processing speed of 44.37 FPS, which can meet the real-time requirements, providing effective technical support for the detection and segmentation of fruits in complex unstructured environments for fruit harvesting robots.

LostNet: A smart way for lost and find.

The rapid population growth in urban areas has led to an increased frequency of lost and unclaimed items in public spaces such as public transportation, restaurants, and other venues. Services like Find My iPhone efficiently track lost electronic devices, but many valuable items remain unmonitored, resulting in delays in reclaiming lost and found items. This research presents a method to streamline the search process by comparing images of lost and recovered items provided by owners with photos taken when items are registered as lost and found. A photo matching network is proposed, integrating the transfer learning capabilities of MobileNetV2 with the Convolutional Block Attention Module (CBAM) and utilizing perceptual hashing algorithms for their simplicity and speed. An Internet framework based on the Spring Boot system supports the development of an online lost and found image identification system. The implementation achieves a testing accuracy of 96.8%, utilizing only 0.67 GFLOPs and 3.5M training parameters, thus enabling the recognition of images in real-world scenarios and operable on standard laptops.

DA-Net: A classification-guided network for dental anomaly detection from dental and maxillofacial images

Dental abnormalities (DA) are frequent signs of disorders of the mouth that cause discomfort, infection, and loss of teeth. Early and reasonably priced treatment may be possible if defective teeth in the oral cavity are automatically detected. Several research works have endeavored to create a potent deep learning model capable of identifying DA from pictures. However, because of the following problems, aberrant teeth from the oral cavity are difficult to detect: 1) Normal teeth and crowded dentition frequently overlap; 2) The lesion area on the tooth surface is tiny. This paper proposes a professional dental anomaly detection network (DA-Net) to address such issues. First, a multi-scale dense connection module (MSDC) is designed to distinguish crowded teeth from normal teeth by learning multi-scale spatial information of dentition. Then, a pixel differential convolution (PDC) module is designed to perform pathological tooth recognition by extracting small lesion features. Finally, a multi-stage convolutional attention module (MSCA) is developed to integrate spatial information and channel information to obtain abnormal teeth in small areas. Experiments on benchmarks show that DA-Net performs well in dental anomaly detection and can further assist doctors in making treatment plans. Specifically, the DA-Net method performs best on multiple detection evaluation metrics: IoU, PRE, REC, and mAP. In terms of REC and mAP indicators, the proposed DA-Net method is 1.1% and 1.3% higher than the second-ranked YOLOv7 method.

High-Precision and Lightweight Model for Rapid Safety Helmet Detection.

This paper presents significant improvements in the accuracy and computational efficiency of safety helmet detection within industrial environments through the optimization of the you only look once version 5 small (YOLOv5s) model structure and the enhancement of its loss function. We introduce the convolutional block attention module (CBAM) to bolster the model's sensitivity to key features, thereby enhancing detection accuracy. To address potential performance degradation issues associated with the complete intersection over union (CIoU) loss function in the original model, we implement the modified penalty-decay intersection over union (MPDIoU) loss function to achieve more stable and precise bounding box regression. Furthermore, considering the original YOLOv5s model's large parameter count, we adopt a lightweight design using the MobileNetV3 architecture and replace the original squeeze-and-excitation (SE) attention mechanism with CBAM, significantly reducing computational complexity. These improvements reduce the model's parameters from 15.7 GFLOPs to 5.7 GFLOPs while increasing the mean average precision (mAP) from 82.34% to 91.56%, demonstrating its superior performance and potential value in practical industrial applications.

Dual-feature enhanced hybrid convolutional network for imbalanced fault diagnosis of rolling bearings

Abstract Deep learning has been extensively applied in the rolling bearing fault diagnosis domain due to its superior data analysis and feature extraction capabilities. However, in practical applications, the normal operating state occupies most of the service life of equipment, and the occurrence probability of each kind of fault is different, leading to imbalanced data that significantly degrades the performance of the neural network. In order to solve this problem, a dual-feature enhanced hybrid convolutional network (DEHCNet) is proposed. Firstly, an impulse segment enhancement module (ISEM) is constructed to enhance impulse segment features in the raw data, helping the network to learn fault features more accurately. Then, a hybrid convolutional module (HCM) is designed to fully mine discriminant fault features of minority classes from imbalanced data. In addition, a feature-enhanced combinational pooling module (FCPM) is devised to guide the network to focus more on the critical features and maximize the retention of key features in dimensionality reduction operations, thereby reducing the influence of data imbalance on the classifier. Finally, two distinct datasets are used to verify DEHCNet. Experimental results show that this network has better diagnostic accuracy and robustness under conditions of data imbalance.

Lightweight concrete crack detection based on spiking neural networks

PurposeMost existing methods for concrete crack detection are based on deep learning techniques such as convolutional neural networks. However, these models, due to their large memory footprint, high power consumption and insufficient feature extraction capabilities, face challenges in mobile applications. To address these issues, this paper proposes a lightweight spiking neural network detection model.Design/methodology/approachThis model achieves fast and accurate crack detection. Firstly, the Gabor-Spiking (GS) module preprocesses input images, extracting texture features and edge features of crack images through Gabor filter convolution modules and spiking convolution modules, respectively. Next, the multiscale residual (MR) module is designed, composed of convolutional layers and residual modules of various scales, to process the fused features and perform crack detection.FindingsExperimental results demonstrate that the model’s size can be reduced to 4.6 MB, achieving accuracy improvements to 87.3 and 96.4% on the SDNET and OCD datasets, respectively.Originality/valueThis paper proposes a lightweight spiking neural network detection model based on the GS module for edge texture feature fusion and the MR module for crack detection.

Research on tool wear and breakage state recognition of heavy milling 508III steel based on ResNet-CBAM

Milling water chamber head material 508 III steel belongs to extreme manufacturing, and the tool failure is very serious in the process of machining. How to monitor and identify the tool wear damage state in time and effectively is a key problem to be solved urgently. Identifying chip morphology changes under machining conditions plays a very important role in characterizing tool wear and breakage. Tool wear has an important influence on the chip morphology during the process of milling 508III steel, and changes in chip morphology can also reflect the tool wear and breakage state. Therefore, this paper takes the chip morphology image as an important feature for tool wear and breakage recognition, and uses Gaussian fuzzy estimation morphology method to preprocess the chip morphology image. Build a ResNet network combined by convolutional block attention module (ResNet-CBAM) tool wear and breakage recognition model, and explore the selection of the model backbone network, determination of the ResNet backbone network depth, and fusion methods of different attention mechanisms. Through the ablation experiment and visual analysis of the attention mechanism module, it is verified that the ResNet-CBAM model proposed in this paper has an accuracy rate of 96.67% in tool wear and breakage state recognition, especially in the stage of severe tool wear and breakage. This study realizes the prediction and early warning of tool life, and provides effective guarantee for the efficient cutting of large parts of high-end equipment and the stable operation of manufacturing system.

A Small-Scale Object Detection Algorithm in Intelligent Transportation Scenarios

In response to the problem of poor detection ability of object detection models for small-scale targets in intelligent transportation scenarios, a fusion method is proposed to enhance the features of small-scale targets, starting from feature utilization and fusion methods. The algorithm is based on the YOLOv4 tiny framework and enhances the utilization of shallow and mid-level features on the basis of Feature Pyramid Network (FPN), improving the detection accuracy of small and medium-sized targets. In view of the problem that the background of the intelligent traffic scene image is cluttered, and there is more redundant information, the Convolutional Block Attention Module (CBAM) is used to improve the attention of the model to the traffic target. To address the problem of data imbalance and prior bounding box adaptation in custom traffic data sets that expand traffic images in COCO and VOC, we propose a Copy-Paste method with an improved generation method and a K-means algorithm with improved distance measurement to enhance the model’s detection ability for corresponding categories. Comparative experiments were conducted on a customized 260-thousand traffic data set containing public traffic images, and the results showed that compared to YOLOv4 tiny, the proposed algorithm improved mAP by 4.9% while still ensuring the real-time performance of the model.

DCS-YOLO: Defect detection model for new energy vehicle battery current collector.

The future trend in global automobile development is electrification, and the current collector is an essential component of the battery in new energy vehicles. Aiming at the misjudgment and omission caused by the confusing distribution, a wide range of sizes and types, and ambiguity of target defects in current collectors, an improved target detection model DCS-YOLO (DC-SoftCBAM YOLO) based on YOLOv5 is proposed. Firstly, the detection rate of defects with different scales is improved by adding detection layers; Secondly, we use the designed DC module as the backbone network to help the model capture the global information and semantic dependencies of the target, and effectively improve the generalization ability and detection performance of the model. Finally, in the neck part, we integrate our designed Convolutional Block Attention Module (SoftPool Convolutional Block Attention Module, SoftCBAM), which can adaptively learn the importance of channels, enhance feature representation, and enable the model to better deal with target details. Experimental results show that the mAP50 of the proposed DCS-YOLO model is 92.2%, which is 5.1% higher than the baseline model. The FPS reaches 147.1, and the detection accuracy of various defect categories is improved, especially Severely bad and No cover, and the detection recall rate reaches 100%. This method has high target detection model efficiency and meets the requirements of real-time detection of battery collector defects.

Research on Deep Learning Detection Model for Pedestrian Objects in Complex Scenes Based on Improved YOLOv7.

Pedestrian detection is very important for the environment perception and safety action of intelligent robots and autonomous driving, and is the key to ensuring the safe action of intelligent robots and auto assisted driving. In response to the characteristics of pedestrian objects occupying a small image area, diverse poses, complex scenes and severe occlusion, this paper proposes an improved pedestrian object detection method based on the YOLOv7 model, which adopts the Convolutional Block Attention Module (CBAM) attention mechanism and Deformable ConvNets v2 (DCNv2) in the two Efficient Layer Aggregation Network (ELAN) modules of the backbone feature extraction network. In addition, the detection head is replaced with a Dynamic Head (DyHead) detector head with an attention mechanism; unnecessary background information around the pedestrian object is also effectively excluded, making the model learn more concentrated feature representations. Compared with the original model, the log-average miss rate of the improved YOLOv7 model is significantly reduced in both the Citypersons dataset and the INRIA dataset. The improved YOLOv7 model proposed in this paper achieved good performance improvement in different pedestrian detection problems. The research in this paper has important reference significance for pedestrian detection in complex scenes such as small, occluded and overlapping objects.

Flowering Index Intelligent Detection of Spray Rose Cut Flowers Using an Improved YOLOv5s Model

Addressing the current reliance on manual sorting and grading of spray rose cut flowers, this paper proposed an improved YOLOv5s model for intelligent recognition and grading detection of rose color series and flowering index of spray rose cut flowers. By incorporating small-scale anchor boxes and small object feature output, the model enhanced the annotation accuracy and the detection precision for occluded rose flowers. Additionally, a convolutional block attention module attention mechanism was integrated into the original network structure to improve the model’s feature extraction capability. The WIoU loss function was employed in place of the original CIoU loss function to increase the precision of the model’s post-detection processing. Test results indicated that for two types of spray rose cut flowers, Orange Bubbles and Yellow Bubbles, the improved YOLOv5s model achieved an accuracy and recall improvement of 10.2% and 20.0%, respectively. For randomly collected images of spray rose bouquets, the model maintained a detection accuracy of 95% at a confidence threshold of 0.8.

Vehicular Mini-LED backlight display inspection based on residual global context mechanism

Mini-LED backlight has emerged as a promising technology for high performance LCDs, yet the massive detection of dead pixels and precise LEDs placement are constrained by the miniature scale of the Mini-LEDs. The high-resolution network (Hrnet) with mixed dilated convolution and dense upsampling convolution (MDC-DUC) module and a residual global context attention (RGCA) module has been proposed to detect the quality of vehicular Mini-LED backlights. The proposed model outperforms the baseline networks of Unet, Pspnet, Deeplabv3+, and Hrnet, with a mean intersection over union (Miou) of 86.91%. Furthermore, compared to the four baseline detection networks, our proposed model has a lower root-mean-square error (RMSE) when analyzing the position and defective count of Mini-LEDs in the prediction map by canny algorithm. This work incorporates deep learning to support production lines improve quality of Mini-LED backlights.Graphical abstract

A method for recognizing abnormal behaviors of personnel at petroleum stations based on GTB-ResNet

PurposeThe abnormal behaviors of staff at petroleum stations pose significant safety hazards. Addressing the challenges of high parameter counts, lengthy training periods and low recognition rates in existing 3D ResNet behavior recognition models, this paper proposes GTB-ResNet, a network designed to detect abnormal behaviors in petroleum station staff.Design/methodology/approachFirstly, to mitigate the issues of excessive parameters and computational complexity in 3D ResNet, a lightweight residual convolution module called the Ghost residual module (GhostNet) is introduced in the feature extraction network. Ghost convolution replaces standard convolution, reducing model parameters while preserving multi-scale feature extraction capabilities. Secondly, to enhance the model's focus on salient features amidst wide surveillance ranges and small target objects, the triplet attention mechanism module is integrated to facilitate spatial and channel information interaction. Lastly, to address the challenge of short time-series features leading to misjudgments in similar actions, a bidirectional gated recurrent network is added to the feature extraction backbone network. This ensures the extraction of key long time-series features, thereby improving feature extraction accuracy.FindingsThe experimental setup encompasses four behavior types: illegal phone answering, smoking, falling (abnormal) and touching the face (normal), comprising a total of 892 videos. Experimental results showcase GTB-ResNet achieving a recognition accuracy of 96.7% with a model parameter count of 4.46 M and a computational complexity of 3.898 G. This represents a 4.4% improvement over 3D ResNet, with reductions of 90.4% in parameters and 61.5% in computational complexity.Originality/valueSpecifically designed for edge devices in oil stations, the 3D ResNet network is tailored for real-time action prediction. To address the challenges posed by the large number of parameters in 3D ResNet networks and the difficulties in deployment on edge devices, a lightweight residual module based on ghost convolution is developed. Additionally, to tackle the issue of low detection accuracy of behaviors amidst the noisy environment of petroleum stations, a triple attention mechanism is introduced during feature extraction to enhance focus on salient features. Moreover, to overcome the potential for misjudgments arising from the similarity of actions, a Bi-GRU model is introduced to enhance the extraction of key long-term features.