Recognition Speed Research Articles

Essential visual skills required for boxing: A review

Background: Boxing requires athletes to possess specific visual skills to effectively evaluate situations and react in a timely manner. Understanding these skills is crucial for the training and development of athletes.Aim: This review aimed to summarise and analyse the essential visual skills required for boxing, highlighting their importance beyond basic visual acuity.Method: Studies were obtained from various databases, including Science Direct, Google Scholar, CISTI Source, SPORTDiscus and PubMed. A comprehensive keyword search was conducted, searching Medical Subject Headings (MeSH) ‘visual skills’, ‘hand-eye coordination’, ‘peripheral vision’, ‘eye movements’, ‘boxing’, ‘boxing rules’, ‘reaction time’, ‘speed of recognition’, ‘accommodation facility’, ‘visual memory’, ‘combat sport’, ‘depth perception’, ‘motor learning’, and ‘cognitive function’.Results: A total of 35 full-text English-language articles were included in the review after the removal of duplicates and the full-text review process. The review found that the essential visual skills required for boxing are depth perception, accommodation facility, saccadic eye movements, hand-eye coordination, peripheral awareness, speed and span of recognition, reaction time, visual memory and anticipation.Conclusion: The findings suggest that training programmes for boxers should incorporate exercises to enhance these visual skills. Further research is needed to develop specific training protocols and evaluate their effectiveness.Contribution: This review provides a comprehensive overview of visual skills in boxing, offering insights into their role in athletic performance. The study contributes to the development of more effective training strategies and highlights the need for further research in sports vision.

Band Correlation‐Based Multichannel Multiscale Convolution Network for Intelligent Interference Recognition

AbstractIn recent years, with the development and extensive application of wireless communication technology, the communication system should have stronger anti‐Jamming ability. Therefore, interference recognition is particularly important as a prerequisite for anti‐interference. However, the existing traditional and intelligent interference recognition algorithms have problems such as complicated feature extraction and low recognition accuracy under low interference‐to‐noise ratio. In order to solve the above problems, this paper introduces parallel multi‐channel multi‐scale convolution to improve the speed and accuracy of network recognition. In addition, combined with frequency band correlation and long‐short‐term memory network (LSTM), an innovative wireless communication interference identification model based on frequency band correlation is proposed, which uses LSTM to detect the frequency band correlation of interference signals and improve the accuracy of interference identification under low Jamming noise ratio (JNR). Experiments prove that the model proposed in this article has faster recognition speed and better generalization. The introduction of frequency band correlation increases the recognition accuracy to more than 99% with low JNR. Therefore, the model proposed in this paper is an effective and available model in complex electromagnetic environments. © 2024 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.

Computer Vision System for Multi-Robot Construction Waste Management: Integrating Cloud and Edge Computing

Sorting is an important construction waste management tool to increase recycling rates and reduce pollution. Previous studies have used robots to improve the efficiency of construction waste recycling. However, in large construction sites, it is difficult for a single robot to accomplish the task quickly, and multiple robots working together are a better option. Most construction waste recycling robotic systems are developed based on a client-server framework, which means that all robots need to be continuously connected to their respective cloud servers. Such systems are low in robustness in complex environments and waste a lot of computational resources. Therefore, in this paper, we propose a pixel-level automatic construction waste recognition platform with high robustness and low computational resource requirements by combining multiple computer vision technologies with edge computing and cloud computing platforms. Experiments show that the computing platform proposed in this study can achieve a recognition speed of 23.3 fps and a recognition accuracy of 90.81% at the edge computing platform without the help of network and cloud servers. This is 23 times faster than the algorithm used in previous research. Meanwhile, the computing platform proposed in this study achieves 93.2% instance segmentation accuracy on the cloud server side. Notably, this system allows multiple robots to operate simultaneously at the same construction site using only a single server without compromising efficiency, which significantly reduces costs and promotes the adoption of automated construction waste recycling robots.

Soft and Integrable Multimodal Artificial Mechanoreceptors Toward Human Sensor of Skin

AbstractHerein, the development and characterization of three distinct artificial mechanoreceptor sensors meticulously engineered is reported to emulate human skin. By mimicking the morphology, structure, and response characteristics (including preferential sensitivity, adaptation profile, and frequency response) of biological mechanoreceptors, artificial Meissner, Merkel, and Ruffini sensors capable of detecting pressure, shear, and tensile deformations with high fidelity are successfully fabricated. In situ experiments, designed to mimic physiological conditions, demonstrate that the integrated sensor array, mimicking human fingertips, can accurately discriminate seven Braille characters, five distinct surface textures, a grating with ridges, and four‐step delivery stages of an object. Furthermore, a woolen glove incorporating 15 multimodal sensors are developed, which exhibits enhanced classification capabilities for eight objects of varying sizes and surface roughness. Notably, the trimodal sensor integration demonstrates superior recognition speed and precision compared to uni‐ or bimodal configurations, while also improving tactile identification intuition. This biomimetic mechanoreceptor sensor system demonstrates comprehensive and synergistic recognition of diverse stimuli and objects, potentially overcoming technological limitations in applications requiring human‐like tactile perception, such as advanced prosthetics, robotics, and immersive augmented and virtual reality interfaces.

High-resolution image processing and entity recognition algorithm based on artificial intelligence

Abstract Objective With the popularity of high-resolution devices such as high-definition, ultra-high-definition televisions, and smartphones, the demand for high-resolution images is also increasing, which puts forward higher requirements for high-resolution image processing and entity recognition technology. Method This article introduced the research progress and application of high-resolution image processing and entity recognition algorithms from the perspective of artificial intelligence (AI). First, the important role of AI in high-resolution image processing and entity recognition was introduced, and then the applications of deep learning-based algorithms in high-resolution image grayscale equalization, denoising, and deblurring were introduced. Subsequently, the application of AI-based object detection and image segmentation algorithms in entity recognition was explored, and the superiority of AI-based high-resolution image processing and entity recognition algorithms was verified through training and testing. The accuracy of the model was verified through testing experiments. Finally, a summary and outlook were made on high-resolution image processing and entity recognition algorithms based on AI. Result After experimental testing, it was found that high-resolution image processing and entity recognition based on AI had higher efficiency, and the overall image recognition ability was improved by 29.6% compared to traditional image recognition models. The recognition speed and accuracy were also improved. Conclusion High-resolution image processing and element recognition algorithms based on AI enabled observers to see the detailed information in the image more clearly, thus improving the efficiency and accuracy of image analysis. Through continuous improvement of algorithm performance, real-time application, and expansion of cross-disciplinary applications, people can look forward to the development of more advanced and powerful image processing and entity recognition technologies, which will bring huge impetus to research and application in various fields.

Action recognition in rehabilitation: combining 3D convolution and LSTM with spatiotemporal attention.

This study addresses the limitations of traditional sports rehabilitation, emphasizing the need for improved accuracy and response speed in real-time action detection and recognition in complex rehabilitation scenarios. We propose the STA-C3DL model, a deep learning framework that integrates 3D Convolutional Neural Networks (C3D), Long Short-Term Memory (LSTM) networks, and spatiotemporal attention mechanisms to capture nuanced action dynamics more precisely. Experimental results on multiple datasets, including NTU RGB + D, Smarthome Rehabilitation, UCF101, and HMDB51, show that the STA-C3DL model significantly outperforms existing methods, achieving up to 96.42% accuracy and an F1 score of 95.83% on UCF101, with robust performance across other datasets. The model demonstrates particular strength in handling real-time feedback requirements, highlighting its practical application in enhancing rehabilitation processes. This work provides a powerful, accurate tool for action recognition, advancing the application of deep learning in rehabilitation therapy and offering valuable support to therapists and researchers. Future research will focus on expanding the model's adaptability to unconventional and extreme actions, as well as its integration into a wider range of rehabilitation settings to further support individualized patient recovery.

Fast Quality Detection of Astragalus Slices Using FA-SD-YOLO

Quality inspection is a pivotal component in the intelligent sorting of Astragalus membranaceus (Huangqi), a medicinal plant of significant pharmacological importance. To improve the precision and efficiency of assessing the quality of Astragalus slices, we present the FA-SD-YOLO model, an innovative advancement over the YOLOv8n architecture. This model introduces several novel modifications to enhance feature extraction and fusion while reducing computational complexity. The FA-SD-YOLO model replaces the conventional C2f module with the C2F-F module, developed using the FasterNet architecture, and substitutes the SPPF module with the Adaptive Inverted Fusion (AIFI) module. These changes markedly enhance the model’s feature fusion capabilities. Additionally, the integration of the SD module into the detection head optimizes parameter efficiency while improving detection performance. Performance evaluation highlights the superiority of the FA-SD-YOLO model. It achieves accuracy and recall rates of 88.6% and 89.6%, outperforming the YOLOv8n model by 1.8% and 1.3%, respectively. The model’s F1 score reaches 89.1%, and the mean average precision (mAP) improves to 93.2%, reflecting increases of 1.6% and 2.4% over YOLOv8n. These enhancements are accompanied by significant reductions in model size and computational cost: the parameter count is reduced to 1.58 million (a 47.3% reduction), and the FLOPS drops to 4.6 G (a 43.2% reduction). When compared with other state-of-the-art models, including YOLOv5s, YOLOv6s, YOLOv9t, and YOLOv11n, the FA-SD-YOLO model demonstrates superior performance across key metrics such as accuracy, F1 score, mAP, and FLOPS. Notably, it achieves a remarkable recognition speed of 13.8 ms per image, underscoring its efficiency and suitability for real-time applications. The FA-SD-YOLO model represents a robust and effective solution for the quality inspection of Astragalus membranaceus slices, providing reliable technical support for intelligent sorting machinery in the processing of this important medicinal herb.

Temporal–spatial correlation and graph attention-guided network for micro-expression recognition in English learning livestreams

Micro-expressions, fleeting facial movements lasting 1/25 to 1/3 of a second, offer crucial insights into genuine emotions, particularly valuable in online education settings. The rapid growth of English learning livestreams has heightened the need for accurate, real-time micro-expression recognition to enhance learner engagement and instructional effectiveness. However, existing methods need help with the subtle nature of these expressions, especially in dynamic, low-resolution streaming environments. This paper presents TSG–MER–ELL, a novel end-to-end network for micro-expression recognition in English learning livestreams, integrating temporal–spatial correlation and graph attention mechanisms. The framework addresses the unique challenges of real-time emotion analysis in online language education, where subtle facial cues are crucial in understanding learner engagement and comprehension. The temporal–spatial correlation module employs action units with spatio-temporal graph convolution to aggregate features from diverse facial regions, while transformer encoders construct long-range correlations. The graph attention module builds upon local facial areas to guide self-attention computations, yielding precise local correlation features. These global and local features are fused for the final micro-expression classification. We introduce an adaptive loss function that balances accuracy, efficiency, and relevance to linguistic context. Extensive experiments on SMIC, CASME II, and SAMM datasets, adapted for English learning scenarios, demonstrate TSG–MER–ELL’s superior performance over ten state-of-the-art baselines. The TSG–MER–ELL framework achieves top UF1 and UAR scores across all datasets, significantly improving recognition speed and accuracy. Ablation studies and visualizations of temporal–spatial features and graph attention weights provide insights into the framework’s effectiveness in capturing subtle emotional cues. TSG–MER–ELL’s robust performance in varied online learning conditions highlights its potential to enhance engagement, personalize instruction, and improve overall outcomes in virtual English language education.

Intelligent road surface state recognition method based on multi-layer attention residual network

Abstract Data-driven road surface state recognition enhances the efficiency and accuracy of road management, contributing to increased safety and reliability in road traffic. However, traditional machine learning and deep learning-based road surface state recognition typically rely on extensive data for model training, making it challenging to adapt to complex tasks in diverse scenarios. Therefore, this paper proposes a Multi-layer Attention Residual Network (MARN)-based intelligent road surface state recognition method. First, a Residual Convolutional Neural Network (ResNet) is constructed as the backbone model of MARN to mitigate the gradient vanishing problem, allowing the network to extract deeper features. Subsequently, an adaptive multi-layer attention mechanism is introduced in each convolutional layer, enabling adaptive weighting of each feature channel in the dataset to enhance the model’s focus on different features for better feature extraction. Furthermore, a cosine annealing learning rate adjuster is designed to improve the accuracy, robustness, and convergence during the model training process. Finally, the proposed MARN is validated using an image dataset containing six different road surface states. Comparative studies are conducted on the recognition accuracy of the proposed MARN, original ResNet, Visual Geometry Group network (VGG16), and Convolutional Neural Network (CNN). The impact of different batch sizes on the convergence speed of road surface state recognition under MARN is also analyzed. Results demonstrate that MARN achieves a training set accuracy of over 95%, surpassing VGG16 and CNN with accuracies below 85%. Compared to ResNet, MARN exhibits a 1.3% higher training set accuracy and a 0.25 lower validation set loss, showcasing superior accuracy and robustness in road surface state recognition.

Continuous sign language recognition algorithm based on object detection and variable-length coding sequence

Currently, continuous sign language recognition faces challenges such as difficulty in acquiring skeletal data, long training time for Three-Dimensional convolutional neural networks, and easy occlusion and blurring of hands. To address these problems, this paper proposes a continuous sign language recognition method based on target detection and coding sequence. The algorithm uses Dual-branch Shuffle Attention Mechanis-You Only Look Once version X (DSA-YOLOX) detection network to detect the head and hands, and encodes the sign language video according to the partition to achieve the transformation from Three-Dimensional to One-Dimensional; and then uses the proposed Bi-directional Long Short-Term Memory (BiLSTM) hand coding sequence classification model with jointly weighted Fast Dynamic Time Warping (FastDTW) to extract hand coding similarity and features while reducing the number of parameters to achieve the classification and recognition of unequal-length hand coding sequences. From the results of ablation experiments and comparison experiments, all parts of the improvement perform well. The word error rate (WER) of this paper’s method is reduced by 21.26% compared to Dynamic Time Warping-Hidden Markov Model (DTW-HMM) and 11.53% compared to Long Short-Term Memory-A(LSTM-A); the Giga Floating-point Operations Per Second(GFLOPs) of the algorithm are reduced dramatically, which is about 1/13 of the Visual Alignment Constraint(VAC) model and 1/57 of the Spatial-Temporal Multi-Cue(STMC) model; and the algorithm takes better account of the speed and accuracy of sign language recognition.

Comparing visio-spatial intelligence in amateur rugby and netball players using hand-eye coordination tests

Background: Research investigating the differences in visio-spatial skills (VSS) between athletes and non-athletes, as well as variations across sports, presents conflicting findings.Aim: The objective of this study was to determine if there exist significant differences in visio-spatial intelligence (VSI) skills between rugby players and netball players, and whether such disparities are present when comparing both groups to non-athletes.Setting: All participants, including the control group (non-athletes), rugby players and netball players, were recruited from the Zululand region of South Africa’s KwaZulu-Natal province, specifically from all premier league rugby and netball teams in the area.Methods: Participants underwent an optometric assessment, followed by an evaluation of VSS using six established tests: The Hart Near Far Rock, saccadic eye movement, evasion, accumulator, flash memory and ball wall toss tests.Results: The results revealed that rugby players significantly outperformed netball players in speed of recognition, peripheral awareness and hand-eye coordination (P = 0.000). Moreover, both rugby players and netball players performed significantly better than non-athletes in five of the six tests (P = 0.000), with the exception being the visual memory test (P = 0.809).Conclusion: This discrepancy in performance suggests that certain VSS are superior in athletes compared to non-athletes, highlighting potential implications for theories of vision, test selection and the development of sport-specific VSS testing batteries. Furthermore, the use of a hand-eye coordination-specific VSS test battery effectively differentiated between different sports.Contribution: The identified pattern was not consistent across all VSS tests, indicating that further research should explore the training methods employed by both sports, as these factors may contribute to the observed differences.

Research on Target Hybrid Recognition and Localization Methods Based on an Industrial Camera and a Depth Camera in Complex Scenes

In order to improve the target visual recognition and localization accuracy of robotic arms in complex scenes with similar targets, hybrid recognition and localization methods based on an industrial camera and depth camera are proposed. First, according to the speed and accuracy requirements of target recognition and localization, YOLOv5s is introduced as the basic algorithm model for target hybrid recognition and localization. Then, in order to improve the accuracy of target recognition and coarse localization based on an industrial camera (eye-to-hand), the AFPN feature fusion module, simple and parameter-free attention module (SimAM), and soft non-maximum suppression (Soft NMS) are introduced. In order to improve the accuracy of target recognition and fine localization based on a depth camera (eye-in-hand), the SENetV2 backbone network structure, dynamic head module, deformable attention mechanism, and chain-of-thought prompted adaptive enhancer network are introduced. After that, on the basis of constructing a dual camera platform for target hybrid recognition and localization, the hand–eye calibration, collection and production of image datasets required for model training are completed. Finally, for the docking of the oil filling port, the hybrid recognition and localization experimental tests are completed in sequence. The test results show that in target recognition and coarse localization based on the industrial camera, the recognition accuracy of the designed model reaches 99%, and the average localization errors in the horizontal and vertical directions are 2.22 mm and 3.66 mm, respectively. In target recognition and fine localization based on the depth camera, the recognition accuracy of the designed model reaches 98%, and the average errors in depth, horizontal, and vertical directions are 0.12 mm, 0.28 mm, and 0.16 mm, respectively. These not only verify the effectiveness of the target hybrid recognition and localization methods based on dual cameras, but also demonstrate that they meet the high-precision recognition and localization requirements in complex scenes.

Behavior recognition algorithm based on a dual-stream residual convolutional neural network

Abstract In the process of behavior recognition, the recognition operation may be carried out in various environments such as sunny, cloudy, and night. Since traditional recognition algorithms are judged by identifying the pixels of the image, the intensity of the light will affect the image. The brightness and contrast of the display thus interfere with the recognition results. Therefore, traditional algorithms are easily affected by the lighting environment around the recognition object. To improve the accuracy and recognition rate of the behavior recognition algorithm in different lighting environments, a convolutional neural network (CNN) algorithm using a dual-stream method of time flow and spatial flow is studied here. First, we collect behavioral action data sets and preprocess the data. The core of the behavior recognition algorithm of the dual-stream residual CNN is to use the time stream and the spatial stream to fuse behavioral features and eliminate meaningless data features. After processing Perform feature selection on the data, select the acoustic wave and light-sensing features of the data, and finally, use the extracted features to classify and identify using the two-stream residual CNN and the traditional behavior recognition method. The behavior recognition algorithm based on the dual-stream residual CNN was tested on the data of four groups of people. For the behavioral feature map with a data volume of 50, the behavior recognition algorithm of the dual-stream residual CNN was effective in various environments under different lighting conditions. The recognition accuracy can reach 83.5%, which is 12.3% higher than the traditional. The behavior recognition algorithm of the dual-stream residual CNN takes 17.25 s less than the conventional recognition algorithm. It is concluded that behavior recognition based on dual-stream residual CNNs can indeed improve the recognition accuracy and recognition speed in environments with different lighting conditions than traditional behavior recognition.

YOLOX-S-TKECB: A Holstein Cow Identification Detection Algorithm

Accurate identification of individual cow identity is a prerequisite for the construction of digital farms and serves as the basis for optimized feeding, disease prevention and control, breed improvement, and product quality traceability. Currently, cow identification faces challenges such as poor recognition accuracy, large data volumes, weak model generalization ability, and low recognition speed. Therefore, this paper proposes a cow identification method based on YOLOX-S-TKECB. (1) Based on the characteristics of Holstein cows and their breeding practices, we constructed a real-time acquisition and preprocessing platform for two-dimensional Holstein cow images and built a cow identification model based on YOLOX-S-TKECB. (2) Transfer learning was introduced to improve the convergence speed and generalization ability of the cow identification model. (3) The CBAM attention mechanism module was added to enhance the model’s ability to extract features from cow torso patterns. (4) The alignment between the apriori frame and the target size was improved by optimizing the clustering algorithm and the multi-scale feature fusion method, thereby enhancing the performance of object detection at different scales. The experimental results demonstrate that, compared to the traditional YOLOX-S model, the improved model exhibits a 15.31% increase in mean average precision (mAP) and a 32-frame boost in frames per second (FPS). This validates the feasibility and effectiveness of the proposed YOLOX-S-TKECB-based cow identification algorithm, providing valuable technical support for the application of dairy cow identification in farms.

Automated production of batched unclonable micro-patterns anti-counterfeiting labels with strong robustness and rapid recognition speed

Anti-counterfeiting technologies are indeed crucial for information security and protecting product authenticity. Traditional anti-counterfeiting methods have their limitations due to their clonable nature. Exploring new technologies, particularly those based on pixel-level textures, is a promising avenue to address the clonable issue due to high encoding capacity. However, research in this field is still in its early stages. This work introduces a new fluorescent anti-counterfeiting label technology with four key characteristics: efficient laser etching, high-throughput fabrication and segmentation, robustness aided by data augmentation, and an exceptionally high recognition speed. Specifically, the etching achieves a speed of 1,200 labels/3s, the high-throughput procedures yield a rate of 2,400 labels/4 min, and the total count is about 5.2 × 104 labels. The number of labels is further augmented to about 5.2 × 106 by implementing arbitrary rotation and brightness variation to enhance the robustness in the recognition procedure. We divide these labels into 44 categories based on differences in patterns. Utilizing machine learning methods, we have achieved a total recognition (including extraction and search process) time per label averaging 421.96 ms without classification, and 40.13 ms with classification. Specifically, the search process with classification is nearly fiftieth times shorter than the non-classification method, reaching 8.52 ms in average. The overall recognition time is much faster than previous works, and achieves an accuracy of over 98.7%. This work significantly increases the practical significance of pixel-level anti-counterfeiting labels.

자율주행 차량 사고 원인 분석을 통한 안전성 제고 방안

Although the convenience of drivers is maximizing due to the full-fledged self-driving service, complex problems are still emerging in various areas, including legal, ethical, and social aspects. As a result, social awareness and acceptability of self-driving cars have not been sufficiently formed, and efforts to improve stability are essential at this point. Therefore, in this study, only records of accidents while driving autonomously were selected among the accident data of autonomous vehicles provided by California DMV and the causes of accidents were analyzed for about 400 cases. As a result, main causes were verified as low recognition rate of parked vehicles, inability to grasp changes in vehicle size due to attachment of convenience goods, insufficient defense driving system, low recognition speed in special road conditions, etc. To mitigate these problems, it is expected to increase vehicle recognition rate, reconstruct safety area range, devise additional vehicle functionalities for defensive driving, and emphasize the importance of communication between self-driving cars and smart roads. These measures aim to enhance the safety of autonomous vehicles and contribute to increasing positive social perception and trust in autonomous driving.

Home Door Security System with Face Recognition Using ESP 32 Cam

The development of science and technology, especially in the fields of Electronics and Informatics, is expected to enhance room or home security systems, particularly home door security systems. Many door security technologies have been developed to prevent and ensure the comfort of homeowners when they leave their homes. Some commonly used door security systems include CCTV cameras, fingerprint sensors, voice sensors, and RFID. However, with technological advancements, these security systems have shown vulnerabilities, such as frequent errors in fingerprint and voice sensors. Based on this, to create a more secure and innovative home door security system, the authors developed a security system with face detection and a remote warning notification system using Telegram. The researchers created a door security system using the ESP 32-CAM for face detection and as a controller for the Internet of Things (IoT) system. Additionally, a manual system using a keypad for password input was also created. The results of this study include the accuracy and speed of face recognition and the speed of sending security system warning information through the Telegram application. Keywords— Door Lock System, Face Recognition, ESP32 Camera, Microcontroller, Internet of Things.

Prompt Update Algorithm Based on the Boolean Vector Inner Product and Ant Colony Algorithm for Fast Target Type Recognition

In recent years, data mining technology has become increasingly popular, evolving into an independent discipline as research deepens. This study constructs and optimizes an association rule algorithm based on the Boolean vector (BV) inner product and ant colony optimization to enhance data mining efficiency. Frequent itemsets are extracted from the database by establishing BV and performing vector inner product operations. These frequent itemsets form the problem space for the ant colony algorithm, which generates the maximum frequent itemset. Initially, data from the total scores of players during the 2022–2024 regular season was analyzed to obtain the optimal lineup. The results obtained from the Apriori algorithm (AA) were used as a standard for comparison with the Confidence-Debiased Adversarial Fuzzy Apriori Method (CDAFAM), the AA based on deep learning (DL), and the proposed algorithm regarding their results and required time. A dataset of disease symptoms was then used to determine diseases based on symptoms, comparing accuracy and time against the original database as a standard. Finally, simulations were conducted using five batches of radar data from the observation platform to compare the time and accuracy of the four algorithms. The results indicate that both the proposed algorithm and the AA based on DL achieve approximately 10% higher accuracy compared with the traditional AA. Additionally, the proposed algorithm requires only about 25% of the time needed by the traditional AA and the AA based on DL for target recognition. Although the CDAFAM has a similar processing time to the proposed algorithm, its accuracy is lower. These findings demonstrate that the proposed algorithm significantly improves the accuracy and speed of target recognition.

Semantic Segmentation Model-Based Boundary Line Recognition Method for Wheat Harvesting

The wheat harvesting boundary line is vital reference information for the path tracking of an autonomously driving combine harvester. However, unfavorable factors, such as a complex light environment, tree shade, weeds, and wheat stubble color interference in the field, make it challenging to identify the wheat harvest boundary line accurately and quickly. Therefore, this paper proposes a harvest boundary line recognition model for wheat harvesting based on the MV3_DeepLabV3+ network framework, which can quickly and accurately complete the identification in complex environments. The model uses the lightweight MobileNetV3_Large as the backbone network and the LeakyReLU activation function to avoid the neural death problem. Depth-separable convolution is introduced into Atrous Spatial Pyramid Pooling (ASPP) to reduce the complexity of network parameters. The cubic B-spline curve-fitting method extracts the wheat harvesting boundary line. A prototype harvester for wheat harvesting boundary recognition was built, and field tests were conducted. The test results show that the wheat harvest boundary line recognition model proposed in this paper achieves a segmentation accuracy of 98.04% for unharvested wheat regions in complex environments, with an IoU of 95.02%. When the combine harvester travels at 0~1.5 m/s, the normal speed for operation, the average processing time and pixel error for a single image are 0.15 s and 7.3 pixels, respectively. This method could achieve high recognition accuracy and fast recognition speed. This paper provides a practical reference for the autonomous harvesting operation of a combine harvester.

A New Kiwi Fruit Detection Algorithm Based on an Improved Lightweight Network

To address the challenges associated with kiwi fruit detection methods, such as low average accuracy, inaccurate recognition of fruits, and long recognition time, this study proposes a novel kiwi fruit recognition method based on an improved lightweight network S-YOLOv4-tiny detection algorithm. Firstly, the YOLOv4-tiny algorithm utilizes the CSPdarknet53-tiny network as a backbone feature extraction network, replacing the CSPdarknet53 network in the YOLOv4 algorithm to enhance the speed of kiwi fruit recognition. Additionally, a squeeze-and-excitation network has been incorporated into the S-YOLOv4-tiny detection algorithm to improve accurate image extraction of kiwi fruit characteristics. Finally, enhancing dataset pictures using mosaic methods has improved precision in the characteristic recognition of kiwi fruits. The experimental results demonstrate that the recognition and positioning of kiwi fruits have yielded improved outcomes. The mean average precision (mAP) stands at 89.75%, with a detection precision of 93.96% and a single-picture detection time of 8.50 ms. Compared to the YOLOv4-tiny detection algorithm network, the network in this study exhibits a 7.07% increase in mean average precision and a 1.16% acceleration in detection time. Furthermore, an enhancement method based on the Squeeze-and-Excitation Network (SENet) is proposed, as opposed to the convolutional block attention module (CBAM) and efficient channel attention (ECA). This approach effectively addresses issues related to slow training speed and low recognition accuracy of kiwi fruit, offering valuable technical insights for efficient mechanical picking methods.