Convolutional Network Research Articles

A method for asteroid detection using convolutional neural networks on VST images

The study of asteroids, particularly near-Earth asteroids, is key to gaining insights into our Solar System and can help prevent dangerous collisions. Beyond finding new objects, additional observations of known asteroids will improve our knowledge of their orbit. We have developed an automated pipeline to process and search for asteroid trails in images taken with OmegaCAM, the wide-field imager mounted on the VLT Survey Telescope (VST), on the European Southern Observatory's Cerro Paranal. The pipeline inputs a FITS image and outputs the position, length, and angle of all the asteroids trails detected. A convolutional neural network was trained on a set of synthetic asteroid trails, with trail lengths 5–120 pixels (1– arcsecond ) and S/Ns 3–20. Its performance was tested on synthetic trails and validated using real trails, chosen from the Solar System Object Image Search of the Canadian Astronomy Data Centre. On the synthetic trails, the pipeline achieved a completeness of 70 % for trails with length ≥15 pixels ( arcsecond ), with a precision of 82 %. On the real trails, the pipeline achieved a completeness of 65 %, with a precision of 44 %, a lower value likely due to the higher presence of contaminants and stars in the field. The pipeline was able to detect both low- and high-S/N asteroid trails. Our method shows a strong potential to make new discoveries and precoveries in VST data across the S/N range studied, especially in the fainter end, which remains largely unexplored.

Comparative study of CNN-based and conventional fault interpretation methods: a study of the deep-water Orange Basin, South Africa

Image-based deep learning methods, especially convolutional neural networks (CNNs), are gaining traction in seismic interpretation, but their application still demands manual validation. This study compares a U-Net structured CNN, called Fault-Net, with conventional edge-enhancing seismic attributes of variance and chaos, which serve as a scientific baseline. We adopted two seismic interpretation workflows: (1) conventional attributes to enhance fault features, and (2) a deep learning-based workflow for fault segmentation using CNNs. Both workflows were applied to a high-resolution 3D seismic dataset from the structurally complex deep-water Orange Basin (South Africa). While deep learning-based software packages are commercially available, it is unclear whether they are suitable for the Orange Basin and in an academic setting due to their proprietary architectures and generally closed training data. This study provides public evidence of the feasibility of automated structural interpretation in complex seismic datasets using deep learning, revealing both key benefits and limitations. Where high-quality labelled data are available, the deep learning approach is faster and tends to produce a cleaner and more accurate depiction of larger faults compared to conventional methods. The open availability of Fault-Net makes deep learning-based interpretation particularly advantageous for academic settings, offering significant time and resource efficiency while enhancing the understanding of complex subsurface structures.

A Comprehensive Review of Vision-Based Sensor Systems for Human Gait Analysis

Analysis of the human gait represents a fundamental area of investigation within the broader domains of biomechanics, clinical research, and numerous other interdisciplinary fields. The progression of visual sensor technology and machine learning algorithms has enabled substantial developments in the creation of human gait analysis systems. This paper presents a comprehensive review of the advancements and recent findings in the field of vision-based human gait analysis systems over the past five years, with a special emphasis on the role of vision sensors, machine learning algorithms, and technological innovations. The relevant papers were subjected to analysis using the PRISMA method, and 72 articles that met the criteria for this research project were identified. A detailing of the most commonly used visual sensor systems, machine learning algorithms, human gait analysis parameters, optimal camera placement, and gait parameter extraction methods is presented in the analysis. The findings of this research indicate that non-invasive depth cameras are gaining increasing popularity within this field. Furthermore, depth learning algorithms, such as convolutional neural networks (CNNs) and long short-term memory (LSTM) networks, are being employed with increasing frequency. This review seeks to establish the foundations for future innovations that will facilitate the development of more effective, versatile, and user-friendly gait analysis tools, with the potential to significantly enhance human mobility, health, and overall quality of life. This work was supported by [GOBIERNO DE ESPANA/PID2023-150967OB-I00].

A Deep-Learning Method for Remaining Useful Life Prediction of Power Machinery via Dual-Attention Mechanism

Remaining useful life (RUL) prediction is a cornerstone of Prognostic and Health Management (PHM) for power machinery, playing a crucial role in ensuring the reliability and safety of these critical systems. In recent years, deep learning techniques have shown great promise in RUL prediction, providing more reliable and accurate outcomes. However, existing models often struggle with comprehensive feature extraction, especially in capturing the complex behavior of power machinery, where non-linear degradation patterns arise under varying operational conditions. To tackle this limitation, we propose a multi-feature fusion model leveraging a dual-attention mechanism. Initially, convolutional neural networks (CNNs) and channel attention mechanisms are employed to preliminarily extract spatial features. Subsequently, a layer combining a Gate Recurrent Unit (GRU) and self-attention mechanisms is used to further extract and integrate temporal features. Finally, RUL values are predicted via regression. The effectiveness of the proposed method was validated on C-MAPSS datasets, and its superior performance in RUL prediction was demonstrated through comparative analysis with other methods.

Fast prediction of curing residual stress in polymer‐matrix composites by a deep learning method

AbstractCuring residual stress (CRS) is common in polymer‐matrix composites due to the anisotropic properties of materials. Finite element method (FEM), the most extensively used approach for curing behavior prediction, is usually complicated and time‐consuming. To achieve a fast prediction of process‐induced stresses, a convolutional neural network (CNN) is established based on the FEM. Firstly, a fully coupled methodology is built and validated through distortions of experimentally manufactured laminates. Then, it is applied to generate models with different stacking layers, and the residual stress is computed and serves as the dataset of the deep learning model. Finally, the construction and hyperparameters are determined, and the good generalization performance proves the high accuracy of the current model. Besides, the CNN method (<1 s) greatly reduces the computational time compared with the FEM (>14 min). The supervised machine learning method shows great potential in promoting efficiency in stacking sequence designing and optimization of composites.Highlights A fully coupled model was built to predict the curing behavior of composites. The numerical model was validated by the deformation of unsymmetrical composites. A FEM‐CNN method was proposed for the fast prediction of curing residual stress. Compared with FEM, this machine learning method is accurate and efficient.

Winter Wheat Yield Prediction Using Satellite Remote Sensing Data and Deep Learning Models

Accurate crop yield prediction is crucial for formulating agricultural policies, guiding agricultural management, and optimizing resource allocation. This study proposes a method for predicting yields in China’s major winter wheat-producing regions using MOD13A1 data and a deep learning model which incorporates an Improved Gray Wolf Optimization (IGWO) algorithm. By adjusting the key parameters of the Convolutional Neural Network (CNN) with IGWO, the prediction accuracy is significantly enhanced. Additionally, the study explores the potential of the Green Normalized Difference Vegetation Index (GNDVI) in yield prediction. The research utilizes data collected from March to May between 2001 and 2010, encompassing vegetation indices, environmental variables, and yield statistics. The results indicate that the IGWO-CNN model outperforms traditional machine learning approaches and standalone CNN models in terms of prediction accuracy, achieving the highest performance with an R2 of 0.7587, an RMSE of 593.6 kg/ha, an MAE of 486.5577 kg/ha, and an MAPE of 11.39%. The study finds that April is the optimal period for early yield prediction of winter wheat. This research validates the effectiveness of combining deep learning with remote sensing data in crop yield prediction, providing technical support for precision agriculture and contributing to global food security and sustainable agricultural development.

AI-driven autonomous adaptative feedback welding machine

Abstract The gas tungsten arc welding (GTAW) is a primary method for nuclear component fabrication and repair. Recent advancements in monitoring and automation technologies have made the shift toward fully automated arc welding more feasible, reducing the necessity for continuous human oversight. Two artificial intelligence-based networks were developed that utilize sensor-based feedback on a mechanized GTAW head. We present an image-based semantic segmentation convolutional neural network that identifies crucial features such as the weld pool, groove, wire, and electrode based on which geometric measurements are derived. A separate novel neural network predicts the weld bead geometry for multi-pass welds and inconsistent groove geometries. The application of both neural networks is a pre-requisite that enables the autonomous planning and execution of multi-pass welds to fill a groove.

CONF-RCNN: a conformer and faster region-based convolutional neural network model for multi-label classification of tomato leaves disease in real field environment

CONF-RCNN: a conformer and faster region-based convolutional neural network model for multi-label classification of tomato leaves disease in real field environment

An ensemble learning model for detection of pulmonary hypertension using electrocardiogram, chest X-ray, and brain natriuretic peptide

Abstract Aims Delayed diagnosis of pulmonary hypertension (PH) is a known cause of poor patient prognosis. We aimed to develop an artificial intelligence (AI) model, using ensemble learning method to detect PH using electrocardiography (ECG), chest X-ray (CXR), and brain natriuretic peptide (BNP), facilitating accurate detection and prompting further examinations. Methods and results We developed a convolutional neural network model using ECG data to predict PH, labelled by ECG from seven institutions. Logistic regression was used for the BNP prediction model. We referenced a CXR deep learning model using ResNet18. Outputs from each of the three models were integrated into a three-layer fully connected multimodal model. Ten cardiologists participated in an interpretation test, detecting PH from patients’ ECG, CXR, and BNP data both with and without the ensemble learning model. The area under the receiver operating characteristic curves of the ECG, CXR, BNP, and ensemble learning model were 0.818 [95% confidence interval (CI), 0.808–0.828], 0.823 (95% CI, 0.780–0.866), 0.724 (95% CI, 0.668–0.780), and 0.872 (95% CI, 0.829–0.915). Cardiologists’ average accuracy rates were 65.0 ± 4.7% for test without AI model and 74.0 ± 2.7% for test with AI model, a statistically significant improvement (P < 0.01). Conclusion Our ensemble learning model improved doctors’ accuracy in detecting PH from ECG, CXR, and BNP examinations. This suggests that earlier and more accurate PH diagnosis is possible, potentially improving patient prognosis.

Integration of unpaired single cell omics data by deep transfer graph convolutional network

The rapid advance of large-scale atlas-level single cell RNA sequences and single-cell chromatin accessibility data provide extraordinary avenues to broad and deep insight into complex biological mechanism. Leveraging the datasets and transfering labels from scRNA-seq to scATAC-seq will empower the exploration of single-cell omics data. However, the current label transfer methods have limited performance, largely due to the lower capable of preserving fine-grained cell populations and intrinsic or extrinsic heterogeneity between datasets. Here, we present a robust deep transfer model based graph convolutional network, scTGCN, which achieves versatile performance in preserving biological variation, while achieving integration hundreds of thousands cells in minutes with low memory consumption. We show that scTGCN is powerful to the integration of mouse atlas data and multimodal data generated from APSA-seq and CITE-seq. Thus, scTGCN shows high label transfer accuracy and effectively knowledge transfer across different modalities.

Forecasting Agricultural Trade Based on TCN-LightGBM Models: A Data-Driven Decision

Facing the increasing complexity and dynamic fluctuations of the global agricultural trade market, accurate forecasting plays a key role in supporting agricultural policy formulation, stabilising the market and optimising resource allocation. In order to increase the precision and stability of agricultural trade predictions, this research suggests a hybrid model built on a temporal convolution network (TCN) and a lightweight gradient boosting tree (LightGBM). The TCN module effectively captures the long-term dependence characteristics of time series data through dilated convolution, which improves the model’s ability to identify seasonal and periodic trends. The LightGBM module, on the other hand, makes use of the characteristics of gradient boosting decision trees and excels at efficiently handling nonlinear relationships and avoiding overfitting. Experimental results show that the TCN-LightGBM model outperforms traditional models in terms of mean square error (MSE), mean absolute error (MAE) and prediction accuracy. Specifically, compared with ARIMA, LSTM, TCN alone or LightGBM alone, TCN-LightGBM achieves a prediction accuracy of 91.3% on the test data, with MSE and MAE of 0.021 and 0.115 respectively, significantly improving prediction accuracy and stability. In addition, parameter sensitivity analysis shows that the TCN-LightGBM model maintains a highly consistent prediction trend under different parameter configurations, which verifies the robustness of the model and its practical application value. This study provides a data-driven decision support tool with high accuracy and strong stability, providing a new solution for agricultural trade forecasting and other complex time series prediction tasks.

A Review of CNN Applications in Smart Agriculture Using Multimodal Data

This review explores the applications of Convolutional Neural Networks (CNNs) in smart agriculture, highlighting recent advancements across various applications including weed detection, disease detection, crop classification, water management, and yield prediction. Based on a comprehensive analysis of more than 115 recent studies, coupled with a bibliometric study of the broader literature, this paper contextualizes the use of CNNs within Agriculture 5.0, where technological integration optimizes agricultural efficiency. Key approaches analyzed involve image classification, image segmentation, regression, and object detection methods that use diverse data types ranging from RGB and multispectral images to radar and thermal data. By processing UAV and satellite data with CNNs, real-time and large-scale crop monitoring can be achieved, supporting advanced farm management. A comparative analysis shows how CNNs perform with respect to other techniques that involve traditional machine learning and recent deep learning models in image processing, particularly when applied to high-dimensional or temporal data. Future directions point toward integrating IoT and cloud platforms for real-time data processing and leveraging large language models for regulatory insights. Potential research advancements emphasize improving increased data accessibility and hybrid modeling to meet the agricultural demands of climate variability and food security, positioning CNNs as pivotal tools in sustainable agricultural practices. A related repository that contains the reviewed articles along with their publication links is made available.

FusionNet remote a hybrid deep learning ensemble model for remote image classification in multispectral images

Remote sensing plays a vital role in various industries, including smart cities, agriculture, and environmental monitoring, by capturing multispectral images of the Earth’s surface and analyzing its features. Accurate classification of these images is essential for extracting valuable information and making informed decisions. However, traditional image classification models, such as Convolutional Neural Networks (CNNs), often struggle to fully capture the complex spectral and spatial information inherent in multispectral images. Addressing this research gap, we propose FusionNet-Remote, a novel hybrid deep learning ensemble model that integrates CNNs with Random Forests (RF) to enhance the accuracy and robustness of remote image classification. The primary objective of this study is to develop a model that combines the strengths of CNNs for spatial feature extraction with the robust classification capabilities of RFs. Through comprehensive evaluations using LANDSAT data, FusionNet-Remote demonstrates superior performance compared to existing models, achieving an outstanding training accuracy of 99.8% and a testing accuracy of 98.7%. The model also exhibits the lowest training loss (0.0001) and testing loss (0.0043), significantly outperforming traditional models such as standalone CNNs, VGG16, and Inceptionv3. These results highlight the effectiveness of the hybrid ensemble approach in overcoming the limitations of conventional methods. This research contributes to the field by introducing an innovative ensemble technique that significantly improves remote image classification accuracy and reliability. The model’s high performance underscores its potential for critical applications such as land cover mapping, vegetation analysis, and disaster monitoring. Future work will explore the integration of more advanced CNN architectures and alternative fusion techniques to further enhance the model's performance and adaptability across different remote sensing scenarios.

Convolutional neural network for determining the flow field around an airfoil and blunt-based models

The convolutional neural network is widely applied in the classification of images and medicine. Some current networks are used in aerospace engineering and show a high potential in determining aerodynamic forces and flow fields. This article constructs a convolutional neural network for predicting pressure and velocity fields around a two-dimensional aircraft wing model (airfoil model). Training data is computed using the Reynolds-averaged method, and then extracted, focusing on the flow around the wing. Input data includes geometric parameters, and airfoil inlet velocity, and output data includes pressure field and flow velocity around the airfoil. The convolutional neural network is based on improving the U-Net network model, commonly used in medical applications. The results show that the convolutional neural network accurately predicts flow around the airfoil, with an average error below 3%. Therefore, this network can be used and further developed to predict flow around the wing. The network is then applied to predict the pressure and pressure fields around a blunt-based model with different aspect ratios. The main feature of the flow can be extracted from the network. Results related to pressure distribution, velocity, and method error are presented and discussed in the study. This study also suggests improving the network and applying it to pressure and velocity fields in aerospace engineering

Enhanced Diagnosis of Skin Cancer from Dermoscopic Images Using Alignment Optimized Convolutional Neural Networks and Grey Wolf Optimization

Skin cancer (SC) is a highly serious kind of cancer that, if not addressed swiftly, might result in the patient’s demise. Early detection of this condition allows for more effective therapy and prevents disease development. Deep Learning (DL) approaches may be used as an effective and efficient tool for SC detection (SCD). Several DL-based algorithms for automated SCD have been reported. However, more efficient models are needed to improve accuracy. As a result, this paper introduces a new strategy for SCD based on Grey Wolf optimization (GWO) methodologies and CNN. The proposed methodology has four stages: preprocessing, segmentation, feature extraction, and classification. The proposed method utilizes a Convolutional Neural Network (CNN) to extract features from Regions of Interest (ROIs). CNN is employed for feature categorization, whereas the GWO approach enhances accuracy by refining edge detection and segmentation. This technique utilizes a probabilistic model to accelerate the convergence of the GWO algorithm. Employing the GWO model to optimize the structure and weight vectors of CNNs can enhance diagnostic accuracy by a minimum of 5%, based on evaluation outcomes. The application of the proposed strategy and its performance comparison with other methods indicate that the proposed method with GWO predicted SC with an average accuracy of 95.11% and without GWO an Accuracy of 92.66%, respectively, enhancing accuracy by a minimum of 2.5% when we train our model with GWO.

Integrated Machine Learning Approaches for Landslide Susceptibility Mapping Along the Pakistan–China Karakoram Highway

The effectiveness of data-driven landslide susceptibility mapping relies on data integrity and advanced geospatial analysis; however, selecting the most suitable method and identifying key regional factors remains a challenging task. To address this, this study assessed the performance of six machine learning models, including Convolutional Neural Networks (CNNs), Random Forest (RF), Categorical Boosting (CatBoost), their CNN-based hybrid models (CNN+RF and CNN+CatBoost), and a Stacking Ensemble (SE) combining CNN, RF, and CatBoost in mapping landslide susceptibility along the Karakoram Highway in northern Pakistan. Twelve geospatial factors were examined, categorized into Topography/Geomorphology, Land Cover/Vegetation, Geology, Hydrology, and Anthropogenic Influence. A detailed landslide inventory of 272 occurrences was compiled to train the models. The proposed stacking ensemble and hybrid models improve landslide susceptibility modeling, with the stacking ensemble achieving an AUC of 0.91. Hybrid modeling enhances accuracy, with CNN–RF boosting RF’s AUC from 0.85 to 0.89 and CNN–CatBoost increasing CatBoost’s AUC from 0.87 to 0.90. Chi-square (χ2) values (9.8–21.2) and p-values (<0.005) confirm statistical significance across models. This study identifies approximately 20.70% of the area as from high to very high risk, with the SE model excelling in detecting high-risk zones. Key factors influencing landslide susceptibility showed slight variations across the models, while multicollinearity among variables remained minimal. The proposed modeling approach reduces uncertainties, enhances prediction accuracy, and supports decision-makers in implementing effective landslide mitigation strategies.

Prompt Assessment of Facial Expressions to Foresee Behavioral Aptitudes in Video Interviews

This endeavor seeks to engineer a sophisticated real-time image and video processor, augmented with an advanced artificial intelligence (AI) agent, capable of prognosticating a candidate's behavioral competencies through the nuanced analysis of their facial expressions. This achievement leverages a real-time, video-recorded interview utilizing a histogram of oriented gradients and support vector machine (HOG-SVM) in conjunction with convolutional neural network (CNN) recognition. Diverging from the traditional paradigm of emotional state recognition, this avant-garde prototype system is designed to autonomously decipher a job candidate’s behaviors through their micro expressions, grounded in the behavioral ecology view of facial displays (BECV), within the framework of employment interviews. This manuscript introduces a pioneering methodology for evaluating a candidate's performance in a video interview. It elucidates a comprehensive study of sentiment analysis and eye-tracking techniques, whereby results can be synthesized on a single display to facilitate the selection of the most suitable candidate for hire. An empirical study was executed at a enterprise, wherein video recordings and competency evaluations were gathered from the organization's staff and hiring executives. The findings revealed that our proposed system exhibits superior predictive capabilities compared to traditional human-structured interviews, personality assessments, occupational interest evaluations, and assessment center. Consequently, our innovative methodology can be effectively employed as a screening tool, leveraging a personal-value-based competency framework. Key Words: Behavioural ecology view of facial displays (BECV) · Convolutional neural network (CNN) · Employment selection · Histogram of oriented gradients (HOG) · Real-time image and video processing · Support vector machine (SVM), , Eye Tracking Technique, AVI(Automatic Visual Inspection).

Prediction of FeO content in sintered ore based on ICEEMDAN and CNN-BiLSTM-AM

FeO content of sintered ore is an important reference index for measuring the performance of sintered ore. It significantly impacts the ironmaking process, iron quality, and energy consumption. Aiming at the current problem of delayed and poor accuracy of sintered ore FeO content detection results, this article proposes a hybrid network model that incorporates improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN), convolutional neural network (CNN), bidirectional long short-term memory (BiLSTM), and attention mechanism (AM) for FeO content prediction. First, the FeO content time series were decomposed using ICEEMDAN to obtain sub-layers with different frequencies. Then, the features with higher correlation with FeO content were selected by feature selection as model inputs, followed by predicting the decomposition sequences of FeO content using CNN-BiLSTM-AM for the feature-selected variables, respectively. Finally, all predicted sublayer predictions were reconstructed into the final prediction by summation. The proposed model effectively captures the essence of the sequence through the ICEEMDAN algorithm, extracts deep features from FeO content data using CNN, captures contextual information through BiLSTM, and enhances feature extraction capability with AM. The experimental results show that the collaboration of CNN, BiLSTM, and AM effectively enhances the modelling capability of the model and significantly improves the prediction accuracy. Additionally, the ICEEMDAN decomposition algorithm is employed to enhance the prediction performance further, offering advantages over other decomposition techniques. The MAE, MAPE, RMSE, RRMSE, and R² of the new ICEEMDAN-CNN-BiLSTM-AM (ICBA) model are 0.0751, 0.846%, 0.0937, 1.0500%, and 0.9646, respectively, demonstrating a significant improvement in prediction accuracy and outperforming the relevant comparison models.

Climate model downscaling in central Asia: a dynamical and a neural network approach

Abstract. High-resolution climate projections are essential for estimating future climate change impacts. Statistical and dynamical downscaling methods, or a hybrid of both, are commonly employed to generate input datasets for impact modelling. In this study, we employ COSMO-CLM (CCLM) version 6.0, a regional climate model, to explore the benefits of dynamically downscaling a general circulation model (GCM) from the Coupled Model Intercomparison Project Phase 6 (CMIP6), focusing on climate change projections for central Asia (CA). The CCLM, at 0.22° horizontal resolution, is driven by the MPI-ESM1-2-HR GCM (at 1° spatial resolution) for the historical period of 1985–2014 and the projection period of 2019–2100 under three Shared Socioeconomic Pathways (SSPs), namely the SSP1-2.6, SSP3-7.0, and SSP5-8.5 scenarios. Using the Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) gridded observation dataset as a reference, we evaluate the performance of CCLM driven by ERA-Interim reanalysis over the historical period. The added value of CCLM, compared to its driving GCM, is evident over mountainous areas in CA, which are at a higher risk of extreme precipitation events. The mean absolute error and bias of climatological precipitation (mm d−1) are reduced by 5 mm d−1 for summer and 3 mm d−1 for annual values. For winter, there was no error reduction achieved. However, the frequency of extreme precipitation values improved in the CCLM simulations. Additionally, we employ CCLM to refine future climate projections. We present high-resolution maps of heavy precipitation changes based on CCLM and compare them with the CMIP6 GCM ensemble. Our analysis indicates an increase in the intensity and frequency of heavy precipitation events over CA areas already at risk of extreme climatic events by the end of the century. The number of days with precipitation exceeding 20 mm increases by more than 90 by the end of the century, compared to the historical reference period, under the SSP3-7.0 and SSP5-8.5 scenarios. The annual 99th percentile of total precipitation increases by more than 9 mm d−1 over mountainous areas of central Asia by the end of the century, relative to the 1985–2014 reference period, under the SSP3-7.0 and SSP5-8.5 scenarios. Finally, we train a convolutional neural network (CNN) to map a GCM simulation to its dynamically downscaled CCLM counterpart. The CNN successfully emulates the GCM–CCLM model chain over large areas of CA but shows reduced skill when applied to a different GCM–CCLM model chain. The scientific community interested in downscaling CMIP6 models could use our downscaling data, and the CNN architecture offers an alternative to traditional dynamical and statistical methods.

A traffic prediction method for missing data scenarios: graph convolutional recurrent ordinary differential equation network

Traffic prediction plays an increasingly important role in intelligent transportation systems and smart cities. Both travelers and urban managers rely on accurate traffic information to make decisions about route selection and traffic management. Due to various factors, both human and natural, traffic data often contains missing values. Addressing the impact of missing data on traffic flow prediction has become a widely discussed topic in the academic community and holds significant practical importance. Existing spatiotemporal graph models typically rely on complete data, and the presence of missing values can significantly degrade prediction performance and disrupt the construction of dynamic graph structures. To address this challenge, this paper proposes a neural network architecture designed specifically for missing data scenarios—graph convolutional recurrent ordinary differential equation network (GCRNODE). GCRNODE combines recurrent networks based on ordinary differential equation (ODE) with spatiotemporal memory graph convolutional networks, enabling accurate traffic prediction and effective modeling of dynamic graph structures even in the presence of missing data. GCRNODE uses ODE to model the evolution of traffic flow and updates the hidden states of the ODE through observed data. Additionally, GCRNODE employs a data-independent spatiotemporal memory graph convolutional network to capture the dynamic spatial dependencies in missing data scenarios. The experimental results on three real-world traffic datasets demonstrate that GCRNODE outperforms baseline models in prediction performance under various missing data rates and scenarios. This indicates that the proposed method has stronger adaptability and robustness in handling missing data and modeling spatiotemporal dependencies.