Neighboring Pixels Research Articles

A Modified LBP Operator-Based Optimized Fuzzy Art Map Medical Image Retrieval System for Disease Diagnosis and Prediction.

Medical records generated in hospitals are treasures for academic research and future references. Medical Image Retrieval (MIR) Systems contribute significantly to locating the relevant records required for a particular diagnosis, analysis, and treatment. An efficient classifier and effective indexing technique are required for the storage and retrieval of medical images. In this paper, a retrieval framework is formulated by adopting a modified Local Binary Pattern feature (AvN-LBP) for indexing and an optimized Fuzzy Art Map (FAM) for classifying and searching medical images. The proposed indexing method extracts LBP considering information from neighborhood pixels and is robust to background noise. The FAM network is optimized using the Differential Evaluation (DE) algorithm (DEFAMNet) with a modified mutation operation to minimize the size of the network without compromising the classification accuracy. The performance of the proposed DEFAMNet is compared with that of other classifiers and descriptors; the classification accuracy of the proposed AvN-LBP operator with DEFAMNet is higher. The experimental results on three benchmark medical image datasets provide evidence that the proposed framework classifies the medical images faster and more efficiently with lesser computational cost.

Demonstration of Neural Networks to Reconstruct Temperatures from Simulated Fluorescent Data Toward Use in Bio-microfluidics.

Biological systems often have a narrow temperature range of operation, which require highly accurate spatially resolved temperature measurements, often near ±0.1 K. However, many temperature sensors cannot meet both accuracy and spatial distribution requirements, often because their accuracy is limited by data fitting and temperature reconstruction models. Machine learning algorithms have the potential to meet this need, but their usage in generating spatial distributions of temperature is severely lacking in the literature. This work presents the first instance of using neural networks to process fluorescent images to map the spatial distribution of temperature. Three standard network architectures were investigated using non-spatially resolved fluorescent thermometry (simply-connected feed-forward network) or during image or pixel identification (U-net and convolutional neural network, CNN). Simulated fluorescent images based on experimental data were generated based on known temperature distributions where Gaussian white noise with a standard deviation of ±0.1 K was added. The poor results from these standard networks motivated the creation of what is termed a moving CNN, with an RMSE error of ±0.23 K, where the elements of the matrix represent the neighboring pixels. Finally, the performance of this MCNN is investigated when trained and applied to three distinctive temperature distributions characteristic within microfluidic devices, where the fluorescent image is simulated at either three or five different wavelengths. The results demonstrate that having a minimum of 10 3.5 data points per temperature and the broadest range of temperatures during training provides temperature predictions nearest to the true temperatures of the images, with a minimum RMSE of ±0.15 K. When compared to traditional curve fitting techniques, this work demonstrates that greater accuracy when spatially mapping temperature from fluorescent images can be achieved when using convolutional neural networks.

A High-Precision Power Line Recognition and Location Method Based on Structured-Light Binocular Vision

To complete the wiring operation of the main transmission line stripped of its insulating skin in a live power distribution system, a structured-light binocular vision method is utilized to identify and locate the line. First, aiming at the interference of the background information, a depth threshold segmentation method is used to filter the background area. Second, a mean filtering method is proposed to filter out the mismatch noise of a binocular vision camera in an outdoor environment. The Canny algorithm is then utilized to extract the contour, the central axis of the main transmission line is fitted, and the difference in the neighborhood pixel value is used to recognize the stripping area. Finally, the spatial equation and attitude of the central axis of the fitting transmission line are obtained along with the central coordinates of the stripping area, guiding the robot to carry out the wiring.

Classification of Land Cover Remote-Sensing Images Based on Pattern Recognition

With the development of remote sensing technology, remote sensing image data plays an active role in the dynamic monitoring of global resource changes and land cover utilization. Remote sensing image land cover classification is an important application direction of remote sensing data; how to further improve the accuracy of remote sensing image land cover classification is very important for the effective application of remote sensing data. The traditional remote sensing image land cover classification is mainly to classify remote sensing data according to the spectral data of ground objects. However, due to the complex environment of remote sensing images and the dynamic changes of the environment, traditional classification methods based only on pixel spectral data are often unable to achieve. A satisfactory classification result is achieved. In addition, some researchers have also proposed to combine pixel neighborhood texture information to supplement spectral feature data. Although the traditional classification method based on spectral features solves the problem of time-consuming visual interpretation, to a certain extent, due to the limited semantic expression ability and poor generalization ability of the design features, the classification accuracy is still not very satisfactory. This paper mainly studies the classification method of land cover remote sensing image based on pattern recognition. This paper is based on the experimental results of remote sensing data in Nanjing Yuhuatai District in 2018 and 2019. The ground resolution of the data is 2.5 meters. Data is projected, corrected, and equalized. Half of the images covering 43.75 square kilometers are used as training samples, and the remaining 50 square kilometers are used for detection. In the classification results of this IndianPines data, OA only increased by nearly 10% to 86.2%, AA increased by 13%, r was 82.77%, and Kappa coefficient was 0.84. In the classification results of Salinas data, both OA and AA increase by about 5%, and the optimization effect is not obvious.

Detection of degraded forests in Guinea, West Africa, based on Sentinel-2 time series by inclusion of moisture-related spectral indices and neighbourhood effect

Unlike deforestation, forest degradation implies a change in forest structure, with no modification in land use; thus, it is subtle and less visible than deforestation. Moreover, forest degradation is a major concern, as the potential for carbon sequestration is reduced. Degradation is related to location, anthropogenic drivers, climate, and forest types, and there is no unique detection methodology to map this phenomenon on a global scale. This study proposes a procedure for detecting degraded forests using Sentinel-2 (S2) multispectral imagery in Guinea, West Africa, and demonstrates that consideration of context improves degraded forest detection. In Guinea, the primary degradation process is selective logging in the massif forest, in addition to forest fragmentation caused by land use change. Monitoring forest degradation using remote sensing methods usually requires significant overhead in terms of photo-interpretation expertise. We proposed an approach to monitoring degradation through a semi-automated classification procedure with minimum photo-interpretation costs. First, the radiometric values of the Sentinel-2 data were explored using a photo-interpreted reference map and ground observations carried out with a local forester. Many of the ground observations found that forests considered as intact in the photo-interpreted reference map were degraded. This suggests photo-interpretation can be affected when investing in a single-date image. Our results showed that radiometric moisture-related indices (canopy water content [CWC], moisture stress index [MSI]), associated with leaf area index (LAI) retrieved from S2 data, were strongly related to the degree of degradation. Second, we improved the mapping of degraded forests through photo-interpretation of S2 data. Third, classification methodologies involving LAI, MSI, and CWC were established, and the results were compared with the improved photo-interpreted map. Two types of classification were compared: standard pixel-based classification and a hybrid method of classification that took contextual information into account by including neighbour pixel values in the algorithm to consider the pixel context, similar to the photo-interpretation methods. The inclusion of neighbour pixels in the semi-automated classification process led to a substantial improvement in the results. When contextual information was included, photo-interpretation of 1.6% of the latest year's data was sufficient to reach an overall accuracy above 85% requiring no use of past reference maps. Our method strongly supports the photo-interpretation procedure for less time-consuming and expensive map production, offering a significant foundation for degraded forest delineation in regions under similar conditions of canopy opening.

Face Recognition Attendance System Method Based on Fusion of LBP and HOG

As one of the hot topics in the field of computer vision research, face recognition technology has received significant attention due to its potentiality for a wide range of applications in government as well as commercial purposes. In practical applications, although several existing face recognition methods have achieved good performances in specific scenes, they easily suffer from a sharp decline in recognition rate if affected by different conditions of light, expression, posture and occlusion. Among many factors, influences of complex illuminations on face recognition are particularly significant. To further improve the performance of the existing local binary pattern (LBP) operator, neighbourhood weighted average LBP (NWALBP) is first proposed for fully considering the strong correlations between pixel pairs in the neighbourhood, which extends the traditional LBP uni-layer neighbourhood template window to the bi-layer neighbourhood template window and calculates the weighted average of bi-layer neighbourhood pixels in each direction. Then, inspired by centre symmetric LBP (CS-LBP), centre symmetric NWALBP (CS-NWALBP) is further proposed, which can effectively reduce computation complexity by only comparing the weighted average values of the neighbourhood pixels that are symmetric about the centre pixel. Finally, by combining the merit of histogram of oriented gradient (HOG), a feature fusion algorithm named CS-NWALBP+HOG is suggested. Several experiments have eventually demonstrated that our proposed algorithms have more robust performance under complex illumination conditions if compared with many other latest algorithms.

Nonlocal means filter based on cosine similarity applied in speckle reduction of digital holography.

In the field of digital holography, the speckle caused by coherent light greatly disturbs the quality of the reconstruction. This paper presents an innovative method to efficiently reduce speckle noise with a nonlocal means filter based on cosine similarity that determines the weight of each traversal pixel to the target pixel by comparing the similarity between the target pixel neighborhood and the traversal pixel neighborhood. Experimental results with qualitative and quantitative analysis indicate that the proposed method significantly improves noise reduction performance while preserving the details of the original image. Compared with other general image-processing methods, this well-directed method is more in line with the characteristics of holographic speckle noise and has obvious advantages in various metrics.

Local neighborhood gradient pattern: A feature descriptor for content based image retrieval

Local feature descriptors are efficient encoders for capturing repeated local patterns in many of the computer vision applications. Majority of such descriptors consider only limited local neighborhood pixels to encode a pattern. One of the major issues while considering more number of neighborhood pixels is that it increases the dimensionality of the feature descriptor. The proposed descriptor addresses these issues by describing an effective encoding pattern with optimal feature vector length. In this paper, we have proposed Local Neighborhood Gradient Pattern (LNGP) for Content-Based Image Retrieval (CBIR) in which the relationship between a set of neighbours and the centre pixel is considered to obtain a compact 8-bit pattern in the respective pixel position. The relationship of the gradient information of immediate, next-immediate, and diagonal neighbours with the centre pixel is considered for pattern formation, and thus the local information based on pixels in three directions are captured. The experiments are conducted on benchmarked image retrieval datasets such as Wang’s 1K, Corel 5K, Corel 10K, Salzburg (Stex), MIT-Vistex, AT & T, and FEI datasets and it is observed that the proposed descriptor yields average precision of 71.88%, 54.57%, 40.66%, 71.85%, 86.12%, 82.54%, and 68.54% respectively in the mentioned datasets. The comparative analysis of the recent descriptors indicates that the proposed descriptor performs efficiently in CBIR applications.

An IoT Method for Telemedicine: Lossless Medical Image Compression Using Local Adaptive Blocks

Image compression gains prominence in picture archiving and communication systems for storage and transmission of data. The advancements in technology enable the usage of computer-assisted algorithms for disease diagnosis and therapeutic planning. In this research work, a prediction-based compression algorithm was proposed to compress DICOM images. The local adaptive block-based predictor compression technique utilizes neighborhood pixel gray values for prediction and the Huffman coder was employed for the encoding of predicted coefficients. The compression results were considered effective, in contrast with the JPEG lossless, Kernel Least Mean Square, Context Adaptive Block-Based Prediction, and Least Mean square Prediction approaches. The performance validation by metrics reveals the effectiveness of the proposed local adaptive block prediction-based approach. The proposed compression algorithm was implemented in Raspberry Pi B+ embedded processor as IoT application. The compressed images are transferred through the cloud to other nodes in the network, thus facilitating the teleradiology for disease diagnosis by physicians.

Multi-modality image fusion for medical assistive technology management based on hybrid domain filtering

Medical image fusion technique combines images from different imaging modalities to enhance the reliability and accuracy of the medical diagnoses that play an increasingly significant role in clinical applications. In the domain of medical image fusion, it has made huge advancement in the span of past few years. However medical image fusion stills remain a challenging task since the recent methods still suffers from colour distortion, blurring and noise in their fusion results. In order to address the shortcomings of recent methods, this paper presents a novel method to fuse multimodal medical images by employing hybridization of cross bilateral filter and edge aware filtering using pixel-based fusion rule. This method proposed a novel algorithm for computation of weights for final fusion rule. First, a cross bilateral filter (CBF) is employed to both source images which considers both geometric closeness and gray level of neighbourhood pixels by avoiding smoothing the edges and utilizes one image to find the kernel and another to filter and vice versa. Detail images are obtained after subtraction of CBF output from the corresponding source images. With the help of domain transform filter, small scale features of the detail images are retrieved from the surrounding of large-scale structures of these images. These detail images are further processed via rolling guidance filter (RGF) which eliminates small-scale structures and refine edges to preserve large-scale structures of detail images. The weights are computed while measuring the strength of details in the detail images. The computed weights are then directly multiplied with source images followed by weight normalization in order to obtain final fusion result. Qualitative and quantitative experiments are performed on the publicly available datasets, where the results demonstrate that the proposed method outperforms the state-of-the-art, in terms of both visual effect and objective evaluation metrics. Our method generates clear and clean fused images which do not suffers from noise and artifacts while preserving essential information of the multimodal images.

Detection & recognition of veiled and unveiled human face on the basis of eyes using transfer learning.

Face detection and recognition are the most substantial research areas in computer vision and transfer learning due to the inspiring nature of faces as an object. In this paper, we show that we can obtain promising results on the standard face databanks when the features are extracted merely from the eye. The contributions of this work are divided into three parts, specifically face detection, eyes detection and recognition for individual identification. The key features for face recognition, used in this study are the eyes, nostrils, and mouth. The key features for eyes recognition are center of left eye, center of right eye, midpoint of eyes and extraction of eyebrows. Extracted Local Binary Pattern Histogram (LBPH) method is used to extract the facial features of face images whose computational complexity is very low and these features contain simple pixel values. Furthermore, neighborhood pixels are calculated to extract effective facial feature to realize eyes recognition and person verification. This study is able to identify an individual on the basis of even a single eye. The algorithm finds the brighter eye from the face and then, on the basis of that eye, the person is identified and the name of person is provided. The experimental results of this study show that faces are recognized accurately and LBPH method has achieved 98.2% accuracy.

Aspect dependent-based ghost suppression for extended targets in through-the-wall radar imaging under compressive sensing framework

Several approaches have been proposed to suppress multipath ghost in through-the-wall radar imaging (TWRI). One classical approach, called Aspect Dependent (AD), exploits locations of ghosts in the images without demanding prior knowledge of the reflecting geometry. This operation strategy makes the method superior over multipath exploitation-based approaches. However, the AD method assumes a point target that emulates unreal environment. Therefore, reconstructing extended targets with this method leads to incorrect scene interpretation. This work proposes a ghost suppression method for extended targets based on the AD feature that exploits duo sub-apertures. Firstly, we evaluate the best suppression method using a performance metric called relative clutter peak. Next, the evaluated method is extended to encompass a target extent during sub-images reconstruction. Following this strategy, an effective image fusion method suitable for extended targets is proposed. The method considers pixel neighborhood to effectively recover the given extended target. Simulation results show that, when signal-to-noise ratio is 20dB, the proposed method significantly improves signal-to-clutter ratio and relative clutter peak by 8.8% and 23.8%, respectively, relative to the existing AD based methods under point target assumption.

A New LBP Variant: Corner Rhombus Shape LBP (CRSLBP).

The local binary model is a straightforward, dependable, and effective method for extracting relevant local information from images. However, because it only uses sign information in the local region, the local binary pattern (LBP) is ineffective at capturing discriminating characteristics. Furthermore, most LBP variants select a region with one specific center pixel to fill all neighborhoods. In this paper, a new variant of a LBP is proposed for texture classification, known as corner rhombus-shape LBP (CRSLBP). In the CRSLBP approach, we first use three methods to threshold the pixel’s neighbors and center to obtain four center pixels by using sign and magnitude information with respect to a chosen region of an even block. This helps determine not just the relationship between neighbors and the pixel center but also between the center and the neighbor pixels of neighborhood center pixels. We evaluated the performance of our descriptors using four challenging texture databases: Outex (TC10,TC12), Brodatz, KTH-TIPSb2, and UMD. Various extensive experiments were performed that demonstrated the effectiveness and robustness of our descriptor in comparison with the available state of the art (SOTA).

Cloud Removal with SAR-Optical Data Fusion and Graph-Based Feature Aggregation Network

In observations of Earth, the existence of clouds affects the quality and usability of optical remote sensing images in practical applications. Many cloud removal methods have been proposed to solve this issue. Among these methods, synthetic aperture radar (SAR)-based methods have more potential than others because SAR imaging is hardly affected by clouds, and can reflect ground information differences and changes. While SAR images used as auxiliary information for cloud removal may be blurred and noisy, the similar non-local information of spectral and electromagnetic features cannot be effectively utilized by traditional cloud removal methods. To overcome these weaknesses, we propose a novel cloud removal method using SAR-optical data fusion and a graph-based feature aggregation network (G-FAN). First, cloudy optical images and contemporary SAR images are concatenated and transformed into hyper-feature maps by pre-convolution. Second, the hyper-feature maps are inputted into the G-FAN to reconstruct the missing data of the cloud-covered area by aggregating the electromagnetic backscattering information of the SAR image, and the spectral information of neighborhood and non-neighborhood pixels in the optical image. Finally, post-convolution and a long skip connection are adopted to reconstruct the final predicted cloud-free images. Both the qualitative and quantitative experimental results from the simulated data and real data experiments show that our proposed method outperforms traditional deep learning methods for cloud removal.

Correlation between Blood Oxygen Level-Dependent Magnetic Resonance Imaging Images and Prognosis of Patients with Multicenter Diabetic Nephropathy on account of Artificial Intelligence Segmentation Algorithm.

This study was aimed to analyze the correlation between blood oxygen level-dependent magnetic resonance imaging (BOLD-MRI) images and prognosis of patients with diabetic nephropathy (DN) based on artificial intelligence (AI) segmentation algorithm, so as to provide references for diagnosis and treatment as well as prognosis analysis of patients DN. In this study, a kernel function-based fuzzy C-means algorithm (KFCM) model was proposed, and the FCM algorithm based on neighborhood pixel information (BCFCM) and the FCM algorithm based on efficiency improvement (EnFCM) were introduced for comparison to analyze the image segmentation effects of three algorithms. The results showed that the partition coefficient (Vpc) and partition entropy (Vpe) of the KFCM algorithm were 0.801 and 0.602, respectively, which were better than those of the traditional FCM, BCFCM, and EnFCM algorithm. At the same time, the effects of correlation between renal cortex R2∗ (RC-R2∗), renal medulla R2∗ (RM-R2∗), renal cortex D (RC-D), renal medulla D (RM-D) and renal function on the prognosis were compared. The results showed that the correlation coefficients between RC-R2∗, RM-R2∗, RC-D, RM-D and renal function were 0.57, 0.62, 0.49, and 0.38, respectively; among them, RC-R2∗ and RM-R2∗ were negatively correlated to the estimated glomerular filtration rate (eGFR), and the difference between the groups was statistically significant (P <0.05). Among the factors affecting the prognosis of DN patients, the GFR, hemoglobin (Hb), RC-R2∗, RM-R2∗, and RC-D were all related to the prognosis of DN, and the difference between groups was statistically obvious (P <0.05). It suggested that the KFCM algorithm proposed in this study showed the relatively best segmentation effect on BOLD-MRI images for DN patients; an increase in R2∗ indicated a poor prognosis, and an increase in the RC-D value indicated a better prognosis.

Intelligent framework for mineral segmentation and fluid-accessible surface area analysis in scanning electron microscopy

Imaging is powerful means of sample characterization where mineral abundances and surface areas can be quantified from mineral maps. Images are typically manually processed by domain experts, which is time-consuming, labor intensive, and subjective. Emerging techniques, such as machine learning based image processing, can potentially address these limitations and accelerate image processing but the performance of these models for accurate sample characterization and surface area analysis has not been completely evaluated. This study evaluates the potential of Random Forest and U-Net machine learning methods for mineral characterization and surface area analysis of six sandstone samples. Various input variable sets including filter extracted features, scanning electron microscopy (SEM) backscatter electron (BSE) images and SEM-energy dispersive x-ray spectroscopy images (EDS) images were considered. The evaluation was conducted by providing an intelligent framework that not only evaluates the accuracy of prediction for each pixel but also investigates the accuracy of predicted neighboring pixels. In addition, a new methodology is proposed to distinguish the more susceptible places to dissolution on the surface of a given mineral using a ranked mineral dissolution risk assessment map. The results showed both methods had an acceptable performance with the U-Net model outperforming Random Forest. Both methods showed an improved accuracy when filter extracted features were added to the dataset as input variables. The models’ performance predicting mineral abundances and accessibility agreed well with ground truth data for majority classes (e.g., quartz) compared to minority classes. Finally, the proposed methodology was shown to reliably identify the locations susceptible for dissolution indicated via proposed risk assessment maps. The intelligent segmentation and surface area analysis framework is a promising tool for accelerating the processing of SEM data and reactivity assessment of samples.

Spatial Perception Correntropy Matrix for Hyperspectral Image Classification

With the development of the hyperspectral imaging technique, hyperspectral image (HSI) classification is receiving more and more attention. However, due to high dimensionality, limited or unbalanced training samples, spectral variability, and mixing pixels, it is challenging to achieve satisfactory performance for HSI classification. In order to overcome these challenges, this paper proposes a feature extraction method called spatial perception correntropy matrix (SPCM), which makes use of spatial and spectral correlation simultaneously to improve the classification accuracy and robustness. Specifically, the dimension reduction is carried out firstly. Then, the spatial perception method is designed to select the local neighbour pixels. Thus, local spectral-spatial correlation is characterized by the correntropy matrix constructed using the selected neighbourhoods. Finally, SPCM representations are fed into the support vector machine for classification. The extensive experiments carried out on three widely used data sets have revealed that the proposed SPCM performs better than several state-of-the-art methods, especially when the training set is small.

Magnetic Resonance Features of Acquired Immune Deficiency Syndrome Involving Central Nervous System Diseases by Intelligent Fuzzy C-Means Clustering (FCM) Algorithm.

This study was aimed to explore the application of fuzzy C-means (FCM) algorithm in MR images of acquired immune deficiency syndrome (AIDS) patients. Sixty AIDS patients with central nervous disease were selected as the research object. A method of brain MR image segmentation based on FCM clustering optimization was proposed, and FCM was optimized based on the neighborhood pixel correlation of gray difference. The correlation was introduced into the objective function to obtain more accurate pixel membership and segmentation features of the image. The segmented image can retain the original image information. The proposed algorithm can clearly distinguish gray matter from white matter in images. The average time of image segmentation was 0.142 s, the longest time of level set algorithm was 2.887 s, and the running time of multithreshold algorithm was 1.708 s. FCM algorithm had the shortest running time, and the average time was significantly better than other algorithms (P < 0.05). FCM image segmentation efficiency was above 90%, and patients can clearly display the location of lesions after MRI imaging examination. In summary, FCM algorithm can effectively combine the spatial neighborhood information of the brain image, segment the BRAIN MR image, analyze the characteristics of AIDS patients from different directions, and provide effective treatment for patients.

Recognition of Skin Lesions in Dermoscopic Images using Local Binary Patterns and Multinomial Logistic Regression

Skin cancer, such as melanoma, affects more than 13 million people annually and causes more than 65,000 deaths. As with almost all types of cancer, it can be treated effectively if it's diagnosed in it's early stages. The development of computer-aided diagnostic tools is, therefore, a valuable and urgent need to suport the decision of dermatologists, minimize human error, and accelerate the time to diagnosis and the prognosis of patients. The present work proposes the automatic recognition of three skin lesions, namely Common Nevus, Atypical Nevus, and Melanoma, from dermoscopic images. First, a feature vector representing texture is extracted from the dermoscopic images using the local binary pattern operator (LBP). Then, a multinomial logistic regression (MRL) model is used to exploit the feature vector to perform the multiclass classification. The hyperparameters of the LBP operator, namely the radius and the pixel neighborhoods, were modified to provide the feature set to the MRL model that maximizes the classification performance. Moreover, to assure the generalizability of the model and the reproducibility of the results, a 10-iteration Monte Carlo cross-validation (MCCV) was employed, by dividing the dataset into 70% for training and 30% for testing, with different random seeds in each iteration. The optimal LBP-MRL classification model was finally embeddeb in a graphical user interface (GUI) to facilitate the interaction between the user and the classification system. An average accuracy of 90%, averge recall of 89.23%, 84.75% and 100%, and average precision of 89.95%, 88.12% and 100% were obtained, to identify Common Nevus, Atypical Nevus and Melanoma, respectively, in the MCCV strategy. It should be highlighted that the proposed classification system can perfectly differentiate melanoma from the other two skin lesions.

Keypoint Message Passing for Video-Based Person Re-identification

Video-based person re-identification~(re-ID) is an important technique in visual surveillance systems which aims to match video snippets of people captured by different cameras. Existing methods are mostly based on convolutional neural networks~(CNNs), whose building blocks either process local neighbor pixels at a time, or, when 3D convolutions are used to model temporal information, suffer from the misalignment problem caused by person movement. In this paper, we propose to overcome the limitations of normal convolutions with a human-oriented graph method. Specifically, features located at person joint keypoints are extracted and connected as a spatial-temporal graph. These keypoint features are then updated by message passing from their connected nodes with a graph convolutional network~(GCN). During training, the GCN can be attached to any CNN-based person re-ID model to assist representation learning on feature maps, whilst it can be dropped after training for better inference speed. Our method brings significant improvements over the CNN-based baseline model on the MARS dataset with generated person keypoints and a newly annotated dataset: PoseTrackReID. It also defines a new state-of-the-art method in terms of top-1 accuracy and mean average precision in comparison to prior works.