Search

A hemi-closing-wedge and hemi-opening-wedge, inverted V-shaped high tibial osteotomy with local bone graft has been reported to be an effective surgical procedure for medial osteoarthritis of the knee. In this procedure, an inverted V-shaped osteotomy is made and a thin wedged bone block is resected from the lateral side and implanted in the medial opening space created after valgus correction. This procedure can provide sufficient valgus correction of the knee with severe varus deformity more easily than can closing-wedge high tibial osteotomy. The inverted V-shaped osteotomy does not change the posterior tibial slope, the patellar height, or the length of the lower limb at all because the center of tibial alignment correction by the inverted V-shaped osteotomy is located near the center of rotation of angulation of the lower-limb deformity. We recently modified this procedure by performing biplanar osteotomy, developing useful cutting guides, and fixing the tibia with a lateral locking compression plate. The surgical technique is described to enable the reproducible creation of the hemi-closing-wedge and hemi-opening-wedge, inverted V-shaped osteotomy with the locking plate for medial osteoarthritic knees with moderate or severe varus deformity.

In this paper, we propose a novel approach for traffic accident anticipation through (i) Adaptive Loss for Early Anticipation (AdaLEA) and (ii) a large-scale self-annotated incident database for anticipation. The proposed AdaLEA allows a model to gradually learn an earlier anticipation as training progresses. The loss function adaptively assigns penalty weights depending on how early the model can an- ticipate a traffic accident at each epoch. Additionally, we construct a Near-miss Incident DataBase for anticipation. This database contains an enormous number of traffic near- miss incident videos and annotations for detail evaluation of two tasks, risk anticipation and risk-factor anticipation. In our experimental results, we found our proposal achieved the highest scores for risk anticipation (+6.6% better on mean average precision (mAP) and 2.36 sec earlier than previous work on the average time-to-collision (ATTC)) and risk-factor anticipation (+4.3% better on mAP and 0.70 sec earlier than previous work on ATTC)., Accepted to CVPR 2018

In this paper we propose a novel approach for video-based person re-identification that exploits convolutional neural networks to learn the similarity of persons observed from video camera. We take 3-dimensional convolutional neural networks (3D CNN) to extract fine-grained spatiotemporal features from the video sequence of a person. Unlike recurrent neural networks, 3D CNN preserves the spatial patterns of the input, which works well on re-identification problem. The network maps each video sequence of a person to a Euclidean space where distances between feature embeddings directly correspond to measures of person similarity. By our improved parameter learning method called entire triplet loss, all possible triplets in the mini-batch are taken into account to update network parameters. This parameter updating method significantly improves training, enabling the embeddings to be more discriminative. Experimental results show that our model achieves new state of the art identification rate on iLIDS-VID dataset and PRID-2011 dataset with 82.0%, 83.3% at rank 1, respectively.

In this paper, we propose a method to estimate 6D pose (3D position and pose) of the everyday objects, especially tools, even if there are not the same 3D model of the target object. We assume that the tools that belong to same category have common spatial relation in part-affordances, the role of each part. The proposed method uses this assumption for pose estimation. By minimizing two types of consistencies, the spatial relationship of part-affordances and the shapes, we have confirmed that the proposed method achieved higher performance compared with traditional pose estimation method.

The demand for sports video analysis is expanding. Sports video analysis is used to analyze their own play and play of opponent players, and visualize movement and ability of players. Scientific and objective analysis become possible by incorporating video analysis. It is desired to improve the competition level by feeding back the analysis results to the players. Therefore, in this research, in the case of feeding back the analysis result to the athlete, research purpose is to realize a method which can understand and evaluate the details of the form hitting the ball in detail. First, joint position coordinates are estimated from input RGB tennis images by a posture estimation method. The joint position coordinates in each frame at the time of shot are classified using unsupervised method and represented by BoW. The feature vector is designed by combining this with the shot position. The probability of shot success is predicted using this feature vector. By visualizing BoW of the shot with the high probability of success and the high failure probability, it is possible to extract and compare poses that are likely to appear in each case without giving correct labels.

Google earth with high-resolution imagery basically takes months to process new images before online updates. It is a time consuming and slow process especially for post-disaster application. The objective of this research is to develop a fast and effective method of updating maps by detecting local differences occurred over different time series; where only region with differences will be updated. In our system, aerial images from Massachusetts’s road and building open datasets, Saitama district datasets are used as input images. Semantic segmentation is then applied to input images. Semantic segmentation is a pixel-wise classification of images by implementing deep neural network technique. Deep neural network technique is implemented due to being not only efficient in learning highly discriminative image features such as road, buildings etc., but also partially robust to incomplete and poorly registered target maps. Then, aerial images which contain semantic information are stored as database in 5D world map is set as ground truth images. This system is developed to visualise multimedia data in 5 dimensions; 3 dimensions as spatial dimensions, 1 dimension as temporal dimension, and 1 dimension as degenerated dimensions of semantic and colour combination dimension. Next, ground truth images chosen from database in 5D world map and a new aerial image with same spatial information but different time series are compared via difference extraction method. The map will only update where local changes had occurred. Hence, map updating will be cheaper, faster and more effective especially post-disaster application, by leaving unchanged region and only update changed region.

In recent years, analysis of remote sensing imagery is imperatives in the domain of environmental and climate monitoring primarily for the application of detecting and managing a natural disaster. Satellite imagery or aerial imagery is beneficial because it can widely capture the condition of the surface ground and provides a massive amount of information in a piece of satellite imagery. Since obtaining satellite imagery or aerial imagery is getting more ease in recent years, landslide detection and flood detection is highly in demand. In this paper, we propose automatic natural disaster detection particularly for landslide and flood detection by implementing convolutional neural network (CNN) in extracting the feature of disaster more effectively. CNN is robust to shadow, able to obtain the characteristic of disaster adequately and most importantly able to overcome misdetection or misjudgment by operators, which will affect the effectiveness of disaster relief. The neural network consists of 2 phases: training phase and testing phase. We created training data patches of pre-disaster and post-disaster by clipping and resizing aerial imagery obtained from Google Earth Aerial Imagery. We are currently focusing on two countries which are Japan and Thailand. Training dataset for both landslide and flood consist of 50000 patches. All patches are trained in CNN to extract region where changes occurred or known as disaster region occurred without delay. We obtained accuracy of our system in around 80%–90% of both disaster detections. Based on the promising results, the proposed method may assist in our understanding of the role of deep learning in disaster detection.

The mechanism behind collisions between vehicles and pedestrians must be thoroughly studied in order to prevent future traffic accidents. In particular, preventing collisions where pedestrian steps out onto the road from behind an obstruction such as buildings, walls or vehicles is a challenging problem. To tackle this problem, we propose situation dependent topic model (SDTM), a regression model that predicts dangerous vehicle-pedestrian encounter in response to different driving situations, which also provides a framework to analyze and understand the underlying factors that lead to dangerous situations. Complex nature of situations where collisions with pedestrians happen can be expressed well by defining how dangerous situations arise differently for each driving situation pattern retrieved using statistical topic modeling. In experiments, we compare the performance of SDTM with orthodox logistic regression models using vehicle-pedestrian encounters in near-miss incidents. We also show the result of acquired knowledge that can form the basis of many other researches concerning pedestrian safety.

Recently, a demand for video analysis on vehicle-mounted driving recorders has been increasing in vision-based safety systems, such as for autonomous vehicles. The technology must be positioned one of the most important task, however, the conventional traffic datasets (e.g. KITTI, Caltech Pedestrian) are not included any dangerous scenes (near-miss scenes), even though the objective of a safety system is to avoid danger. In this paper, (i) we create a pedestrian near-miss dataset on vehicle-mounted driving recorders and (ii) propose a method to jointly learns to predict pedestrian detection and its danger level {high, low, no-danger} with convolutional neural networks (CNN) based on the ResNets. According to the result, we demonstrate the effectiveness of our approach that achieved 68% accuracy of joint pedestrian detection and danger label prediction, and 58.6fps processing time on the self-collected pedestrian near-miss dataset.

The percentage of pedestrian deaths in traffic accidents is on the rise in Japan. In recent years, there have been calls for measures to be introduced to protect vulnerable road users such as pedestrians and cyclists. In this study, a method to detect and track pedestrians using an in-vehicle camera is presented. We improve the technology of detecting pedestrians by using the highly accurate images obtained with a monocular camera. In the detection step, we employ ECoHOG as the feature descriptor; it accumulates the integrated gradient intensities. In the tracking step, we apply an effective motion model using optical flow and the proposed feature descriptor ECoHOG in a tracking-by-detection framework. These techniques were verified using images captured on real roads. key words: Pedestrian Active Safety, Tracking-by-detection, ECoHOG, Particle Filter, Vehicle Motion Model

SUMMARY This paper describes a method for accurately interpolating a low-resolution depth image using a high-resolution color image. In our method, first, tangent planes on each superpixel are estimated from sparse depth information and dense color information. Then, the neighboring superpixels that have smooth-connectable tangent planes are connected, and image segmentation to smooth surfaces is achieved. Finally, the low-resolution depth image is interpolated by using this smooth surface segmentation. In experiments with images from the Middlebury stereo data sets, our method interpolates each image at a high rate, and achieves the lowest error when compared to existing techniques.

The purpose of this study was to examine the effect of changing toe direction on knee kinetics and kinematics associated with anterior cruciate ligament injury during drop vertical jumps. Fourteen females performed drop vertical jumps under three toe conditions (natural, toe-in, and toe-out). The knee kinetics and kinematics during landing were evaluated using a motion analysis system. Results under three toe conditions were compared using a one-way repeated measures analysis of variance and a post hoc Bonferroni test. Toe-in landing was associated with a significantly greater knee abduction angle, tibial internal rotation angle, and knee abduction moment than the natural and toe-out conditions. Toe-out landing was associated with significantly greater tibial internal rotational angular velocity. Changing toe direction significantly affects knee kinetics and kinematics during landing. It is important to avoid changing toe direction excessively inward or outward during landing to prevent the increases in knee abduction and tibial internal rotation which might increase the risk of ACL injury. Prognosis, Level IV.

It has generally been assumed that metals usually fail as a result of microvoid nucleation induced by particle fracture. Here, we concentrate on high-density micropores filled with hydrogen in aluminum, existence of which has been largely overlooked until quite recently. These micropores exhibit premature growth under external loading, thereby inducing ductile fracture, whereas the particle fracture mechanism operates only incidentally. Conclusive evidence of a micropore mechanism is provided by the observation of an instantaneous release of gas at failure. We can therefore conclude that the growth of micropores dominates ductile fracture. Since the material we used has a standard pore density, we can assume that an identical fracture mechanism operates in other aluminum alloys. This finding suggests that intense heat treatment, which is generally believed to enhance the mechanical properties through homogenization, may have entirely the opposite effect. This revelation will have a major impact on the engineering design of metals.

Utilization of synchrotron X-ray microtomography has enabled the virtual reconstruction of invisible microstructural units and their deformation during the compaction, compression and recovery of flexible graphite. Particles artificially affixed, in advance, to the surface of a given microstructural unit were connected by plane triangles to create a 3D mesh in the reconstruction. High density 3D mapping of strain and displacement has also been achieved, by tracing the physical displacement of the surface particles. Micro-mechanisms of compression and recovery behavior in flexible graphite have been examined involving the direct observations and measurements. The microstructural unit was found to be composed of thin expanded graphite discs with slightly misaligned basal planes of graphite. Highly localized deformation of a thin expanded graphite disc was observed during the compression and recovery processes, suggesting strong interaction with surrounding expanded graphite discs. The macroscopic compression behavior of flexible graphite is attributable to a combination of bending and thickness reduction/recovery of the discs. It has, however, been shown that only the thickness recovery of the individual discs, due to the presence of internal entrapped air, accounts for the macroscopic recovery behavior of flexible graphite.

Cheyne-Stokes respiration is especially prevalent in preterm newborns, but its severity may not be recognized. It is characterized by apnea and cyclical weakening and strengthening of the breathing. We developed a method for detecting apnea and this abnormal respiration and for estimating its malignancy. Apnea was detected based on a “difference” feature (calculated from wavelet coefficients) and a modified maximum displacement feature (related to the respiratory waveform shape). The waveform is calculated from vertical motion of the thoracic and abdominal region during respiration using a vision sensor. Our proposed detection method effectively detects apnea (sensitivity 88.4%, specificity 99.7%).

It has recently been revealed that high-density pre-existing hydrogen micropores, formed during production processes, exhibit premature growth and coalescence under external loading at room temperature, thereby inducing ductile fracture. This process is incidentally supplemented by the well-established ductile fracture mechanism based on particle damage. It is reasonable to assume that the pre-existing hydrogen micropores may also contribute to damage evolution at high temperatures. In the present study, synchrotron X-ray microtomography was applied to the in situ observation of deformation and fracture in Al–Mg alloys at a high temperature. High-density hydrogen micropores were observed in the alloys. Flow localization controlled deformation through the mechanism of solute drag creep. A combined effect of grain boundary sliding and heterogeneous nucleation on particles was also confirmed to accelerate the growth of pre-existing hydrogen micropores and cavities. Although continuous nucleation occurred together with the growth of pre-existing hydrogen micropores, the effects of the pre-existing hydrogen micropores, especially those located on grain boundaries, were predominant in the overall damage evolution. It seemed likely that supersaturated hydrogen in the aluminum alloys might also make an appreciable contribution to cavitation during high-temperature loading.

We propose the DeepPose-based pose estimation system that is flexible with the change of bounding-box range for top-view images. Our purpose is to link person detection system and pose estimation system. We introduce Bounding-box Curriculum Learning (BCL) and Recurrent Pose Estimation (RPE). BCL is a learning technique of CNN inspired from Curriculum Learning. RPE is a recurrent process of pose estimation that fixes the bounding-box range in response to the estimated results. We show the effect of proposed methods compared to normal learned CNN-based pose estimator on our original top-view dataset.

Analysis of satellite images plays an increasingly vital role in environment and climate monitoring, especially in detecting and managing natural disaster. In this paper, we proposed an automatic disaster detection system by implementing one of the advance deep learning techniques, convolutional neural network (CNN), to analysis satellite images. The neural network consists of 3 convolutional layers, followed by max-pooling layers after each convolutional layer, and 2 fully connected layers. We created our own disaster detection training data patches, which is currently focusing on 2 main disasters in Japan and Thailand: landslide and flood. Each disaster's training data set consists of 30000∼40000 patches and all patches are trained automatically in CNN to extract region where disaster occurred instantaneously. The results reveal accuracy of 80%∼90% for both disaster detection. The results presented here may facilitate improvements in detecting natural disaster efficiently by establishing automatic disaster detection system.

In this paper, we propose a framework for recognizing human activities that uses only in-topic dominant codewords and a mixture of intertopic vectors. Latent Dirichlet allocation (LDA) is used to develop approximations of human motion primitives; these are mid-level representations, and they adaptively integrate dominant vectors when classifying human activities. In LDA topic modeling, action videos (documents) are represented by a bag-of-words (input from a dictionary), and these are based on improved dense trajectories. The output topics correspond to human motion primitives, such as finger moving or subtle leg motion. We eliminate the impurities, such as missed tracking or changing light conditions, in each motion primitive. The assembled vector of motion primitives is an improved representation of the action. We demonstrate our method on four different datasets., in CVPRW16

We present a technique to address the new challenge of activity prediction in computer vision field. In activity prediction, we infer the next human activity through “classified activities” and “activity data analysis. Moreover, the prediction should be processed in real-time to avoid dangerous or anomalous activities. The combination of space–time convolutional neural networks (ST-CNN) and improved dense trajectories (iDT) are able to effectively understand human activities in image sequences. After categorizing human activities, we insert activity tags into an activity database in order to sample a distribution of human activity. A naive Bayes classifier allows us to achieve real-time activity prediction because only three elements are needed for parameter estimation. The contributions of this paper are: (i) activity prediction within a Bayesian framework and (ii) ST-CNN and iDT features for activity recognition. Moreover, human activity prediction in real-scenes is achieved with 81.0% accuracy.

Recent advances in X-ray microtomography have created the opportunity to image the interior of materials. Microstructural images that are similar to or about an order of magnitude higher in resolution than those currently obtained with light microscopy can now be obtained in three-dimensions using synchrotron radiation. Local strain mapping is readily enabled by processing these high-resolution tomographic images using either the microstructural tracking technique or the digital volume correlation technique. This article is a review of the methodology behind these techniques and discusses recent experimental research on three-dimensional (3D) strain mapping. Potential future research directions are also outlined.

In material engineering, it is widely recognized that deformation and fracture (D/F) characteristics are important because the safety of developed materials can be determined on the basis of D/F characteristics. The D/F characteristic is defined as the load required to break the material and the strain caused by applying a particular load. To observe the effect of grain boundary slip at the micron level, we have proposed a method of obtaining displacement vectors of internal structures from submicron 3D CT images. In this paper, we introduce an improved method for accurately acquiring D/F characteristics. The results of a simulation and a real test confirm the effectiveness of the improved method.

In the past, deformation/fracture (D/F) characteristics, defined as load-deformation relationships until the materials are fractured, have been analyzed and evaluated on the surface. The D/F characteristics are affected by more than 10,000 micro-scale internal structures like air bubbles (pores), dispersed particles and cracks in 1 mm3; therefore, it is required to analyze nano-scale D/F characteristics inside materials. In this paper, we propose an analysis method by obtaining displacement vectors of dispersed particles from nano-order 3D-CT images. A problem of matching over 10,000 dispersed particles between deformation is solved by a stratified matching.

The annihilation behavior of micro pores in an aluminum alloy casting during surface cold working was continuously observed using the synchrotron radiation microtomography. To analyze micro pore annihilation behavior, the displacement of artificially dispersed particles was measured and thereby local strain distributions were mapped in high-density. A peening treatment annihilated most of the micro pores with a diameter of over 10 mm in the upper part of the specimen that could initiate fatigue cracks. The annihilation behavior of the micro pores could be understood as a function of effective strain locally accumulated around them, and not as a function of local hydrostatic strain. The effective strain varied significantly, with some large pores remaining where local effective plastic strain was relatively low. The complete annihilation of large pores in the surface layer suggests that the application of sufficiently long surface cold working is effective in improving the high-cycle fatigue properties. [doi:10.2320/matertrans.M2010069]

For the analysis of soccer videos, positional information of players and ball is very essential. In order to obtain trajectories of the players, robust tracking for multiple players is highly required. In this paper, we propose a method to track multiple players in a soccer video which is captured by a single camera. In the case of capturing the video from a single view, there might be a lot of occluded situations of players. In order to overcome this problem, we propose a robust tracking method for multiple players by combining Particle Filter and Real AdaBoost Classifier. First, each target player is tracked by using Particle Filter. Next, an occluded situation is automatically detected by checking a distance between the target players. Then we resample the center of the gravities by classifier's detection after Particle Filter tracking. We show the experimental results and effectiveness of our proposed method.

Introduction Traditional cephalometric radiographs can analyze facial soft-tissues 2 dimensionally. Because they cannot provide information about the nose, lips, cheeks, and mouth, another method is needed to analyze these soft tissues. We introduce a new method for analyzing the 3-dimensional (3D) shape and size of facial soft-tissue morphology. Methods A 3D average face model was constructed based on 3D computed tomography images of Japanese male and female adult volunteers who had well-balanced faces and normal occlusions. To test the feasibility of evaluating the quantitative effects of surgery, preoperative and postoperative 3D computed tomography images of facial soft tissues of 1 man and 1 woman were superimposed on the average faces. Results This quantitative assessment provided a comprehensive evaluation of the characteristics that separate size and shape. It was possible to view the superimposed images from any desired angle on a personal computer. Conclusions This method provides easy-to-understand information for patients and appears to be useful for clinical diagnosis and pretreatment and posttreatment soft-tissue morphologic evaluations of patients with jaw deformities.

In recent years, a large number of research papers and monographs on the analysis of hedge fund returns have been published. Typically, the authors of these studies implicitly or explicitly treat monthly returns of hedge funds as independent and identically distributed observations. The Hedge Fund Index might be able to serve that role. But the returns of an individual hedge fund are not like that. They behave autoregressively depending on the time periods. This stochastic behavior should be modeled as a combined/regime switching stochastic process of two processes: i.i.d. process and autoregressive process. This paper first depicts the autoregressiveness of hedge fund returns. Then we introduce our statistical model for returns of an individual hedge fund and then, with our retrospective view, we perform several data analyses for individual hedge funds’ return data.

In recent years, the images captured by AVHRR (Advanced Very High Resolution Radiometer) on the NOAA (National Oceanic and Atmospheric Administration) series of satellites have been used very widely for environment and land cover monitoring. In order to use NOAA images, they need to be accurately transformed from the image coordinate system into map coordinate system. This paper proposes a geometric correction method that corrects the errors caused by this transformation. In this method, the errors in NOAA image are corrected in the image coordinate system before transforming into the map coordinate system. First, the elevation values, which are read from GTOPO30 database, are verified to divide data into flat and rough blocks. Next, in order to increase the number of GCPs (Ground Control Points), besides the GCPs in the database, more GCPs are generated based on the feature of the coastline. After using reference images to correct the missing lines and noise pixels in the top and bottom parts of the image, the elevation errors of the GCP templates are corrected and GCP template matching is applied to find the residual errors for the blocks that match GCP templates. Based on these blocks, the residual errors of other flat and rough blocks are calculated by affine and Radial Basis Function transform respectively. According to the residual errors, all pixels in the image are moved to their correct positions. Finally, data is transformed from image into map by bilinear interpolation. With the proposed method, the average values of the error after correction are smaller than 0.2 pixels on both latitude and longitude directions. This result proved that the proposed method is a highly accurate geometric correction method.

We introduce a new four-dimensional (4D) visualizing system of stomatognathic function that combines an individual's 3D head model, reconstructed from cephalograms, and mandibular movement data recorded with a six degrees-of-freedom jaw movement analyzer. The individual's head model was constructed by transforming a standard head model according to the 3D coordinate values calculated from the individual's cephalograms. The accuracy of the 3D reconstructed head model was determined. The error averaged 1.02%, with a range of 0.39–1.58%, and the accuracy was sufficient for clinical use. This method performs dynamic and precise simulation of mandibular movement. We believe that our system will be useful for evaluating the functional background of patients with malocclusions and maxillofacial deformities.

The local elastic and plastic strain during deformation are very complicated and different form the macroscopic strain, because most materials have inhomogeneous microstructure. In this study, local strain distribution in three dimensions has been measured using the new developed method based on image analysis in high-resolution synchrotron radiation computed tomography (SR-CT). Model and practical specimens, which made of cupper alloy and aluminum alloy, respectively, were prepared for a development procedure and application of local strain measurements. Gauging intervals of microstructural features before and after deformation gave us information of inhomogeneous local strain distribution in three dimensions. High strain was observed in a necking region appeared after tensile deformation in the model sample. A combination of non-destructive measurements by SR-CT and three-dimensional analysis revealed inhomogeneous strain distributions in practical aluminum samples.

A tracking procedure for the high-resolution X-ray computed tomography (CT) has been developed in order to measure 3-D local strain within a deforming material in high-density. A dispersion-strengthened copper alloy model sample with alumina particles, which contains micropores, was visualized by the synchrotron radiation CT. The pores observed in reconstructed CT volumes were used as tracking markers. The developed tracking method using a set of matching parameters, which classifies matched, pended and rejected markers, exhibited high ratio of success tracking. Furthermore, the ratio was improved by applying the spring model method, which is one of the particle image velocity (PIV) methods utilized in the field of the fluid mechanics, to the pended markers. The method based on the image analysis of CT imaging volumes provides us 3-D high-density strain mapping.

International audience; Measurement of interior strain distribution has been developed based on the synchrotron radiation computed tomography (SR-CT). In metal deformation, local strain distribution differs from macroscopic strain due to microstructural factors such as grains, grain boundaries, particles, pores, voids, and cracks. A model sample, which was made from a copper alloy strengthened with alumina, was prepared containing artificial pores. Tensile loading was applied to the sample step by step. High-resolution tomographic experiment was performed at the third-generation synchrotron radiation facility (SPring-8) in Japan. Gravity center position, volume and surface area in the pores, which were regarded as markers in a tracking procedure, were measured by 3-D digital image analysis in the SR-CT images. The markers before and after the deformation were provided for registration and macroscopic strain correction before the tracking procedure. The marker tracking was carried out by means of matching parameter that was described as functions of distance, volume and surface at markers. The ratio of success tracking was evaluated in order to clarify whether the tracking method developed in this study was reliable. The 3-D strain distribution was represented successfully by the tracking results. A combination of high-resolution SR-CT and tracking of microstructural features is effective to visualize interior strain distribution in materials in 3-D.

In the past, D/F characteristics, load-deformation relationships until the materials are fractured, have been analyzed on the surface. The D/F characteristics are affected by more than ten thousand micro-scale internal structures like air bubbles (pores), cracks and particles; therefore, it is required to analyze nano-scale D/F characteristics inside materials. In this paper, we propose a method that automatically obtains the corresponding relations of the particles from nano-order 3DCT images at each deformation stage. The particles are deformation-proof and may have different geometries. First of all, some big particles are considered as landmarks and matched between pre- and post-deformation. The results of landmark matching make it easy to match many remaining particles and pores.