In real Internet of Things (IoT) environments, IoT devices vary widely in data types and needs. IoT devices participating in Personalized Federated Learning (PFL) all have their own unique data characteristics, but there exist some similarities. Previous work enhances Personalized Local Models (PLMs)’ performance by clustering IoT devices’ PLM while ignoring dimensionality and communication volume, resulting in lower Global Model (GM) accuracy and PLMs’ performance. To this end, we propose a Personalized Federated Learning method based on Model Head Clustering (FedMHC). Specifically, FedMHC groups IoT devices with similar data characteristics and distributes different GMs to different device groups. FedMHC allows each IoT device to obtain a GM that best fits its local data characteristics and guides PLM training. FedMHC only clusters model head parameters on the server side. Thus, the edge server only needs to transmit the head parameters and a single shared feature extractor parameters during communication with IoT devices. The improvement can effectively address the issues of dimensionality and high communication volume. Experiments on CIFAR-100, Tiny-ImageNet, and AG News datasets demonstrate that FedMHC enhances the model accuracy by 1.79% and 5.9% in pathological heterogeneous scenarios, and by 1.43%, 0.92%, and 0.94% in practical heterogeneous scenarios, compared to the topperforming methods among 9 baselines.

Semi-supervised anomaly detection is a data mining task which aims at learning features from partially-labeled datasets. We propose Deep Anomaly Detection and Search (DADS) with reinforcement learning. During the training process, the agent searches for possible anomalies in unlabeled dataset to enhance performance. Empirically, we compare DADS with several methods in the settings of leveraging known anomalies to detect both other known and unknown anomalies. Results show that DADS achieves good performance.

To handle a large amount of unlabeled data, batch active learning (BAL) queries humans for the labels of a batch of the most valuable data points at every round. Most current BAL strategies are based on human-designed heuristics, such as uncertainty sampling or mutual information maximization. However, there exists a disagreement between these heuristics and the ultimate goal of BAL, i.e., optimizing the model's final performance within the query budgets. This disagreement leads to a limited generality of these heuristics. To this end, we formulate BAL as an MDP and propose a data-driven approach based on deep reinforcement learning. Our method learns the BAL strategy by maximizing the model's final performance. Experiments on the UCI benchmark show that our method can achieve competitive performance compared to existing heuristics-based approaches.

Feature selection (FS) is a crucial procedure in machine learning pipelines for its significant benefits in removing data redundancy and mitigating model overfitting. Since concept drift is a widespread phenomenon in streaming data and could severely affect model performance, effective FS on concept drifting data streams is imminent. However, existing state-of-the-art FS algorithms fail to adjust their selection strategy adaptively when the effective feature subset changes, making them unsuitable for drifting streams. In this paper, we propose a dynamic FS method that selects effective features on concept drifting data streams via deep reinforcement learning. Specifically, we present two novel designs: (i) a skip-mode reinforcement learning environment that shrinks action space size for high-dimensional FS tasks; (ii) a curiosity mechanism that generates intrinsic rewards to address the long-horizon exploration problem. The experiment results show that our proposed method outperforms other FS methods and can dynamically adapt to concept drifts.

As deepfake technology has become increasingly sophisticated and accessible, making it easier for individuals with malicious intent to create convincing fake content, which has raised considerable concern in the multimedia and computer vision community. Despite significant advances in deepfake video detection, most existing methods mainly focused on model architecture and training processes with little focus on data perspectives. In this paper, we argue that data quality has become the main bottleneck of current research. To be specific, in the pre-training phase, the domain shift between pre-training and target datasets may lead to poor generalization ability. Meanwhile, in the training phase, the low fidelity of the existing datasets leads to detectors relying on specific low-level visual artifacts or inconsistency. To overcome the shortcomings, (1). In the pre-training phase, pre-train our model on high-quality facial videos by utilizing data-efficient reconstruction-based self-supervised learning to solve domain shift. (2). In the training phase, we develop a novel spatio-temporal generator that can synthesize various high-quality "fake" videos in large quantities at a low cost, which enables our model to learn more general spatio-temporal representations in a self-supervised manner. (3). Additinally, to take full advantage of synthetic "fake" videos, we adopt diversity losses at both frame and video levels to explore the diversity of clues in "fake" videos. Our proposed framework is data-efficient and does not require any real-world deepfake videos. Extensive experiments demonstrate that our method significantly improves the generalization capability. Particularly on the most challenging CDF and DFDC datasets, our method outperforms the baselines by 8.88% and 7.73% points, respectively.

Knowledge amalgamation (KA) is a novel deep model reusing task aiming to transfer knowledge from several well-trained teachers to a multi-talented and compact student. Currently, most of these approaches are tailored for convolutional neural networks (CNNs). However, there is a tendency that Transformers, with a completely different architecture, are starting to challenge the domination of CNNs in many computer vision tasks. Nevertheless, directly applying the previous KA methods to Transformers leads to severe performance degradation. In this work, we explore a more effective KA scheme for Transformer-based object detection models. Specifically, considering the architecture characteristics of Transformers, we propose to dissolve the KA into two aspects: sequence-level amalgamation (SA) and task-level amalgamation (TA). In particular, a hint is generated within the sequence-level amalgamation by concatenating teacher sequences instead of redundantly aggregating them to a fixed-size one as previous KA approaches. Besides, the student learns heterogeneous detection tasks through soft targets with efficiency in the task-level amalgamation. Extensive experiments on PASCAL VOC and COCO have unfolded that the sequence-level amalgamation significantly boosts the performance of students, while the previous methods impair the students. Moreover, the Transformer-based students excel in learning amalgamated knowledge, as they have mastered heterogeneous detection tasks rapidly and achieved superior or at least comparable performance to those of the teachers in their specializations.

Recent studies have made dozens of attempts to apply multi-agent deep reinforcement learning (MARL) for large-scale traffic signal control. However, most related studies have ignored how to master arterial traffic signal control. We cannot easily extract useful information and search solution space because the arterial traffic control problem has large state-action spaces. Here we tackle these issues by proposing a multi-agent attention-base soft actor-critic (MASAC) model to master arterial traffic control. Specifically, we implement the attention mechanism in the actor and critic network to enhance traffic information extraction ability. More importantly, we are the first to apply the soft actor-critic (SAC) algorithm to train the arterial traffic control model to search more solution spaces. Testing results indicate that the MASAC method significantly outperforms existing MARL algorithms and the multiband-based method. These findings can help researchers to design better model structures for other MARL problems.

Few-shot image classification aims to classify unseen classes with limited labelled samples. Recent works benefit from the meta-learning process with episodic tasks and can fast adapt to class from training to testing. Due to the limited number of samples for each task, the initial embedding network for meta-learning becomes an essential component and can largely affect the performance in practice. To this end, most of the existing methods highly rely on the efficient embedding network. Due to the limited labelled data, the scale of embedding network is constrained under a supervised learning(SL) manner which becomes a bottleneck of the few-shot learning methods. In this paper, we proposed to train a more generalized embedding network with self-supervised learning (SSL) which can provide robust representation for downstream tasks by learning from the data itself. We evaluate our work by extensive comparisons with previous baseline methods on two few-shot classification datasets (i.e., MiniImageNet and CUB) and achieve better performance over baselines. Tests on four datasets in cross-domain few-shot learning classification show that the proposed method achieves state-of-the-art results and further prove the robustness of the proposed model. Our code is available at https://github.com/phecy/SSL-FEW-SHOT.

Neural trees aim at integrating deep neural networks and decision trees so as to bring the best of the two worlds, including representation learning from the former and faster inference from the latter. In this paper, we introduce a novel approach, termed as Self-born Wiring (SeBoW), to learn neural trees from a mother deep neural network. In contrast to prior neural-tree approaches that either adopt a pre-defined structure or grow hierarchical layers in a progressive manner, task-adaptive neural trees in SeBoW evolve from a deep neural network through a construction-by-destruction process, enabling a global-level parameter optimization that further yields favorable results. Specifically, given a designated network configuration like VGG, SeBoW disconnects all the layers and derives isolated filter groups, based on which a global-level wiring process is conducted to attach a subset of filter groups, eventually bearing a lightweight neural tree. Extensive experiments demonstrate that, with a lower computational cost, SeBoW outperforms all prior neural trees by a significant margin and even achieves results on par with predominant non-tree networks like ResNets. Moreover, SeBoW proves its scalability to large-scale datasets like ImageNet, which has been barely explored by prior tree networks.

Semantic segmentation models trained on public datasets have achieved great success in recent years. However, these models didn’t consider the personalization issue of segmentation though it is important in practice. In this paper, we address the problem of personalized image segmentation. The objective is to generate more accurate segmentation results on unlabeled personalized images by investigating the data’s personalized traits. To open up future research in this area, we collect a large dataset containing various users’ personalized images called PSS (Personalized Semantic Segmentation). We also survey some recent researches related to this problem and report their performance on our dataset. Furthermore, by observing the correlation among a user’s personalized images, we propose a baseline method that incorporates the inter-image context when segmenting certain images. Extensive experiments show that our method outperforms the existing methods on the proposed dataset. The code and the PSS dataset are available at https://mmcheng.net/pss/.

Exploring the intrinsic interconnections between the knowledge encoded in PRe-trained Deep Neural Networks (PR-DNNs) of heterogeneous tasks sheds light on their mutual transferability, and consequently enables knowledge transfer from one task to another so as to reduce the training effort of the latter. In this paper, we propose the DEeP Attribution gRAph (DEPARA) to investigate the transferability of knowledge learned from PR-DNNs. In DEPARA, nodes correspond to the inputs and are represented by their vectorized attribution maps with regards to the outputs of the PR-DNN. Edges denote the relatedness between inputs and are measured by the similarity of their features extracted from the PR-DNN. The knowledge transferability of two PR-DNNs is measured by the similarity of their corresponding DEPARAs. We apply DEPARA to two important yet under-studied problems in transfer learning: pre-trained model selection and layer selection. Extensive experiments are conducted to demonstrate the effectiveness and superiority of the proposed method in solving both these problems. Code, data and models reproducing the results in this paper are available at https://github.com/zju-vipa/DEPARA.

Video classification problem is a challenging task in computer vision. The performance of this task is highly relied on the scale of training data and the effectiveness of video embedding via a robust embedding network. Unsupervised solutions such as feature average pooling technique, as a simple label-independent and parameter-free based method, cannot efficiently represent the video sequences. While supervised methods, such as RNN, can improve the recognition accuracy. The performance of RNN based methods, however, is decreased with the increasing length of the videos and the hierarchical relationships between frames across events in the video. In this paper, we propose a novel video classification method based on a deep convolutional graph neural network (DCGN). The proposed method utilizes the characteristics of the hierarchical structure of the video, and performed multi-level embedding feature extraction on the video frame sequence through the graph network, and obtained a video representation which reflects the event semantics hierarchically. Experiments on YouTube-8M Large-Scale Video Understanding dataset show that our proposed model outperforms the commonly used RNN based models, verifying its effectiveness for video classification.

Big data computing is a process to handle large volumes of information, which typically crosses different functional units in a distributed system. Like any processes involving distributed systems, it has a concern of reliability problems, such as lossy communication links between functional units and crashed computation nodes inside functional units. The paper focuses on resolving this concern in a particular distributed system scenario where the cross-boundary network connections have a high rate of failure and the internal computation nodes are relatively reliable. We propose a pure client side protocol to achieve exactly once message processing which makes big data computing in the above scenario more reliable. Moreover, we optimize the protocol to be more efficient in resource consumption using methods such as machine learning.

Web bots such as crawlers are widely used to automate various online tasks over the Internet. In addition to the conventional approach of human interactive proofs such as CAPTCHAs, a more recent approach of human observational proofs (HOP) has been developed to automatically distinguish web bots from human users. Its design rationale is that web bots behave intrinsically differently from human beings, allowing them to be detected. This paper escalates the battle against web bots by exploring the limits of current HOP-based bot detection systems. We develop an evasive web bot system based on human behavioral patterns. Then we prototype a general web bot framework and a set of flexible de-classifier plugins, primarily based on application-level event evasion. We further abstract and define a set of benchmarks for measuring our system's evasion performance on contemporary web applications, including social network sites. Our results show that the proposed evasive system can effectively mimic human behaviors and evade detectors by achieving high similarities between human users and evasive bots.

This paper presents a finding and a technique on program behavior prediction. The finding is that surprisingly strong statistical correlations exist among the behaviors of different program components (e.g., loops) and among different types of program level behaviors (e.g., loop trip-counts versus data values). Furthermore, the correlations can be beneficially exploited: They help resolve the proactivity-adaptivity dilemma faced by existing program behavior predictions, making it possible to gain the strengths of both approaches--the large scope and earliness of offline-profiling--based predictions, and the cross-input adaptivity of runtime sampling-based predictions.

As AJAX applications gain popularity, client-side JavaScript code is becoming increasingly complex. However, few automated vulnerability analysis tools for JavaScript exist. In this paper, we describe the first system for exploring the execution space of JavaScript code using symbolic execution. To handle JavaScript code's complex use of string operations, we design a new language of string constraints and implement a solver for it. We build an automatic end-to-end tool, Kudzu, and apply it to the problem of finding client-side code injection vulnerabilities. In experiments on 18 live web applications, Kudzu automatically discovers 2 previously unknown vulnerabilities and 9 more that were previously found only with a manually-constructed test suite.

Modern languages like Java and C# rely on dynamic optimizations in virtual machines for better performance. Current dynamic optimizations are reactive. Their performance is constrained by the dependence on runtime sampling and the partial knowledge of the execution. This work tackles the problems by developing a set of techniques that make a virtual machine evolve across production runs. The virtual machine incrementally learns the relation between program inputs and optimization strategies so that it proactively predicts the optimizations suitable for a new run. The prediction is discriminative, guarded by confidence measurement through dynamic self-evaluation. We employ an enriched extensible specification language to resolve the complexities in program inputs. These techniques, implemented in Jikes RVM, produce significant performance improvement on a set of Java applications.

Software speculation has shown promise in parallelizing programs with coarse-grained dynamic parallelism. However, most speculation systems use offline profiling for the selection of speculative regions. The mismatch with the input-sensitivity of dynamic parallelism may result in large numbers of speculation failures in many applications. Although with certain protection, the failed speculations may not hurt the basic efficiency of the application, the wasted computing resource (e.g. CPU time and power consumption) may severely degrade system throughput and efficiency. The importance of this issue continuously increases with the advent of multicore and parallelization in portable devices and multiprogramming environments. In this work, we propose the use of transparent statistical learning to make speculation cross-input adaptive. Across production runs of an application, the technique recognizes the patterns of the profitability of the speculative regions in the application and the relation between the profitability and program inputs. On a new run, the profitability of the regions are predicted accordingly and the speculations are switched on and off adaptively. The technique differs from previous techniques in that it requires no explicit training, but is able to adapt to changes in program inputs. It is applicable to both loop-level and function-level parallelism by learning across iterations and executions, permitting arbitrary depth of speculations. Its implementation in a recent software speculation system, namely the behavior-oriented parallelization system, shows substantial reduction of speculation cost with negligible decrease (sometimes, considerable increase) of parallel execution performance.

Based on accurate landscape information acquired from multi-temporary satellite imagery, dynamics of landscape spatial patterns of urban-rural ecotone are examined in this study. A case study is carried out in the east part of Beijing city. At first, with a reference of land use classification, landscape is divided into 5 types and 11 subtypes using satellite SPOT images of 1986(SP0T1), 1996(SPOT2) and 2002(SPOT5). Then landscape indices of each year are calculated basing on the classification results of landscape types, including fragmentation index, diversity index, and evenness index for total characters, and fractal dimension index, nearest neighbor index, mean patch nearest distance, proportion of patch area and patch density for single landscape patch measuring. In addition, according to analysis of those indices and dynamics of landscape spatial pattern in that area, driving factors are investigated in order to find the relationships between landscape evolution and urban development.

The complexity inherent in urban emergency response brings lots of challenge for decision making. An effective test bed for analyzing the dynamics of emergency response for different scenarios and evaluating the strength of rescue strategies is crucial for improving disaster planning, preparedness and mitigation. Computer Simulation is becoming an increasingly important tool which provides an effective and low-cost approach to address this problem. The agent based approach has been accepted in various areas and multi agent systems have constituted one of the key technologies in disaster rescue simulations, since interactions with human activities should be implemented within them. In this paper agent-based disaster simulation system and its components are discussed, is s u es t h at simulation system will have when it is applied for practical usages are also discussed. These discussions will delve into future research topics in developing practical disaster simulation systems.

This paper first discusses the technical route for applying 3S technique in post-disaster reconstruction planning, and gives the general framework of the post-disaster reconstruction planning supporting system (PRPSS). Furthermore the database model and the function model of PRPSS is analyzed. At last the paper gives the technical methods to achieve the system and the application cases of the system in the post-disaster reconstruction planning. The practice indicates that PRPSS developed with 3S technique can be used to many aspects of the post-disaster reconstruction planning, such as the information extraction of geological disasters, disaster situation assessment, site selection analysis, thematic mapping and so on. The post-disaster reconstruction planning results can be made more scientific and rational by using PRPSS.

Many studies have shown that the best performer among a set of garbage collectors tends to be different for different applications. Researchers have proposed application-specific selection of garbage collectors. In this work, we concentrate on a second dimension of the problem: the influence of program inputs on the selection of garbage collectors.

In recent years, it was one of the currently hot issues, which study the evolution of river way and hydro-system using 3S technology. Aerial photos, TM/ETM+ and SPOT5 imagery were offered to study the evolution of riverway in QingKou region. Automatic detection for change information method such as spectral variation method and false color composite method were used to detect change information, the results showed that the former method was better. Five Rivers in Qingkou Region were selected for our research, change information such as length and width was extracted separately. On this basis, the driving factors of evolution were analyzed. The riverway gradually become steady due to effective river rectification, the influence caused by man-made factors was greater than natural factors in the evolution.

Traditional land cover performance from remote sensing imagery using the statistical characteristics of the pixel have encountered a lot of difficulties in dealing with the issue of classification of high-resolution images. Object-oriented classification techniques based on image segmentation are a good solution to this problem. Qingkou region, the Beijing-Hangzhou Grand Canal and other eight natural rivers junctions, has take break dramatic changes in the last 50 years, therefore, research on land cover performance in this region have great significance. In this paper, object-oriented classification method was used to extract land cover information from SPOT5 imagery in Qingkou region. The imagery was segmented in two scale parameters to create homogeneous objects for different classes. And the overall classification stability is up to more than 87%. It is available to provide research data for land cover performance evaluation in Beijing-Hangzhou Grand Canal and development research on cities along the canal.

In this paper, the Kuusk-Nilson forest reflectance and transmittance (FRT) model was inverted to retrieve the overstory and understory leaf area index (LAI) of forest stands in the Longmenhe Nature Reserve (Xingshan County, Hubei province, China). Atmospherically and topographically corrected hyperspectral Hyperion imagery and field data had been input to retrieve the overstory and understory LAI simultaneously using FRT inverted model. An uncertainty and sensitivity matrix was used to analyze the sensitivity of the FRT model parameters based on field data. Twenty-one Hyperion bands were selected based on principal-components analysis and band importance from total Hyperion bands. Eight different Hyperion band combinations from 21 Hyperion bands were tested to evaluate the accuracy of the inversion of overstory and understory LAI. Our study showed that the overstory LAI of stands can be better retrieved when considering the understory LAI compared to only use total LAI.

Wetland is a kind of natural resource with fast dynamic changing. The study of wetland information extracting way which is fast and measurable is front of wetland remote sensing. SPOT5 imagery has high spatial resolution and multi-spectral resolution, which provided a rich, reliable and accurate source data for wetland resources investigation. This paper took the Nansihu wetlands as the study area and combining with the characteristics of SPOT5 imagery discussed wetland resources investigation methods. In the study, SWIR band threshold method was used to extract information of water bodies, Spectral Relations method was used to get rid of shadows from extraction of water bodies and Visual Interpretation method based on GIS information was used for the extraction of other type wetlands. Using above method, Nansihu area 12 categories wetlands information were extracted. Practice proved that this method was simple and practical.

The primary goal of this research was to study the method to calculate the green space accessibility, which was used to assess the capacity and chance of inhabitants to approach green space. Many factors affecting the green space accessibility were taken into account, such as attraction of green space, distribution of population, land use pattern and traffic cost. Based on GIS software, the study area was divided into regular grids with the size of 500 meters. For each grid, the resistance for accessing a green space was calculated regarding to qualified factors based on gravity & spatial interaction model and shortest path analysis. Then the green space accessibility of whole city was accumulated by the sum of all grids' computing results. Taking Beijing city as a case study, SPOT imagery in three periods were used to acquire information of green space to assess its accessibility and check the evolution process.

This paper reports the results of a research project on applying geospatial information technology to the protection of China's Grand Canal. The canal is the longest man-made waterway in the world. The earliest section of the canal can be dated back to 500 BC. The course of the canal has been changed many times during its history mainly depending on the locations of China's capitals. The current course of the canal was completed in 1291 running from Beijing in the north to Hangzhou in the south. Its total length is 1200 miles. It is very difficu.lt to use conventional methods to conduct archaeological studies on the canal. This study focuses on the section of the canal near Nanwang township in Shandong province. The archaeological study of this research is based on the remote sensing technology with data from various sources such as ancient maps, historical documents, aerial photographs in 1950s, and recent remote sensing imageries. The results of this research are verified in the field.

This paper presents the major outcomes of a multidisciplinary research project - the Research and Application of Spatial Information Technology for Conservation of Large-scale Heritage Sites, which is within the National Key Technology R&D Program and aims to apply spatial information technology for cultural heritage conservation, especially the large-scale heritage sites, and takes the Grand Canal of China (GCC) to conduct a case study. At first an integrated framework for the application of spatial information technology in conserving cultural heritage was conceived comprehensively. Then some standards were established with the effort of domain experts including the standards for metadata, spatial data and professional data. Besides that a spatial data base of GCC, a PDA based system for data collecting and GIS system for GCC was set up and reviewed in detail. Finally this paper appraises the project's established results objectively; present the research planning in the future.

Grand Canal of China (GCC) is considered as a significant historical cultural heritage in China. Conservation and management of GCC are becoming an extremely urgent issue. In this paper, a design of GCC conservation integrated information system was designed after analyzing GCC conservation of business requirement. The design focuses on the analysis of the system's overall structure, function model and data model. Practice shows that, GCC conservation integrated information system can play an important role in GCC's heritage data collection, data management, data analysis, conservation planning of GCC and the establishment of dynamic monitoring and management.

Remote sensing dynamic monitoring of land use can detect the change information of land use and update the current land use map, which is important for rational utilization and scientific management to land resources. This paper discussed the technological procedure of land use dynamic monitoring, including the process of remote sensed images, the information classification and extraction of remote sensed imagery, and analysis of land use changes. Based on SPOT imagery data in three periods, the paper took Beijing city as an example, extracted the land use information during 1986-2004, and the land use changes were required in the period. The object-oriented method was used to extract information, and contrastive method after classification was used to confirm change zones.

Vegetation cover is of great significance in understanding climate change process due to its vital role in controlling water and carbon cycles. The properties of vegetation's surfaces are usually estimated by remotely sensed data through regression models or physical-based models, which simulates the interactions of solar radiation with the vegetation medium. In real domain, the spectral responses measured by the sensor in forested area are strongly influenced by the different understory natural conditions that limit the possibility of applying both retrieval methods to predict overstory vegetation parameters. Understory information is therefore needed for estimating trees' parameters; moreover from a biodiversity point of view and perspective of forest management, understory represents a critical component of forest ecosystem that needs a better characterization. An experiment has been conducted using hyperion and panchromatic QuickBird data to explore the status of different vegetation's understory under a sparse forest in the Longmenhe Nature Reserve, China. Understory vegetation information of study area is classified into five classes. The novel aspect of the method is the integration of spectral (hyperspectral) domain fusion and spatial domain fusion techniques within a multi-layer perceptron artificial neural network model. Real data from the experiment on a limited ground as well as hyperion and QuickBird data are used as input dataset. A nonlinear artificial neural network achieved a classification accuracy of 80% despite the presence of co-occurring mid-story and understory vegetation. The achieved results show that this method is able to identify the different vegetation information under the tree canopy. Our studies suggest that it is necessary to incorporate the geographic and vegetation community prior information to further improve the accuracy in order to monitor understory vegetation.

Soil erosion is a major environmental problem worldwide, threatening the human sustainable development. Global water erosion and wind erosion affect 1094 and 549 Mha, respectively. Soil erosion concerns multi factors, for example, land cover, climate, vegetation cover, topographic factors. Soil erosion is also different in different spatial and temporal scales. To monitor and assess the extent of soil erosion, multi data concerning these influencing elements need to be considered by combined use. Among these factors that influent the process of soil erosion, vegetation cover and slope steepness are selected. The vegetation cover data in the Chengdu plain have been estimated from normalized difference vegetation index derived from Landsat-7 ETM+ image acquired at 2000-11-02. Slope steepness is computed based on the pixels of DEM (Digital Elevation Model), the pixels of DEM are transformed from the 1:100000 terrain map. Vegetation cover and topographic factor can be combined as a cross tab model. A soil erosion risk map with six grades can be drawn. Using the method the probability and the extent of soil erosion can be measured. Remote sensing data provide a significant information source for mapping, monitoring and predicting current rapid soil erosion. In the analysis process of soil erosion monitoring geography information system takes very important roll. It can fuse different thematic data and different formats of data. This paper gives a brief synthesis of the information obtainable from remote-sensing data and DEM data, and assesses the status of soil erosion in Chengdu plain.