Search

Traditional enhancement techniques can improve the contrast of low-light and low-resolution images, but they fail to recover their resolution. Conversely, traditional super-resolution (SR) algorithms can enhance resolution but not restore contrast. To address this issue, a novel progressive joint enhancement and SR tactic, named PESiT, is proposed to synchronously improve contrast and resolution in low-light and low-resolution images. PESiT comprises an enhanced multi-scale Retinex module followed by a blind SR module with regularization optimization. In the first module, the common logarithm is replaced with an S-function to expand the intensity distribution of images and prevent color inversion. In the latter module, those merits of reconstruction- and learning-based tactics are combined to tackle various unknown degradations by imposing consistency constraints on high- and low-resolution image pairs. Extensive experiments on public datasets demonstrate the robustness and superiority of PESiT in processing low-light and low-resolution images under various scenarios. Compared with state-of-the-art techniques, PESiT achieves superior performance, e.g., the highest peak signal-to-noise ratio, structural similarity index, feature similarity index, and the lowest learned perceptual image patch similarity, highlighting its validity in achieving optimal image quality improvements.

This paper presents a reflective-type phase shifter (RTPS) capable of a full 360° phase shift range with low-loss and compact for 5.47–5.85-GHz WLAN wireless backhaul. Intuitive and concise analyses of common reflective loads, including single-element single-tunable (SEST), dual-element single-tunable (DEST), and three-element single-tunable (TEST), are analyzed, elaborating their limitations. The four-element dual-tunable (FEDT) load technique is proposed to increase the phase shift range. In addition, the influence factors of phase shift step and insertion loss (IL) are comprehensively discussed. Measurement results show that the proposed RTPS provides a phase shift range greater than 360°, a phase shift step less than 15°, and an IL less than 2.3-dB in the entire frequency band. Meanwhile, the size of this RTPS is only $9.5\times 8.2$ mm2. Compared to other RTPS, this design has the advantages of low IL, compact size, wide band, high precision, and easy control.

Public-key dual-receiver encryption (PK-DRE) is a kind of particular public-key encryption for enabling two independent recipients to obtain the same plaintext from the same ciphertext. Due to its dual-receiver property, PK-DRE is quite helpful in many scenarios, such as deniable authentication, global key escrow, security puzzle, and even blockchain. In this paper, we revisit the PK-DRE scheme $\mathtt {CFZ}14$ proposed at CT-RSA 2014 and propose a variant. This variant is original from a new security proof which allows us to remove some steps in $\mathtt {CFZ}14$ . To the best of our knowledge, the obtained variant is more efficient than the existing PK-DRE schemes in terms of public verifiability and key size.

Chest X-ray is one of the most common radiological examinations for screening thoracic diseases. Despite the existing methods based on convolution neural network that have achieved remarkable progress in thoracic disease classification from chest X-ray images, the scale variation of the pathological abnormalities in different thoracic diseases is still challenging in chest X-ray image classification. Based on the above problems, this paper proposes a residual network model based on a pyramidal convolution module and shuffle attention module (PCSANet). Specifically, the pyramid convolution is used to extract more discriminative features of pathological abnormality compared with the standard $3\times 3$ convolution; the shuffle attention enables the PCSANet model to focus on more pathological abnormality features. The extensive experiment on the ChestX-ray14 and COVIDx datasets demonstrate that the PCSANet model achieves superior performance compared with the other state-of-the-art methods. The ablation study further proves that pyramidal convolution and shuffle attention can effectively improve thoracic disease classification performance. The code is published in https://github.com/Warrior996/PCSANet.

In real-world driving scenarios, the model mismatch can severely impair the robustness of the tracking system controlled by Model Predictive Control (MPC). Tube-based MPC (TMPC) addresses this problem by keeping the model mismatch error in an invariant tube. The TMPC algorithms, however, cannot deal with state-dependent uncertainty since TMPC relies on the fixed tubes. This paper presents a practical algorithm for improving the capability of TMPC to handle multiplicative uncertainty. Firstly, this algorithm adopts a Homothetic Tube-based MPC (HTMPC) framework to optimize the system’s future trajectory and tube geometry simultaneously, which dynamically resizes tubes according to uncertainty and the system’s current state. Secondly, this algorithm provides both the feasible formulation of the tube and the homothetic factor with low computational complexity. Thirdly, we aim to systematically evaluate the algorithm’s robustness by the simulations of different scenarios where the system parameters and the measurement noises might change over time. We have conducted and analyzed the Monte-Carlo simulations to compare the robustness and tracking capability of the proposed algorithm and other control algorithms. The comparative analysis shows that the HTMPC algorithm provides a higher level of performance than MPC and TMPC, and it performs closely to the robust controller based on the immersion and invariance (I&I) principle.

Load forecasting is critical for power system operation and market planning. With the increased penetration of renewable energy and the massive consumption of electric energy, improving load forecasting accuracy has become a difficult task. Recently, it was demonstrated that deep learning models perform well for short-term load forecasting (STLF). However, prior research has demonstrated that the hybrid deep learning model outperforms the single model. We propose a hybrid neural network in this article that combines elements of a convolutional neural network (1D-CNN) and a long short memory network (LSTM) in novel ways. Multiple independent 1D-CNNs are used to extract load, calendar, and weather features from the proposed hybrid model, while LSTM is used to learn time patterns. This architecture is referred to as a CNN-LSTM network with multiple heads (MCNN-LSTM). To demonstrate the proposed hybrid deep learning model’s superior performance, the proposed method is applied to Ireland’s load data for single-step and multi-step load forecasting. In comparison to the widely used CNN-LSTM hybrid model, the proposed model improved single-step prediction by 16.73% and 24-step load prediction by 20.33%. Additionally, we use the Maine dataset to verify the proposed model’s generalizability.

The stirrer design is important in a reverberation chamber measurement system. Previous study shows that the rotating radius of the stirrer plays a key role for the stirrer performance. However, to identify the contribution from the structure, optimizing the stirrer structure while keeping the stirring volume unchanged is necessary. In this paper, when the stirring volume is kept invariant, we show that the detailed structure of stirrers can be optimized to improve the performance but the effect is not significant. A comparative study is given to confirm the effect of zigzag boundaries on the stirrers. Both simulations and measurements confirm the performance improvement, key performance indicators such as field uniformity and correlated angles are simulated and measured.

Bearings are important part of aerospace equipments. Therefore, bearings fault diagnosis and fault detection is necessary for the safe operation. Generally, the bearing are core component and runs in a complex and heavy background noise environment. Thence, the bearing signals will be overwhelmed by noise. At present, the diagnosis methods of bearings are based on the experience of experts which costs a lot of labor and time-consuming. Consequently, to address these inadequacies, a novel method based on variational mode decomposition (VMD) algorithms and Alexnet neural network has been presented. Firstly, it decomposes the nonstationary bearing signals into intrinsic mode functions adaptively. However, background noise will seriously affect the number of signal decompositions and reduce accuracy even over-decomposition. Thence, this paper proposed the hierarchical fuzzy entropy method to adaptively extract weak fault characteristics of bearings. Then Alexnet neural network has been utilized to learn the relationship of fault features and bearings health conditions. The architecture of Alexnet diagnosis model is convenient and efficient. Finally, two tentative dataset are adopted to verify the effectiveness and feasibility of the method proposed in this article. The results show that the presented method has a higher fault diagnosis accuracy rate than traditional convolutional neural networks.

Autonomous sailing robots are a new type of green ship that use wind energy to maintain continuous cruising operations. Compared with traditional algorithms, swarm intelligence optimization algorithms have better intelligence and adaptation. An intelligent algorithm acts as one of the most important solutions to the path planning problem of autonomous sailing robots. The beetle swarm optimization, which is a novel intelligent method that combines the search mechanism of a single beetle with the particle swarm optimization algorithm, is utilized to obtain the optimal path. In this study, the track navigation control of an improved mathematical model of a sailing ship is introduced, and the navigation is tested using a downsized prototype of an autonomous sailing robot. The improved beetle swarm optimization is proposed here by dynamically changing the step size factor and the inertia weight formula. In the iteration of the improved beetle swarm optimization algorithm, the location update cooperates with the beetle monomer search mechanism to learn the update strategy of the particle swarm optimization algorithm. Combinatorial strategies can speed up the overall iterative convergence speed and reduce the possibility that the population will fall into a locally optimal solution. The simulation results demonstrate the robustness, efficiency, and feasibility of the improved beetle swarm optimization in different cases. The research results can provide some references and ideas for the autonomous intelligent navigation control design of autonomous sailing robots.

The quality and accessibility of modern financial service have been quickly and dramatically improved, which benefits from the fast development of information technology. It has also witnessed the trend for applying artificial intelligence related technology, especially machine learning to the finance and security industry ranging from face recognition to fraud detection. In particular, deep neural networks have proven to be far superior to traditional algorithms in various application scenarios of computer vision. In this paper, we propose a deep learning-based video analysis system for automated compliance audit in stock brokerage, which in general consists of five modules here: 1) Video tampering and integrity detection; 2) Objects of interest localization and association; 3) Analysis of presence and departure of personnel in a video; 4) Face image quality assessment; and 5) Signature action positioning. To the best of our knowledge, this is the first work that introduces remote automated compliance audit system for the dual-recorded video in finance and security industry. The experimental results suggest our system can identify most of the potential non-compliant videos and has greatly improved the working efficiency of the auditors and reduced human labor costs. The collected dataset in our experiment will be released with this paper.

Solar power forecasting is critical to ensure the safety and stability of the power grid with high photovoltaic power penetration. Machine learning methods are compelling in solar forecasting. These methods can capture the complex coupling relationship between different meteorological factors without physical modeling. Most of the existing machine learning based forecasts follow the batch learning manner. Once the training is completed, the structure and parameters of the model are usually no longer adjusted. However, the climate is complex and dynamic. It is difficult for a fixed model to adapt to the climate characteristics of different regions or periods. Therefore, an online domain adaptive learning approach is proposed in this paper. Knowledge can be selectively accumulated or forgotten in its iterative process. As weather changes, the model can dynamically adjust its structure to adapt to the latest weather conditions. Unlike existing adaptive iterative methods, the proposed adaptive learning approach does not rely on the labels of the test data in the updating process. Experiments show that this method can effectively track changes in data distribution and obtain reliable prediction results.

In this paper, the research of image noise reduction based on semi-supervised learning is carried out, and the neural network is used to reduce the noise of the image, so as to achieve more stable and good image display ability. Based on the convolutional neural network algorithm, the role of activation function optimization network is studied, combined with semi-supervised learning modes such as multi-feature extraction technology, to learn and extract the key features of the input image. Semi-supervised residual learning based on convolutional network is a good image denoising and denoising network model. Compared with other excellent denoising algorithms, it has very good results. At the same time, it greatly improves the image noise pollution and makes the image details clearer. At the same time, compared with other image denoising algorithms, this algorithm can show a good peak signal-to-noise ratio under various noise standard deviations. Through the research in this article, it is verified that the improved convolutional neural network denoising model and multi-feature extraction technology have strong advantages in image denoising.

This paper proposes a new Simultaneous Localization and Mapping (SLAM) method on the basis of graph-based optimization through the combination of the Light Detection and Ranging (LiDAR), RGB-D camera, encoder and Inertial Measurement Unit (IMU). It can conduct joint positioning of four sensors by taking advantaging of the unscented Kalman filter (UKF) to design the related strategy of the 2D LiDAR point cloud and RGB-D camera point cloud. 3D LiDAR point cloud information generated by the RGB-D camera under the 2D LiDAR has been added into the new SLAM method in the sequential registration stage, and it can match the 2D LiDAR point cloud and the 3D RGB-D point cloud by using the method of the Correlation Scan Matching (CSM); In the loop closure detection stage, this method can further verify the accuracy of the loop closure after the 2D LiDAR matching by describing 3D point cloud. Additionally, this new SLAM method has been verified feasibility and availability through the processes of theoretical derivation, simulation experiment and physical verification. As a result, the experiment shows that the multi-sensor SLAM framework designed has a good mapping effect, high precision and accuracy.

Vehicles authentication, the integrity of messages exchanged, and privacy-preserving are essential features in vehicular ad hoc network (VANETs) security. Most of the previously proposed VANETs security solutions do not sufficiently satisfy the security and efficiency requirements. Besides, most of those solutions are heavily dependent on the system key and long-term sensitive data stored in an ideal tamper-proof device, which may not be practical or ideal for resource-constrained onboard units (OBUs), especially in the case of an unexpected cloning or physical attack. Therefore, a robust authentication solution should consider those security issues and the nature of resource-constrained nodes. To satisfy all these requirements, we propose a lightweight multi-factor authentication and privacy-preserving security solution for VANETs. It employs a combination of physically unclonable functions (PUF) and one-time dynamic pseudo-identities as authentication factors. Furthermore, it eliminates the heavy dependency on the system key by decentralising the wide precinct of the certificate authority (CA) into regional domains and achieves robust control of domains keys. A detailed analysis demonstrates that our scheme efficiently meets the VANETs security requirements, and offers more suitable communication and computation costs and features than existing schemes.

This paper presented an adaptive and fractional-order sliding mode control(FOSMC) method for the unmanned quadrotor helicopter. The aircraft system includes actuator fault and external disturbances. The switching sliding mode law enables the system to reach the predefined sliding surface from arbitrary states. Then the equation control law keeps the trajectory stay over the sliding hyperplane. In order to make sure sliding motion from the arbitrary states to the surface within limited time, a novel fractional-order power switching control law is developed. System actuator failures are compensated online with adaptive control laws. The controllers are derived from the Lyapunov theory, which guarantees that the controllability and feasibility. This novel control strategy has higher tracking accuracy through the timely faults and disturbances compensation law. The presented fractional-order sliding mode scheme improves the speed of system convergence and shortens the reaching time. The adaptive strategy estimated the bounds of the disturbances and good robustness has been achieved. Simulation results shown that the presented strategy has numerous advantages in terms of attitude and position tracking.

Deep neural networks have revolutionized many areas of computer vision, but they require notoriously large amounts of labeled training data. For tasks such as semantic segmentation and monocular 3d scene layout estimation, collecting high-quality training data is extremely laborious because dense, pixel-level ground truth is required and must be annotated by hand. In this paper, we present two techniques for significantly reducing the manual annotation effort involved in collecting large training datasets. The tools are designed to allow rapid annotation of entire videos collected by RGBD cameras, thus generating thousands of ground-truth frames to use for training. First, we propose a fully-automatic approach to produce dense pixel-level semantic segmentation maps. The technique uses noisy evidence from pre-trained object detectors and scene layout estimators and incorporates spatial and temporal context in a conditional random field formulation. Second, we propose a semi-automatic technique for dense annotation of 3d geometry, and in particular, the 3d poses of planes in indoor scenes. This technique requires a human to quickly annotate just a handful of keyframes per video, and then uses the camera poses and geometric reasoning to propagate these labels through an entire video sequence. Experimental results indicate that the technique could be used as an alternative or complementary source of training data, allowing large-scale data to be collected with minimal human effort.

Numerous object-tracking and multiple-person-tracking algorithms have been developed in the field of computer vision, but few trackers can properly address the issue of when a pedestrian is partially or fully occluded by other objects or persons. In order to achieve efficient pedestrian tracking in various occlusion conditions, a pedestrian tracking framework is proposed and developed based on the deep learning networks. First, a pedestrian detector is trained as a tracking mechanism based on the Faster R-CNN, which narrows the search range and efficiently improves accuracy, as compared with the traditional gradient descent algorithm. Second, in the process of target matching, a color histogram and scale-invariant feature transform are combined to provide the target model expression, and a full convolution network (FCN) is trained to extract the pedestrian information in the target model, based on an FCN image semantic segmentation algorithm that can remove background noise effectively. Finally, the extensive experiments on a commonly used tracking benchmark show that the proposed method achieves better performance than the other state-of-the-art trackers in various occlusion situations.

Co-occurrence information between words is the basis of training word embeddings; besides, Chinese characters are composed of subcharacters, words made up by the same characters or subcharacters usually have similar semantics, but this internal substructure information is usually neglected in popular models. In this paper, we propose a novel method for learning Chinese word embeddings, which takes full use of external co-occurrence context information and internal substructure information. We represent each word as a bag of subcharacter n-grams, and our model learns the vector representation corresponding to the word and its subcharacter n-grams. The final word embeddings are represented as the sum of these two kinds of vector representation, which makes the learned word embeddings can take into account both the internal structure information and external co-occurrence information possible. The experiments show that our method outperforms state-of-the-art performance on benchmarks.

Vehicles in a vehicular ad-hoc network (VANET) broadcast beacons giving safety-related and traffic information. In an open-access environment, this means that the VANET is susceptible to security and privacy issues. In this paper, we propose a new pseudo-identity-based scheme for conditional anonymity with integrity and authentication in a VANET. The proposed scheme uses a pseudonym in the joining process with the road-side unit (RSU) to protect the real identity even from the RSU, in case it is compromised. All previous identity-based schemes have been prone to insider attackers, and have not met the revocation process. Our scheme resolves these drawbacks as the vehicle signs the beacon with a signature obtained from the RSU. Our scheme satisfies the requirements for security and privacy, and especially the requirements for message integrity and authentication, privacy preservation, non-repudiation, traceability, and revocation. In addition, it provides conditional anonymity to guarantee the protection of an honest vehicle's real identity, unless malicious activities are detected. It is also resistant to common attacks such as modification, replay, impersonation, and man-in-the-middle (MITM) attacks. Although the numerous existing schemes have used a bilinear pairing operation, our scheme does not depend on this due to the complex operations involved, which cause significant computation overhead. Furthermore, it does not have a certification revocation list, giving rise to significant costs due to storage and inefficient communication. Our analysis demonstrates that our scheme can satisfy the security and privacy requirements of a VANET more effectively than previous schemes. We also compare our scheme with the recently proposed schemes in terms of communication and computation and demonstrate its cost-efficiency and appropriateness in working with the VANET. Meanwhile, the computation costs of the beacon signing and verification in our scheme are reduced by 49.9% and 33.3%, respectively.

A high-power borehole-ground electromagnetic transmitter is an important device used to detect deep metal mineral resources. This paper presents the control principle of the transmitter and the arbitrary frequency table transmission technology. The latter includes the synthesis and the predetermined time transmission of an arbitrary frequency table and a seamless frequency conversion based on an embedded system design and hardware programming technology. We used the arbitrary frequency table transmission function to reduce the impact on the transmitting system from processing the frequency conversion and achieve a smooth transition for the transmitting system when converting the high-power frequency table. These features improved the stability and the reliability of the transmitter. We also performed several field tests that support the advantages of the arbitrary frequency table transmission technology. The flexibility of the best customized transmission frequency table in the field improves the efficiency during the electromagnetic exploration and provides high-quality electromagnetic data.

Motion estimation is vital in many computer vision applications. Most existing methods require high quality and large quantity of feature correspondence and may fail for images with few textures. In this paper, a photometric alignment method is proposed to obtain better motion estimation result. Since the adopted photometric constraints are usually limited to the required illumination or color consistency assumption, a new generalized content-preserving warp (GCPW) framework, therefore, is designed to perform photometric alignment beyond color consistency. Similar to conventional content-preserving warp (CPW), GCPW is also a mesh-based framework, but it extends CPW by appending a local color transformation model for every mesh quad, which expresses the color transformation from a source image to a target image within the quad. Motion-related mesh vertexes and color-related mapping parameters are optimized jointly in GCPW to get more robust motion estimation results. Evaluation of tens of videos reveals that the proposed method achieves more accurate motion estimation results. More importantly, it is robust to significant color variation. Besides, this paper explores the performance of GCPW in two popular computer vision applications: image stitching and video stabilization. Experimental results demonstrate GCPW’s effectiveness in dealing with typical challenging scenes for these two applications.

The weighted K-nearest neighbor (WKNN) algorithm is one of the most frequently used algorithms for indoor positioning. However, the traditional WKNN algorithm weights the reference points' coordinates by the inverse of the received signal strength indication (RSSI) difference, which is not accurate enough because of the exponential relationship between RSSI and physical distance. Furthermore, methods based on probabilistic model or data fusion do not consider the uneven spatial resolution of the Wi-Fi RSSI. Therefore, in order to improve the positioning accuracy of traditional location algorithms, this paper proposes a new weighted algorithm based on the physical distance of the RSSI. Experiments were conducted in an office building and the results demonstrate that the proposed method considerably outperforms the KNN, Euclidian-W-KNN, Manhattan-W-KNN, EWKNN, LiFS, and GPR in terms of positioning accuracy, which is defined as the cumulative distribution function of position error.

In this paper, a novel systematic method on the evaluation of quantitative precipitation forecast (QPF) errors from the perspective of rain-area shape verification is proposed. The method aims to improve the accuracy and efficiency of conventional station-based verifications (i.e., standard skill scores), which are insensitive to the biases of station location and rain-area shape and tend to ignore the continuity of precipitation in time and space. The method develops and combines the shape verification indexes, which include the overlap ratio of a forecasted rain-area (Ratiof), the overlap ratio of the ground rainarea (Ratiot), the Jaccard similarity coefficient between the shape of the QPF area, the shape of the ground rain-area (Jaccardshape), the critical success index (CSI) for the rain-area shape (CSIshape), the probability of detection (POD) for the rain-area shape (PODshape), and the false alarm ratio (FAR) for the rain-area shape (FARshape). This definition of QPF verification is applied to a rain event from 2016/08/02 00:30 to 2016/08/02 03:24 in the Guangdong Province. The decomposition of QPF errors into station-based errors and shape error components provides powerful insight into the effects of overall forecasting performance. The experimental results of this investigation show that the proposed method provides an opportunity to assess QPFs objectively and further promote advanced forecasting technologies from this perspective.

With the increasing demands for mineral resources in China, artificial-source electromagnetic prospecting is playing an increasingly significant role. Consequently, electromagnetic instruments, which are crucial for prospecting, have attracted considerable attention. In order to verify the design targets of a self-developed high-power multifunction borehole-ground electromagnetic transmitting system, field tests are carried out in the mining area of Linxi, in Inner Mongolia. Design targets, including the stability of the current waveform, maximum transmitting voltage, maximum transmitting current, maximum output power, the continuous working time of the transmitting system with high-power output, transmission in single and dual boreholes, and multifunction technical target, have been tested. The test results demonstrate that the proposed system attains the expected design targets, and its performance is excellent. Moreover, the transmitting system can meet the requirements of controlled-source electromagnetic exploration and provide valuable reference for researchers in related fields.

Scene text information extraction plays an important role in many computer vision applications. Most features in existing text extraction algorithms are only applicable to one text extraction stage (text detection or recognition), which significantly weakens the consistency in an end-to-end system, especially for the complex Chinese texts. To tackle this challenging problem, we propose a novel text structure feature extractor based on a text structure component detector (TSCD) layer and residual network for Chinese texts. Inspired by the three-layer Chinese text cognition model of a human, we combine the TSCD layer and the residual network to extract features suitable for both text extraction stages. The specialized modeling for Chinese characters in the TSCD layer simulates the key structure component cognition layer in the psychological model. And the residual mechanism in the residual network simulates the key bidirectional connection among the layers in the psychological model. Through the organic combination of the TSCD layer and the residual network, the extracted features are applicable to both text detection and recognition, as humans do. In evaluation, both text detection and recognition models based on our proposed text structure feature extractor achieve great improvements over baseline CNN models. And an end-to-end Chinese text information extraction system is experimentally designed and evaluated, showing the advantage of the proposed feature extractor as a unified feature extractor.

New item or topic profiling and recommendation are useful yet challenging, especially in face of a “cold-start” situation with sparse user-item ratings for the new arrivals. In this paper, a method of acquiring review opinions of the “sentinel” users on the cold-start items is proposed to elicit those items' latent profiles, and thus both user-specific ratings and future popularity of the items can be predicted simultaneously. Specifically, such a joint prediction task is formulated as a two-stage optimization problem, and a sentinel user selection algorithm is devised to facilitate effective latent profiles extraction for both item ratings and popularity predictions. Experiments with microblogging and movie data sets corroborate that the proposed method is capable of mitigating the cold-start problem and it outperforms several competitive peer methods.

The objective of this paper is to estimate the number of moving objects that passes through a specific area without fully decoding the H.265/high-efficiency video coding (HEVC) bitstreams. First, the foreground prediction blocks are extracted according to the motion vectors of the H.265/HEVC bitstreams. Next, these foreground prediction blocks are clustered into the region of interests (ROIs), which are the possible area position of moving objects in the current frame. Finally, the state of moving objects is identified by matching moving objects and these ROIs. In order to estimate the number of moving objects, which move toward a pre-defined direction, a tripwire is set to a detecting area. Any moving objects crossing the tripwire and satisfying the intrusion conditions are counted. With the proposed method, the number of moving objects can be directly estimated in the compressed domain video. This approach significantly increase the processing speed more than 400% at the cost of less than 0.02% accuracy degradation compared with the traditional pixel domain approach. The research results can be applied to traffic management, real-time analysis of surveillance application, and other related areas.