Guofang Lv

Stereo depth estimation under different camera calibration and alignment errors

Depth estimation is a fundamental issue in computational stereo. To obtain accurate stereo depth estimation, all mechanical parameters with a high precision need to be measured in order to achieve subpixel accuracy and to match features between two different images. This paper investigates accurate depth estimation with different mechanical parameter errors, such as camera calibration and alignment errors, which mainly result from camera lens distortion, camera translation, rotation, pitch, and yaw. For each source of the errors, a model for the error description is presented, and the accurate depth estimation due to this error is quantitatively analyzed. Depth estimation algorithms under an individual error, and with all the errors, are given. Experimental results show that the proposed models can rectify the errors and calculate the accurate depths effectively.

An improved similarity measure in particle filters for robust object tracking

Infrared object tracking plays a key role in many research fields, and there is a series of work on applying particle filter to this tracking problem. Most of the PF-based tracking algorithms utilize the Bhattacharyya coefficient as a similarity measure, however, its performance in infrared object tracking is limited due to insufficient discriminative power. In this paper, we present a combined similarity measure under the particle filter framework, which integrates the advantages of the Bhattacharyya coefficient, histogram intersection, and structural similarity. The experimental results are gained by using different infrared image sequences, which show that the proposed measure gives superior discriminative power and achieves more robust and stable tracking performance than the traditional approach.

Method for video incident detection based on biological visual Mechanism

Event detection is an important research in video surveillance technology. This paper proposed a method for traffic event detection based on visual Mechanism on the background of traffic video surveillance applications. In this method, based on the extraction of video target motion characteristics, it extracted abnormal targets mainly through the features merging and significant competitive in video frames. Then it judged the events of abnormal targets. Finally, the simulation results of the method show that the method can effectively simplify the calculation of event detection.

Detecting Saliency in Infrared Images via Multiscale Local Sparse Representation and Local Contrast Measure

For infrared images, it is a formidable challenge to highlight salient regions completely and suppress the background noise effectively at the same time. To handle this problem, a novel saliency detection method based on multiscale local sparse representation and local contrast measure is proposed in this paper. The saliency detection problem is implemented in three stages. First, a multiscale local sparse representation based approach is designed for detecting saliency in infrared images. Using it, multiple saliency maps with various scales are obtained for an infrared image. These maps are then fused to generate a combined saliency map, which can highlight the salient region fully. Second, we adopt a local contrast measure based technique to process the infrared image. It divides the image into a number of image blocks. Then these blocks are utilized to calculate the local contrast to generate a local contrast measure based saliency map. In this map, the background noise can be suppressed effectually. Last, to make full use of the advantages of the above two saliency maps, we propose combining them together using an adaptive fusion scheme. Experimental results show that our method achieves better performance than several state-of-the-art algorithms for saliency detection in infrared images.

Roubust Visual Object Tracking Using Covariance Features In Quasi-Monte Carlo Filter

Abstract Image covariance features, enabled with efficient fusion of several different types of image features without any weighting or normalization, have low dimensions. The covariance-based trackers are robust and versatile with a modest computational cost. This paper investigates an object tracking algorithm using a sequential quasi-Monte Carlo (SQMC) filter combined with covariance features. The covariance features are used not only to model target appearance, but also to model background. The dissimilarity of target and background is integrated in the SQMC filter as an additional measurement for the particle weight. A target model update strategy using the element of Riemannian geometry is proposed for the variation of the target appearance. Comparison experiments are conducted on several image sequences, and the results show that the proposed algorithm can successfully track the object in the presence of appearance changes, cluttered background and even severe occlusions.

Region Covariance Matrices for Object Tracking in Quasi-Monte Carlo Filter

Region covariance matrices (RCMs), categorized as a matrix-form feature in a low dimension, fuse multiple different image features which might be correlated. The region covariance matrices-based trackers are robust and versatile with a modest computational cost. In this paper, under the Bayesian inference framework, a region covariance matrices-based quasi-Monte Carlo filter tracker is proposed. The RCMs are used to model target appearances. The dissimilarity metric of the RCMs are measured on Riemannian manifolds. Based on the current object location and the prior knowledge, the possible locations of the object candidates in the next frame are predicted by combine both sequential quasi-Monte Carlo (SQMC) and a given importance sampling (IS) techniques. Experiments performed on different type of image sequence show our approach is robust and effective.

Spectrum Analysis-Based Traffic Video Synopsis

Camera-based monitoring systems have a wide range of applications in traffic management, since they can collect more informative data in contrast to other sensors. An increasing number of traffic camera systems collect a large volume of traffic video data daily, forming the Big Data of traffic video. One of the challenges for traffic video processing is their high cost of resources and time, which seriously block the development of intelligent transportation systems. This paper proposes a spectrum analysis method for traffic video synopsis, including motion detection and tracking. Our method can largely remove the background noises and correctly extract motion information. Spatial and temporal spectrum analysis (Fourier transformation) are jointly used to detect objects and their motions in traffic videos. Further, the detected motions are tracked by the particle filter, generating trajectories of motions. Motion detection and tracking results given by our method can provide a synopsis for Big Data of traffic videos. The outperformance of our method is demonstrated comparing to the state of art video analysis methods.

Robust object tracking based on local discriminative sparse representation

Despite much success in the application of sparse representation to object tracking, most of the existing sparse-representation-based tracking methods are still not robust enough for challenges such as pose variations, illumination changes, occlusions, and background distractions. In this paper, we propose a robust object-tracking algorithm via local discriminative sparse representation. The key idea in our method is to develop what we believe is a novel local discriminative sparse representation method for object appearance modeling, which can be helpful to overcome issues such as appearance variations and occlusions. Then a robust tracker based on the local discriminative sparse appearance model is proposed to track the object over time. Additionally, an online dictionary update strategy is introduced in our approach for further robustness. Experimental results on challenging sequences demonstrate the effectiveness and robustness of our proposed method.

An On-line Automatic Flow Measurement Method for an Open Channel under Complex Flow Conditions

Buildings flow measurement method using the pre-established upstream and downstream water levels and flow to estimate the flow is the common method for open channel flow measurement. However, due to the changes of import and export, flow pattern, and hydraulic boundary conditions, traditional mechanism modeling-based flow measurement methods which establish the relation between the upstream-downstream water levels and flow by historical records and empirical equation models are usually not able to meet the demands of precision and adaptability. The improvement is based no the neural network (data-driving). However, the neural network based method is commonly offline and the model parameters are constant in the application.If the degree of opening of the weir sluice gate changes frequently, it is hard to construct a neural network model of high precision for on-line and real-time measurement. This research designs a real-time on-line automatic measurement system, for the Pi River canal weir gate, that collects upstream and downstream water levels and the degree of opening of the gate. Moreover, it establishes a three layer BP neural network model based on on-line real-time data correction. This model comprised of a Kalman filter with forgetting factor and a three layer BP neural network data fusion center. In contrast to the standard hydrometric propeller based method, the average relative error is lower than 5%, meeting the “River Discharge Measurement Criterion” proposed by Ministry of Water Resources of the Peoples Republic of China. Both the precision and the repeatability can cater for the engineering applications.

A Novel Method for Detection of Abrupt Signal Changes in Water Quality Sensors Onboard Autonomous Underwater Vehicles

Autonomous underwater vehicles (AUV) are equipped with a variety of water quality sensors. In this paper, we apply wavelet analysis theory to signals acquired by the AUV and develop a tool for identifying such malfunctions. This paper proposes an abrupt signal change detection method based on wavelet transforms. By selecting a threshold for the high-frequency wavelet coefficients, we were able to detect the abrupt signal. Three evaluation criteria were used in this study to assess signal-to-noise quality. These were combined with experimental analysis, noise detection, and a denoising method which is appropriate for detecting abrupt signal changes.