Adaptive Background Research Articles

实时目标跟踪研究

针对图像序列中运动目标检测、跟踪的难点问题，提出了一种实时运动目标检测与跟踪算法。该算法基于自适应背景建模，获取运动目标背景模型和前景图像，从而实现运动目标检测；通过建立运动目标的位置、大小、形状以及颜色分布模型，构造运动目标全局匹配函数，结合目标活力特征，实现多运动目标连续匹配和跟踪。实验结果表明，相对于传统的运动目标跟踪方法，本文方法明显减少了运算时间，增强了环境适应性，实现了复杂场景下运动目标的准确检测和稳定跟踪，对非刚性目标的形变、旋转具有较强的鲁棒性。In this paper, we propose a real-time mobile target detection and tracking algorithm for challenges of mobile target detection and tracking in sequential images. This algorithm based on the adaptive background modeling obtains background model and front-view images of mobile targets, which is the way to achieve target detection. Continuous matching and tracking of multiple mobile targets are realized through constructing position, size, shape and color distribution of the mobile targets, defining a global matching function for those targets, and associating their vitality characteristics. It is demonstrated by experiments that the algorithm presented in this paper, compared to the traditional methods of mobile target tracking, significantly reduces the computation time, improves adaptive feature to environments, achieves accurate detection and robust tracking of mobile targets in complex environments, and shows strong robustness to deformation and rotation of non-rigid targets.

A Recognition Method of Pedestrians’ Running in the Red Light Based on Image

It is dangerous for pedestrians to run when the traffic shows a red light, but in some cases the pedestrians are breaking the rules. This system will be a meaningful thing if the jaywalking behaviors of pedestrians in the road crossing through the monitoring cameras could be recognized. Then drivers can be informed of the situations in advance, and they can take some actions to avoid an accident. The characteristic behavior is the non-construction, and furthermore, due to the change of sunlight, temperature, and weather in the outside environment, and the shaking of cameras themselves, the background images will change as time goes by, which will bring special difficulties in recognizing jaywalking behaviors. In this paper, the method of adaptive background model of mixture Gaussian is used to extract the moving objects in the video. On the base of Histograms of Oriented Gradients (HOG), the pedestrians images and car images from MIT Library are used to train our monitoring system by SVM classifier, and identify the pedestrians in the video. Then, the color histogram, position information and the movement of pedestrians are selected to track them. After that we can identify whether the pedestrians are running in the red lights or not, according to the transportation signals and allocated walking areas. The experiments are implemented to show that the proposed method is effective.

An Adaptive Moving Objects Detection Algorithm Based on Kernel Density Estimation

The detection of moving objects are important research area for video surveillance and other video processing applications. In this paper, we propose an adaptive approach modeling background and segmenting moving object with non-parametric kernel density estimation. Unlike previous approaches to object detection which detect objects by global threshold, we use a local threshold to reflect temporal persistence. With combined of global threshold and local thresholds, the proposed approach can handle scenes containing gradual illumination variations and noise and has no bootstrapping limitations. Experimental results on different types of videos demonstrate the utility and performance of the proposed approach.

Background Subtraction for Online Calibration of Baseline RSS in RF Sensing Networks

Radio frequency (RF) sensing networks are a class of wireless sensor networks (WSNs) which use RF signals to accomplish tasks such as passive device-free localization and tracking. The algorithms used for these tasks usually require access to measurements of baseline received signal strength (RSS) on each link. However, it is often impossible to collect this calibration data (measurements collected during an offline calibration period when the region of interest is empty of targets). We propose adapting background subtraction methods from the field of computer vision to estimate baseline RSS values from measurements taken while the system is online and obstructions may be present. This is done by forming an analogy between the intensity of a background pixel in an image and the baseline RSS value of a WSN link and then translating the concepts of temporal similarity, spatial similarity, and spatial ergodicity, which underlie specific background subtraction algorithms to WSNs. Using experimental data, we show that these techniques are capable of estimating baseline RSS values with enough accuracy that RF tomographic tracking can be carried out in a variety of different environments without the need for a calibration period.

Genomic insights into salt adaptation in a desert poplar

Despite the high economic and ecological importance of forests, our knowledge of the genomic evolution of trees under salt stress remains very limited. Here we report the genome sequence of the desert poplar, Populus euphratica, which exhibits high tolerance to salt stress. Its genome is very similar and collinear to that of the closely related mesophytic congener, P. trichocarpa. However, we find that several gene families likely to be involved in tolerance to salt stress contain significantly more gene copies within the P. euphratica lineage. Furthermore, genes showing evidence of positive selection are significantly enriched in functional categories related to salt stress. Some of these genes, and others within the same categories, are significantly upregulated under salt stress relative to their expression in another salt-sensitive poplar. Our results provide an important background for understanding tree adaptation to salt stress and facilitating the genetic improvement of cultivated poplars for saline soils.

Motion Target Detection of Birds Based on Adaptive Background Update Mechanism

In view of the natural environment variability and uncertainty makes observations of some birds behavior complex and difficult, this paper proposes a Surendra adaptive background updating algorithm based on global threshold, which can update background automatically according to the environmental changes, and then extract the birds target region of interested by background subtraction and morphological filtering. The experimental results show that the birds moving target detection of video-captured has a good extraction effect.

Light adaptation increases response latency of alpha ganglion cells via a threshold-like nonlinearity

Adaptation is an important process of sensory systems to adjust sensitivity to ensure the appropriate information encoding. Sensitivity and kinetics of retinal ganglion cell (RGC) responses have been studied extensively using a brief flash superimposed on different but steady backgrounds. However, it is still unclear if light adaptation exerts any effect on more complex response properties, such as response nonlinearity. In this study, we found that the latency of spike responses to a repeated flashing spot stimulation increased by 30ms in the mouse ON α RGCs (An ON-type RGC is excited when a spot is turned on in the center of its receptive field). A single dimming event preceding the test flash on a steady adapting background could also produce similar effect in increasing latency of light responses. A simple computational model with a linear transformation of the light stimulus and a threshold-like nonlinearity could account for the experimental data. Moreover, the strength of the measured nonlinearity and the response latency were affected by the duration of light adaptation. The possible biological processes underlying this nonlinearity were explored. Voltage clamp recording revealed the presence of the increase in latency and threshold-like nonlinearity in the excitatory input of RGCs. However, no comparable nonlinearity was observed in the light responses of the ON cone bipolar cells. We further excluded GABAergic and glycinergic inhibition, N-methyl-d-aspartate receptor rectification and voltage-gated Na+ channels as potential sources of this nonlinearity by pharmacological experiments. Our results indicate the bipolar cell terminals as the potential site of nonlinearity. Computational modeling constrained by experimental data supports that conclusion and suggests the voltage-sensitive Ca++ channels and Ca++-dependent vesicle release in the bipolar cell terminals as mechanistic basis.

Background Modeling and Fuzzy Clustering for Motion Detection from Video

In this paper, for the modern intelligent video surveillance, we introduce an optimizing motion detection algorithm aim at overcoming the flaw of conventional background subtraction algorithm. We combine adaptive background model in HSV color space with moving object segmentation based on fuzzy clustering to extract moving objects from frame. The adaptive background model is able to restoring the background due to the accurate description of the HSV color space, and then the moving object segmentation based on fuzzy clustering is used to distinguish the moving area and noise area by the adaptive selection of threshold. We also consider SIFT feature to improve the performance of motion detection algorithm. The experiment show that the algorithm alleviates the impairment of noise and time complexity of the motion detection algorithm is low.

Adaptive background model for non-static background subtraction by estimation of the color change ratio

Background modeling, a preliminary processing step for foreground detection, is a challenging task because of the complexity and variety of background regions and unexpected scenarios such as sudden illumination changes, waving trees, rippling water, etc. In this work, we develop a pixel-based background modeling method that uses a probabilistic approach by means of changing color sequences. This method uses two background models in tandem. The first model uses a static background, which is obtained via a probabilistic approach and is a standard from which the foreground is extracted. The second method uses an adaptive background, which is modeled by the degree of color change. This background functions as an additional standard from which the foreground is extracted and is appropriate for eliminating non-static background elements. These models enable the developed method to automatically adapt to various environments. The algorithm was tested on various video sequences and its performance was evaluated by comparison with other state-of-the-art background subtraction methods.

Adaptive Background Modeling Algorithm Based on Objects Dynamicity

Detecting moving objects by using an adaptive background model is a critical component for many vision-based applications. In this paper, we propose a novel background modeling algorithm for traffic video surveillance based on objects dynamicity. Our algorithm is based on the definition of background as a set of static objects and on the observation that a static object is a set of pixels having the same appearance in long time interval. The normalized RGB color is used to describe the appearance under change lighting conditions, and a binary label is introduced to capture the dynamicity of objects over time. Finally, the cardinality function, calculated in each continued time period, is used to select persistent appearance corresponding to background objects. An update formula is proposed to reconstruct background model periodically based on current frame and last background image. Experimental results from several video sequences validate the effectiveness of the proposed algorithm

Object tracking by fusing multiple features based on adaptive background information

It is difficult for the object tracking algorithm based on single feature,to track the object in complex cases.Therefore,this paper proposed an algorithm fusing multiple features for object tracking based on adaptive background information.The algorithm was based on the use of color feature and gray level co-occurrence matrix texture feature to represent the object.Under the frame of particle filter,it analyzed the particle space distribution,particle value distribution and ability to distinguish the background information with different feature.Then it presented an efficient fusion coefficient calculation.According to the object's appearance of changes in the process of tracking,it updated the object template adaptively.The experimental results in different settings show that this algorithm greatly improves the resistance to background interference,under the premise of not reducing the real-time.In all sorts of situations,it has good stability and robustness.

Low false reject rate and false accept rate multi-step fire detection method

This paper proposes an efficient fire detection method for intelligent monitoring of self-service banks. This method includes the following steps: motion region detection, color-based fire detection and fire region refinement. These steps are as follows. First, the method uses four Gaussian distributions to construct an adaptive background model of the scene, and exploits this model to obtain the moving regions of the image. Second, the method uses RGB color and HSI color to determine whether the moving regions are fire candidates or not. These two steps are able to detect almost all of the true fire regions but it also takes a portion of non-fire regions as the fire. Third, the proposed frame difference procedure and the distance of two fire regions of two adjacent frames to eliminate the false fire regions obtained using the first two steps. The experimental results show that the proposed method performs very well in small-scale fire detection. The false reject rate (FRR) and false accept rate (FAR) of our method are 10.4%, and 6.02%, respectively.

Tracking-by-detection of multiple persons by a resample-move particle filter

Camera networks make an important component of modern complex perceptual systems with widespread applications spanning surveillance, human/machine interaction and healthcare. Smart cameras that can perform part of the perceptual data processing improve scalability in both processing power and network resources. Based on these insights, this paper presents a particle filter for multiple person tracking designed for an FPGA-based smart camera. We propose a new joint Markov Chain Monte Carlo-based particle filter (MCMC-PF) with short Markov chains, devoted to each individual particle, in order to sample the particle swarm in relevant regions of the high dimensional state-space with increased particle diversity. Finding an efficient sampling method has become another challenge when designing particle filters, especially for those devoted to more than two or three targets. A proposal distribution, combining diffusion dynamics, learned HOG + SVM person detections, and adaptive background mixture models, limits here the well-known burst in terms of particles and MCMC iterations. This informed proposal based on saliency maps has only been marginally used in the literature in a joint state space PF framework. The presented qualitative and quantitative results--for proprietary and public video datasets--clearly show that our tracker outperforms the well-known MCMC-PF in terms of (1) tracking performances, i.e. robustness and precision, and (2) parallelization capabilities as the MCMC-PF processes the particles sequentially.

Research of CamShift Algorithm to Track Motion Objects

At present, objects detection, identification and tracking are the very popular research areas in computing vision. A real-time objects tracking system was proposed in this paper. This paper focuses on improve CamShift algorithm, which can track the motion objects with high a accuracy. An Adaptive Gaussian Background Model is proposed in the system , which was established in order to automatically update the background and detect the outline of moving objects. By analyzing different algorithms, this paper brings out approaches to promote the performance.The CamShift algorithm to complete motion detection and objects tracking, which applied for static background video sequences. And the experimental results show that this algorithm can detect all motion objects and track almost motion objects. DOI: http://dx.doi.org/10.11591/telkomnika.v11i8.3037

Early Flame Detection by Using Dynamic Accumulated Model

In order to solve the problem of early fire detection in the large space, a dynamic accumulated model is presented. The motion object is firstly obtained by using adaptive background subtraction. Three combined rules for flame detection are utilized. Dynamic accumulated model is implemented to compute the edge flicker of the motion object. Two features including circularity and area change are utilized to distinguished flame from other moving object. Experimental results indicate that the presented method has quick respond and can be used in outdoor environment.

Object Detection Algorithm in Traffic Video Surveillance Based on Compressed Sensing

To adapt the contradiction between the increasing information quantity of highway traffic network monitoring and the limited network bandwidth resources, this paper proposes an object detection algorithm based on Bayesian compressed sensing. Video are sparse in a wavelet base, and a partial Hadamard measurement matrix is adopted to compress the video. The object detection method combines background difference and Bayesian compressed sensing of wavelet tree structure. To get more accurate foreground, an adaptive background model is proposed. Experiments results show the accuracy and effectiveness of the method, and can robustly detect the targets under changing light and reduce the price of video transmission.

Video Object Detection Algorithm Based on Gaussian Mixture Models

The moving object detection of the video image is the basis of sequence image analysis, and it is the research hot issue of today’s foreign and domestic scholars. For detecting the moving object from the scene image in time, a detection algorithm of video moving object based on Gaussian mixture models is proposed in the paper. The pixel values are seen as the combination of the foreground Gaussian distribution and the background Gaussian distribution, and the background estimation and the adaptive background update will be put up. The statistical number of the foreground pixel of the current frame determines whether the light has a larger change, and it combines with the frame-difference method to detect moving object. The experimental results show that the algorithm can quickly and accurately establish the background model and accurately segment the foreground object.

Detection of Abandoned and Stolen Objects in Videos using Mixture of Gaussians

approach to detect abandoned and stolen object in live videos based on background subtraction and foreground analysis has been presented in this paper. The motivation is to produce alerts spontaneously in the presence of abandoned or stolen objects. To detect the abandoned and stolen objects, the focus is to determine the static regions that have recently changed in the scene by performing background subtraction. The most commonly used algorithm for background subtraction is the Gaussian Mixture Model (GMM). In the proposed detection method, an improved Multi- Gaussian Adaptive background model is employed for background subtraction. A simple split and merge method is used to detect the static region from which the static objects are identified. The time and presence of static objects, which may be either abandoned or stolen, are informed to the security officials by triggering an alarm. The method proposed in this research work is tested to run in real time and have shown expected results.

Moving Objects Detection Based on Three Frames Differencing and GMM

The paper proposes a new method for moving objects detection based on fusion of three frames differencing and Gaussian Mixture Model (GMM). In the method, two images are obtained by three frames differencing, then the adaptive background are modeled and updated by GMM for each pixel in the two differencing images. Next, two differencing images are done logic "and" operation to get the shape of the moving object. Finally adopt the mathematical morphology operation to eliminate noise and the small areas of non-objects motion parts. The simulation results show that the proposed method can detect the objects effectively and real-time. So it can be applied in visual surveillance system effectively.

Inhibition to retinal rod bipolar cells is regulated by light levels

The retina responds to a wide range of light stimuli by adaptation of retinal signaling to background light intensity and the use of two different photoreceptors: rods that sense dim light and cones that sense bright light. Rods signal to rod bipolar cells that receive significant inhibition from amacrine cells in the dark, especially from a rod bipolar cell-activated GABAergic amacrine cell. This inhibition modulates the output of rod bipolar cells onto downstream neurons. However, it was not clear how the inhibition of rod bipolar cells changes when rod signaling is limited by an adapting background light and cone signaling becomes dominant. We found that both light-evoked and spontaneous rod bipolar cell inhibition significantly decrease with light adaptation. This suggests a global decrease in the activity of amacrine cells that provide input to rod bipolar cells with light adaptation. However, inhibition to rod bipolar cells is also limited by GABAergic connections between amacrine cells, which decrease GABAergic input to rod bipolar cells. When we removed this serial inhibition, the light-evoked inhibition to rod bipolar cells remained after light adaptation. These results suggest that decreased inhibition to rod bipolar cells after light adaptation is due to decreased rod pathway activity as well as an active increase in inhibition between amacrine cells. Together these serve to limit rod bipolar cell inhibition after light adaptation, when the rod pathway is inactive and modulation of the signal is not required. This suggests an efficiency mechanism in the retina to limit unnecessary signaling.