Parallelizing Multimodal Background Modeling on a Low-Power Integrated GPU

Tracking and segmentation of moving objects are suffering from many problems including those caused by elimination changes, noise and shadows. A modified algorithm for the adaptive background model is proposed by linking Gaussian mixture model with the method of principal component analysis PCA. This approach utilizes the advantage of the PCA method in providing the projections that capture the most relevant pixels for segmentation within the background models. We report the update on both the parameters of the modified method and that of the Gaussian mixture model. The obtained results show the relatively outperform of the integrated method.

This paper proposes a moving target detection scheme suitable for camera motion. Firstly, the background model is initialized by a Gaussian mixture model algorithm. Then Kanade-Lucas-Tomasi Feature Tracker (KLT) method is used to detect optical flow feature points of two adjacent frames, RANdom SAmple Consensus (RANSAC) algorithm is used to filter out the correct matching points and obtain a homography matrix, which can recover the background model matching the current frame, the new background model is used to detect moving target of the current frame. In the foreground detection stage, the current pixel is first compared with its own background model, and then compared with the background model of its 8 neighborhood pixels, the algorithm is speeded up without reducing the detection accuracy in this way; In the update stage of the background model, in order to adapt to the background changes caused by camera motion, an age value variable is set for each pixel. The experimental results show that the improved algorithm has a significant improvement in detection accuracy and running time compared to Gaussian mixture background modeling.

Approximation-free running SVD and its application to motion detection

The discipline of computer vision is becoming more popular as a research subject. In a surveillance-based computer vision application, item identification and tracking are the core procedures. They consist of segmenting and tracking an object of interest from a sequence of video frames, and they are both performed using computer vision algorithms. In situations when the camera is fixed and the backdrop remains constant, it is possible to detect items in the background using more straightforward methods. Aerial surveillance, on the other hand, is characterized by the fact that the target, as well as the background and video camera, are all constantly moving. It is feasible to recognize targets in the video data captured by an unmanned aerial vehicle (UAV) using the mean shift tracking technique in combination with a deep convolutional neural network (DCNN). It is critical that the target detection algorithm maintains its accuracy even in the presence of changing lighting conditions, dynamic clutter, and changes in the scene environment. Even though there are several approaches for identifying moving objects in the video, background reduction is the one that is most often used. An adaptive background model is used to create a mean shift tracking technique, which is shown and implemented in this work. In this situation, the background model is provided and updated frame-by-frame, and therefore, the problem of occlusion is fully eliminated from the equation. The target tracking algorithm is fed the same video stream that was used for the target identification algorithm to work with. In MATLAB, the works are simulated, and their performance is evaluated using image-based and video-based metrics to establish how well they operate in the real world.

Although prostate cancer (PCa) is the most common non-cutaneous form of cancer among Western men, available diagnostic imaging methods are not yet sufficiently reliable to avoid systematic biopsy. In this work, we aim at improving the accuracy of transrectal dynamic contrast-enhanced ultrasonography (DCE-US) for PCa localization by combining local perfusion and dispersion parameters. To this end, ten of these parameters were extracted pixel-by-pixel from 45 DCE-US recordings distributed over 19 patients that were scheduled for radical prostatectomy. Based on 43 benign and 42 malignant histologically-confirmed regions of interest, we produced multiparametric maps using a Gaussian Mixture Model (GMM) algorithm. All possible combinations of one to four parameters were evaluated to select the most suitable subset of parameters. We also tested the GMM algorithm's ability to determine the classification confidence for each pixel and the impact of excluding low-confidence pixels from the images. An accuracy and negative predictive value of 81% and 83%, respectively, are obtained, which improved after pixel exclusion. Even though extended validation on a larger patient group is recommended, multiparametric DCE-US shows high potential in localizing PCa and might become an important tool for guiding targeted biopsy or planning of focal treatment.

Background modeling and subtraction is a basic component of many computer vision and video analysis applications. It has a critical impact on the performance of object tracking and activity analysis. In this paper, we propose an effective and adaptive background modeling and subtraction approach that can deal with dynamic scenes. The proposed approach uses a subspace learning method to model the background and the subspace is updated on-line with a sequential Karhunen-Loeve algorithm. A linear prediction model is also used to make the detection more robust. Experimental results show that the proposed approach is able to model the background and detect moving objects under various type of background scenarios with close to real-time performance.

In this study, the classification of the speakers according to age and gender was discussed. Age and gender classes were first examined separately, and then by combining these classes a classification with a total of 7 classes was made. Speech signals represented by Mel-Frequency Cepstral Coefficients (MFCC) and delta parameters were converted into Gaussian Mixture Model (GMM) mean supervectors and classified with a Support Vector Machine (SVM). While the GMM mean supervectors were formed according to the Maximum-a-posteriori (MAP) adaptive GMM-Universal Background Model (UBM) configuration, the number of components was changed from 16 to 512, and the optimum number of components was decided. Gender classification accuracy of the system developed using aGender dataset was measured as 99.02% for two classes and 92.58% for three classes and age group classification accuracy was measured as 67.03% for female and 63.79% for male. In the classification of age and gender classes together in one step, an accuracy of 61.46% was obtained. In the study, a two-level approach was proposed for classifying age and gender classes together. According to this approach, the speakers were first divided into three classes as child, male and female, then males and females were classified according to their age groups and thus a 7-class classification was realized. This two-level approach was increased the accuracy of the classification in all other cases except when 32-component GMMs were used. While the highest improvement of 2.45% was achieved with 64 component GMMs, an improvement of 0.79 was achieved with 256 component GMMs.

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

We propose a new method for background modeling. Our method is based on the two complementary approaches. One uses the probability density function(PDF) to approximate background model. The PDF is estimated non-parametrically by using Parzen density estimation. And foreground object is detected based on the estimated PDF. The other method is based on the evaluation of the local texture at pixel-level resolution while reducing the effects of variations in lighting. Fusing their approach realize robust object detection under varying illumination. Several experiments show the effectiveness of our approach.

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.

This paper proposes a robust motion detection method for illumination variations which uses histogram of oriented gradients. The detection process is divided into two phases: coarse detection and refinement. In the coarse detection phase, first, a texture-based background model is built which implements a group of adaptive histograms of oriented gradients; then, by comparing the histogram of oriented gradients of each pixel between current frame and background model, a foreground is segmented based on texture feature which is not susceptible to illumination variations, whereas some missing foreground regions exist; finally, the result based on texture is optimized by combining the pixel-wise detection result produced by Gaussian Mixture Model (GMM) algorithm, which greatly improves the detection performance by incorporating efficient morphological operations. In the refinement phase, the above detection result is refined based on the distinction in color feature to eliminate errors like shadows, noises, redundant contour, etc. Experimental results show the effectiveness and robustness of our approach in detecting moving objects in varying illumination conditions.

Many background subtraction techniques have been proposed to detect changes such as moving objects in videos. In this paper, we classify them into two approaches according to forms of the regions detected as moving object. The first one is using frame difference and the second one is exploiting background model. We show characteristics of the two approaches and the relational properties among their products. Then we also show these properties can be used for estimating another result from one approach and identifying the region should be updated in adaptive background model.

This work proposes a novel method for detection of motion based object having dynamic scenario in the background. The suggested scheme has a strong potential for real-time applications especially for rafting, river, sea-beach, swimming pools, ponds, etc. Apart from these, this work is very beneficial for surveillance of border, tunnel, traffic in the sea, forest, restricted zones, deep zones, etc. This work develops a statistical p based background subtraction method and implemented in three stages. In the first stage, a background model is developed using few initial frames. In the second stage, this work classifies the foreground using the difference frame and the appropriate threshold value. An automatic threshold value is generated at run-time and updated iteratively. It also reduces the problem of using a constant threshold. In the third stage, morphological filters and connected component based region filtering technique is applied to enhance the detection quality. The extensive experimental result shows more accurate results of proposed method. It also demonstrates better performance against considered state-of-the-art methods.

Background modelling is a fundamental task in video analytics. This paper presents an adaptive background modelling for real-time indoor and outdoor videos. This proposed method treats background modelling as an optimisation problem and, it fetches multiple peaks from the histogram of the video frame and optimises them using a multi-swarm technique. The background is successfully adapted whenever there is a change in the environment as well as a number of objects in the background. The experimental result of foreground extraction confirms the effectiveness and robustness of the proposed background modelling technique against the various background modelling approaches.

Gaussian-input Gaussian mixture model for representing density maps and atomic models.