No abstract

No abstract

Adaptive background updating is one of the methods used to detect moving objects in video sequences. Many techniques have been presented in this field but there are few mentions about the usage of these methods in real-time applications. We concentrate in the speed of the algorithm and present a method that is fast enough to be used in video surveillance systems. We started from the ideas presented by Gaussian distribution for background generation. Instead of using actively all the pixels in the image we divide the pixels into active and inactive ones. Gaussian distributions are used to model the history of active pixels and to state whether they belong to background or foreground. According to the classification of the previous active pixel also the inactive pixels are classified as a part of the background or foreground. We also reduce the frame frequency and use only every n th frame in the image sequence to construct adaptive background. This article is organised as follows: In Chapter 1 some of the previous work and their results are introduced. In Chapter 2 we first describe the method used by Stauffer and Grimson [3] and then present our new ideas. The results are explained in Chapter 3 and finally a conclusion is given.

H.264 is the video standard needed for new high band requests of multimedia services such as video on demand, videoconference or videosurveillance. These applications require high video quality with the lowest possible band occupation. The new video coding standard H.264 seems to be answer to fulfill these requirements.Anyway, while fulfilling the requested quality-bandwidth ratio, the H.264 encoder complexity is largely increased if compared with previous standards, mainly because of the high computational cost of the Motion Estimation (ME) module that is the core of this standard video encoder.In this paper, we propose an enhancement to our innovative algorithm for a further complexity reduction of the ME module. The main idea is to make dynamically modifiable a previously static H.264 coder parameter: the size of the search window in ME. A Faster Motion Detection (MD) module, pixel precise and without background estimation, has been added to the video coder in order to compute the search window size. Moreover this method is valid for every kind of video stream and not only for sequences having a static background, thanks to the pixel precision of the motion detection.In the proposed solution, the reduction of the number of calculated SAD (Sum-of-Absolute-Differences) allows to decrease the encoder complexity both in low and high complex sequences. This paper illustrates the behavior of the Search Window Estimation (SWE) algorithm in both the old and the new algorithm version. The obtained results are then compared with jm86 full search and fast ME using different quantization values.

In this work, we address the problem of estimating the so-called "Social Distancing" given a single uncalibrated image in unconstrained scenarios. Our approach proposes a semi-automatic solution to approximate the homography matrix between the scene ground and image plane. With the estimated homography, we then leverage an off-the-shelf pose detector to detect body poses on the image and to reason upon their inter-personal distances using the length of their body-parts. Inter-personal distances are further locally inspected to detect possible violations of the social distancing rules. We validate our proposed method quantitatively and qualitatively against baselines on public domain datasets for which we provided groundtruth on interpersonal distances. Besides, we demonstrate the application of our method deployed in a real testing scenario where statistics on the inter-personal distances are currently used to improve the safety in a critical environment.