Moving object detection and tracking using Multiple Webcam

Download Full-Text PDF Cite this Publication

Text Only Version

Moving object detection and tracking using Multiple Webcam

Dr. Anil R. Karwankar

Department of electronics Government College of Engineering Aurangabad.

Aurangabad, India.

Tushar N. Darote

Department of electronics Government College of Engineering Aurangabad.

Aurangabad, India.

Abstract Real time object detection and tracking is an important task in various surveillance applications. Nowadays surveillance systems are very common in offices, ATM centers, shopping malls etc. In this paper, an Automated Video Surveillance system is presented. The system aims at tracking an object in motion and identifies an object in multiple webcam which would increase the area of tracking. The system employs a novel combination of Gaussian Mixture Model based Adaptive Background Modeling Algorithm and a RGB color model used for identifying an object in multiple webcam.

Keywords Moving object detection and tracking; Gaussian Mixture Model; Adaptive Background Modeling; color as a feature; video surveillance.

  1. INTRODUCTION

    Detection, tracking and identifying people in real time videos have become more and more important in the field of computer vision research. It has many applications, such as video based surveillance and humancomputer interaction. Its aim is to locate targets, retrieve their trajectories, and maintain their identities through a video sequence.

    1. Related Work

      In order to solve the challenging problem of human tracking and detection, a huge number of studies are already done related to tracking and detection of moving object. In

      [1] Adaptive Gaussian mixtures have been used for modeling non stationary temporal distributions of pixels in video surveillance applications. Significant improvements are shown on both synthetic and real video data. Incorporating this algorithm into a statistical framework for background subtraction leads to an improved segmentation performance compared to a standard method. In [2] a method employs a region-based approach by processing two foregrounds resulted from gradient and color-based background subtraction methods. In [3] system employs a novel combination of an Adaptive Background Modeling Algorithm (based on the Gaussian Mixture Model) and a Human Detection for Surveillance (HDS) System. The HDS system incorporates a Histogram of Oriented Gradients based human detector which is well known for its performance in detecting humans in still images. In [4], it uses modeling each pixel as a mixture of Gaussians and using an on-line approximation to update the model. The Gaussian

      distributions of the adaptive mixture model are then evaluated to determine which are most likely to result from a background process. Each pixel is classified based on whether the Gaussian distribution which represents it most effective part of the background model. In [5] it uses a method for human tracking using a stereo camera system called Subtraction Stereo and color information. The tracking system using the subtraction stereo, which focuses its stereo matching algorithm to regions obtained by background subtraction, is realized using Kalman filter. To make the tracking system more robust, the new method also uses color information as distinctive information of person.

    2. Our Contribution

      In this paper we are implementing a technique which can continuously track and identify multiple people from one or many place using multiple webcam by using adaptive background subtraction with Gaussian mixture model technique, handoff the color as feature to identify the person in many webcam.

    3. Outline

    In sec. I we describe about introduction, related work and our contribution to the system. In sec. II we introduce overview of system, frame differencing, Gaussian Mixture Model overview and identification of moving object. Subsequently, we provide performance analysis in sec. III. Finally, in sec. IV we conclude.

  2. SYSTEM MODELLING

    1. Overview of system:

      Fig. 1. Flow chart of generalised object detection and tracking.

      The flow chart shows how an image taken from live webcam is processed for tracking and identification purpose. The first process is detection of moving object for that

      Gaussian mixture model is used. Second process is to track a moving object for that blob analysis is used, third process is identification of moving object in multiple webcam for that color feature is used and hand off of the feature is done between the webcam.

    2. Frame Differencing

      A statistical background image of the video scene is obtained. This background image is subtracted from the current frame image and threshold. The foreground regions of interest are extracted from the threshold image after appropriate morphological operations. The algorithm flow for Static Background Subtraction is depicted in Fig. 2.

      Fig. 2. Static Background Subtraction

      The Static Background Subtraction system is not resilient to illumination changes or long lasting changes in the scene. Hence an Adaptive Background Modeling scheme should be adopted. In the following discussion, an implementation of the Gaussian Mixture model algorithm is presented, originally formulated by Stauffer et al [4], and subsequently modified by Harville et al [10].

    3. Gaussian Mixture Model Overview:

      The following algorithm models each individual pixel as a mixture of K-3D Gaussian distributions in the color space.

      Pixel Value

      (1)

      (4)

      (5)

      Where is taken as a learning constant. The distributions are sorted according to the . The first B distributions are chosen from the sorting to represent the background according to the following criteria:

      (6)

      The new pixel value is classified as a foreground pixel if no match is found amongst the B distributions. The least weighted distribution is replaced with the distribution corresponding to the new pixel value.

      Preprocessing of an image

      Morphological operations are done by using structuring element square of matrix 5×5 for smoothing the image.

      bwareaopen(binary image,P)

      Removes from a binary image all connected components (objects) that have fewer than P pixels, producing another binary image this operation is known as an area opening.

    4. Identification

      BlobAnalysis : The BlobAnalysis object computes statistics for connected regions in a binary image

      BBOX: computes the bounding box BBOX of the blobs found in input binary image.

      centroid: gives the co-ordinates of the moving object the coordinates are in the form of M-by-4 matrix of [x y width height] bounding box coordinates, where M represents the number of blobs and [x y] represents the upper left corner of the bounding box.

      The probability of observing the current pixel is

      (2)

      Where, K is the number of distribution, is estimate of the weight at Gaussian in the mixture at time t is the Gaussian probability density function.

      (3)

      Where,

      Fig. 3. Bounding box as green color, green is the color tag as an identity

      mark.

      Extraction region of interest from moving object in this project people are being detected by using color information. So our region of interest is a small mask located at central part of moving object. In that we are extracting its color information. By using centroid information we are able to

      locate the central part of moving object extracting 10×10 matrix color information.

      Region of interest:

      Fig. 4. 10×10 color mask from the Centre of bounding box

      A color tag i inserted to the left upper corner of the bounding box as an identifying mark to that moving object.

      color tag is given as in webcam1. In multiple webcam an multiple person can be tracked using this method.

  3. PERFORMANCE ANALYSIS

    The figure below shows tracking of one person in webcam 2 where the green color rectangle box i.e. bounding box is used to track the moving object. Left upper corner shows the count and the green color tag is inserted at the upper left of the bounding box as an identification mark.

    Fig. 6. one person tracking in webcam2

    Same person is tracked in webcam 1 since both the webcam are placed close to each other and it can be observed that same person are having same color tag that is green color.

    Fig. 5. Flow chart of object detection and tracking using four Webcam

    .

    In this system four webcam are used for identifying and

    tracking two people. A person is entered first in webacm1, this person is identified by frame differencing and Gaussian mixture model, and bounding box is made around the foreground object which gives the co-ordinates of the moving person. For extracting the color mask we take an 10×10 matrix from the center of the bounding box an extracting the mask assume this person is tag as red color in upper left side as an mark of identity. For second person we may tag a green color as an identity mark. The mask are continuously compared with the foreground object and if the mean of difference of two mask is less than threshold then the same

    Fig. 7. tracking of one person in webcam1

    Web

    Mask

    Mask of

    Mask of

    Mea

    Me

    of

    figure 8

    figure 8

    n of

    an

    figure

    (X2)

    (X3)

    diffe

    of

    cam1

    8(X1)

    rence

    diff

    X1 &

    ere

    X2

    nce

    X1

    &

    X3

    33.00

    5.53

    83

    72

    Web

    Mask

    Mask of figure 9 (y2)

    Mask of figure 9 (y3)

    Mea

    Me

    of

    n of

    an

    figure

    diffe

    of

    cam2

    9

    rence

    diff

    (y1)

    Y1 &

    ere

    Y2

    nce

    Y1

    &

    Y3

    39.50

    2.66

    41

    12

    Web

    Mask

    Mask of

    Mask of

    Mea

    Me

    of

    figure 8

    figure 8

    n of

    an

    figure

    (X2)

    (X3)

    diffe

    of

    cam1

    8(X1)

    rence

    diff

    X1 &

    ere

    X2

    nce

    X1

    &

    X3

    33.00

    5.53

    83

    72

    Web

    Mask

    Mask of figure 9 (y2)

    Mask of figure 9 (y3)

    Mea

    Me

    of

    n of

    an

    figure

    diffe

    of

    cam2

    9

    rence

    diff

    (y1)

    Y1 &

    ere

    Y2

    nce

    Y1

    &

    Y3

    39.50

    2.66

    41

    12

    Fig. 8. two person tracking in webcam1

    Fig. 9. two person tracking in webcam2

    The above figure shows two person are been tracked in two webcam continuously. In first webcam the color mask are crop and are compared with the second webcam and if mean difference between the two mask is less than threshold (we set threshold as 7) then same color tag is inserted as a mark for identification.

    Fig. 10. comparison of color mask of two person

    The above table shows the comparison of color mask of two people. The mean difference between each mask is calculated. The mean difference between the color mask of same person in webcam 1 & 2 is less than 7, hence they got same color tag and of different person is greater than 7, and hence they got different color tag. This shows that our algorithm works perfectly for tracking of multiple objects in multiple webcam.

  4. CONCLUSION

The system presented in this paper for Moving object detection and tracking in four webcam are successfully performed in matlab2013. This system successfully tracks two people and identifies them in four webcam continuously by inserting tags for the respective person. Tracking is done by using Gaussian mixture model and for identification purpose we are used rgb color information as a feature.

REFRENCES

    1. Dar-Shyang Lee, Member, IEEE, Effective Gaussian Mixture Learning for Video Background Subtraction, IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 27, NO. 5, MAY 2005

    2. Mohammad Izadi and Parvaneh Saeedi, Robust Region-Based Background Subtraction and Shadow Removing Using Color and Gradient Information.

    3. Prithviraj Banerjee and Somnath Sengupta: Human Motion Detection and Tracking for Video Surveillance.

    4. Stauffer, C. and Grimson, W.E.L,Adaptive Background Mixture Models for Real-Time Tracking, Computer Vision and Pattern Recognition, IEEE Computer Society Conference on,Vol. 2 (06 August 1999), pp. 2246-252 Vol. 2

    5. Yuma Hoshikawa, Kenji Terabayashi, Kazunori Umeda, Human Tracking Using Subtraction Stereo and Color Information

    6. Beymer, D. and Konolige, K., Real-Time Tracking of Multiple People Using Continuous Detection, Proc. Of the 7th Int. Conf. on Computer Vision Frame-rate Workshop, 1999.

    7. Bhattacharyya, A., On a measure of divergence between two statistical populations defined by probability distributions, Bull. Calcutta Math. Soc., Vol.35, pp.99109, 1943.

    8. Schiele, B., et al., Visual People Detection-Different Models, Comparison and Discussion, Proc. of the IEEE ICRA2009 Workshop on People Detection and Tracking, pp.1-8, 2009.

    9. Umeda, K., et al., Subtraction Stereo -A Stereo Camera System That Focuses On Moving Regions -, Proc. Of SPIE-IS&T Electronic Imaging, Vol.7239 Three- DimensionalImaging Metrology, 723908, 2009.

    10. M. Harville, G. Gordon, and J. Woodfill, Foreground Segmentation using adaptive mixture models in color and depth, In Proceedings of the IEEE workshop on Detection and Recognition of Events in Video, 2001.

Leave a Reply

Your email address will not be published. Required fields are marked *