Each researcher is keep pursuit to find the ideal potion of a robust recognition and detection for video system. Thus, an Automated Video Surveillance system or ...
Available online at www.sciencedirect.com
Procedia Engineering
ProcediaProcedia Engineering 00 (2011) Engineering 29 000–000 (2012) 1259 – 1263 www.elsevier.com/locate/procedia
2012 International Workshop on Information and Electronic Engineering
Bandit Detection using Color Detection Method Junoh, A.K2, Mansor, N. 2, Muhamad, W. Z. A. W. 2, Abu, S. 1, Jaafar, W. N.W. 2,Yaacob, S. 2 1
Institut Matematik Kejuruteraan, University Malaysia Perlis, Jalan Sarawak, Taman Bukit Kubu Jaya, 02000, Kuala Perlis, Perlis, Malaysia. 2 Intelligent Signal Processing Group (ISP), University Malaysia Perlis, no 70 &71, Bolok B, Jalan Kangar- Alor setar, 02000, Kangar, Perlis, Malaysia.
Abstract The issue of the actual mechanism for the visual and computational perception of motion in the human are keep grow for the last decade. Each researcher is keep pursuit to find the ideal potion of a robust recognition and detection for video system. Thus, an Automated Video Surveillance system or “ANGSA”is presented in this paper. The “ANGSA” system aims to track an object in motion and classifying it as a human or non-human entity, which would help in subsequent human activity analysis. The system employs a simple method for Human Detection in Surveillance (HDS) System. The HDS system incorporates a color based human detector which is well known for its performance in detecting humans in stationary images. Detailed analysis is carried out on the performance of the system on various test videos. At the end of the study human face successfully to be detected accurately depends on the distance between monitoring camera and the stranger.
© 2011 Published by Elsevier Ltd. Open access under CC BY-NC-ND license. Keywords: Human detecting ; image processing ; security system
1. Introduction Surveillance systems have been existed for more than thousand years. Angsa is a malay word to represent a goose. Goose or “angsa” have been used as pet and also to alert the owner about the existence of the stranger in the house area. The concept of goose behavior is considered since the goose have the criteria that sensitive to suspicious things or strangers and giving an unpleasant sound to alert the owner
1877-7058 © 2011 Published by Elsevier Ltd. Open access under CC BY-NC-ND license. doi:10.1016/j.proeng.2012.01.123
21260
A.K name Junoh/ et al. / Procedia Engineering 29 (2012) 1259 – 1263 Author Procedia Engineering 00 (2011) 000–000
of the house with the stranger especially human. Thus, by adapting this consept of “ANGSA”, a surveliance system have been proposed in order to developed a safety system combining the vision and controller system. The issue of the actual mechanism for the visual and computational perception of motion in the human are kept grow for the last decade. Each reseacrher are keep pursuit to find the ideal algorithm of the robust recognition and detection of video system. However most of the system just able to record the scenario of the event in certain location [1], without further analysis. There are a lot researches have been done to detect the movement of the object in the consicutive frame. However, the objective only to detect the motion of the object in the frame image. Most of the camera can detect the movement of the object, however it still difficult to classify either the object is unliving object, human or animal [2]. 2. Methodology From the sequence images that recorded by camera. Each frame have same size pixel, however each pixel have different color pixel based on RGB (red, green, blue) [3]. However, each image that have been recorded are different in time, the color value are almost same, because the background didn’t change. However, if the objects are moved are relatively to time, the pixel value will change respectively. The system itself also able to recognize and adapt the background changes over time continuously. The moving entities are further classified into human and non-human categories using the color detection method [3]. A brief overview of the system is given in Fig. 1. The foreground is extracted from the video scene by learning a statistical model of the background, and subtracting it from the original frame. The background model learns only the stationary parts of the scene and ignores the moving foreground. The system uses the different color pixel based on RGB (red, green, blue) [3] for modeling the background adaptively. Hence, the motion regions are identified in the frame, which constitute the regions of interest (ROI) for the system. Image Acquisition No
Motion detected? Yes Face Detection? Yes Accuracy Test
Result Validation Fig. 1 System Block Diagram
No
A.K Junohname et al. //Procedia Procedia Engineering Engineering 00 29 (2011) (2012) 000–000 1259 – 1263 Author
2.1. Image Acquisition Acquiring raw data for research can be a long and tedious process. The use of a graphical user interface (GUI) can greatly simplify the process. We have developed a system that captures raw images with key parameters that can be used for statistical analysis. We found the easiest method to implement a capture system was using a device that produced a standard composite video output signal. The MD901 Migix webcam contains a high quality camera that automatically maintains focus on the subject is within a certain range as shown in Fig. 2. The camera used to acquire the images also uses a composite video output. Using a USB Connector we route the signal output of the camera computer that is used for data collection. The data collection computer configured with MATLAB R2009a and the Image Acquisition Toolbox.
Fig. 2 MD901 Migix Webcam
2.2. Motion Detected Based on this image processing method from sequence image, the moving object can be detected and at the same time can be identified. However a robust algorithm are required to recognise the type of the moving objcet either it is a human, animal or unliving things. Thus, a color detection is proposed to differentiaite the different color value between each object. Based on the color detection method the skin was detected in order to determine which type of object which detected in the frame. 2.3. Face Detection Based on YCrCb color space, where the advantages are chroma (CrCb) and luminance (Y) information is stored in different channels due to working process are involving with different lightning conditions. It seems that lighting information could easily lead to false detections, hence the luminance component need to be discarded. In YCbCr color space it is simply to be done by not using one of its channels according to [4].The result of Face Detection is first processed by a decision function based on the chroma components CrCb from YCrCb and Hue from HSV by [5] creating a Skin Map. If all following conditions are true for a pixel, it's marked as skin area. 140 < Cr < 165 140 < Cb < 195
1261 3
41262
A.K name Junoh/ et al. / Procedia Engineering 29 (2012) 1259 – 1263 Author Procedia Engineering 00 (2011) 000–000
0.01 < Hue < 0.1 The image in Fig. 3 below shows a possible result. Skin areas are marked with red box
Fig. 3 Face Detection
3. Result and Discussion
Average Accuracy (%)
Average Accuracy vs Different Distance
120 100 80 60 40 20 0 15
30
45
60
75
Different Distance (cm)
Fig.4 Average Accuracy vs. Different Distance
Fig. 4 shows average accuracy versus different distance (cm) characteristics; the average accuracy was weighted by distance (cm). These results represent the average accuracy for the whole test route. It is apparent that the highest levels of average accuracy are in the range 90% to 99.0% and there is decreasing
A.K Junohname et al.//Procedia Procedia Engineering Engineering 00 29 (2011) (2012) 000–000 1259 – 1263 Author
in accuracy levels between 15cm and 75cm. The average accuracy for all trial is 96%.The highest percentage accuracy is 99 % for 15cm. However at 75cm was the lowest percentage accuracy is 81.25 %. Besides, of the different distance result were received, only one distance (15cm) increased rapidly on distance measures values. The range of this phenomenon is from position 100% to 92.655. However, at 30cm till 60cm in did not show a higher value for the average accuracy (in either direction) for each distance taken. Hence, of the three distances for which accuracy result were received, three such as 30cm, 45cm and 60cm returned almost constant values of average accuracy at range of 93% to 97%. Conversely, for which an accuracy result was received (15cm) always returned higher values of and the highest distance give the lowest accuracy. 4. Conclusion We have proposed simple technical methods for detecting of existence of human in image processing based on color detection methods, by exploiting information coming from motion and color analysis in original images. The efficiency of this proposed approach is believed higher than 90%. Clearly, our methodology deserves further evaluation in control security vision based systems. Acknowledgment This research conducted under Fundamental Research Grant Scheme (FRGS) which contributed by Ministry of Higher Education Malaysia. References [1] T.C Robert, J.L Alan, and K. Takeo, “Introduction to the Special Edition on Video Surveillance”, Proceedings of the IEEE Transactions on Pattern Analysis and Machine Intelligence, Volume 22, No. 8, pp 745-757, August 2000. [2] N. Dalal and B. Triggs, “Histograms of oriented gradients for human.Detection”, Proceedings of the Conference on Computer Vision and Pattern Recognition, San Diego, California, USA, pp. 886–893, 2005. [3] 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. [4] Gong, Shaogang, Stephen J. McKenna, Alexandra Psarrou. Dynamic Vision: “From Images to Face Recognition,” Imperial College Press, (2000). [5] Ilias Maglogiannis, “Face detection and recognition of natural human emotion using Markov random fields”, In Personal and Ubiquitous Computing. Springer London, (2007).
1263 5
