CC BY
36
MOVING VIDEO CAMERA VIGILANCE USING DBSCAN
Moving Video Camera Vigilance Using DBSCAN
Jagdatta Singh1, Gaurav Kumar Srivastava2, Himanshu Pandey3
1Research Scholar, BBDU, Lucknow [email protected] 2Assistant Professor, BBDU, Lucknow [email protected] 3Assistant Professor, BBDNIIT, Lucknow [email protected]
Abstract
The author is trying to develop a model for dynamic or moving video camera vigilance using Density Based Clustering and location sensors. The authors try to exploit the rich functionality exposed by the machine learning paradigm in which the stochastic environment to learn is depicted as a two dimensional graph where the position of an object can be given by its coordinates. The author uses DBSCAN algorithm along with sensor enabled test ground area that keeps the X and Y co-ordinates of the moving objects. The idea here is to capture continuous video of the densest cluster of objects moving together. One practical usage of such system is a wild landscape where groups of animals are moving together to some destination. There will be a somewhat unorganized haphazard movement but we intend to capture only those animals that are greater in number as a group and the camera should move picturing them. This can be achieved by the DBSCAN algorithm
1. Introduction
1.1 Clustering Techniques:
In Clustering we split the data into groups of similar objects. Each group is known as a cluster. The intra-cluster similarity is high while inter-cluster similarity index is low. It is a very important technique in data mining. Traditionally it is seen as part of unsupervised learning. Different types of clusters as reported in the literature [1], [2].
Well Separated Clusters: Every node in this type of cluster is much similar to every other node in the cluster, but different from any other node not in the cluster.
Centre-Based clusters: Every object in the cluster is more similar to the centre also called the centroid than to the centre of any other cluster. Contiguous clusters: A node in a cluster is nearest (or more alike) to one or more other nodes in the cluster as compared to any node that is not in the cluster.
Density based clusters: A cluster is a thick region of points, which is separated by according to the low-density regions, from other regions that is of high density Conceptual clusters: A conceptual cluster shares some common feature, or indicates a particular thought.
MOVING VIDEO CAMERA VIGILANCE USING DBSCAN
1.2. Use of Clustering and Methods
Clustering has wide applications in Image Processing, Document Classification, Pattern Recognition, Spatial Data Analysis, Economic Science and Cluster Web log data to discover similar web access patterns, etc. Various Methods of clustering have been reported in literature [3], [4], [5]: Partitioning method: In literature different Partitioning methods reported are: K-mean method [3], [4], K-Medoids method (PAM) [5], [6], Farthest First Traversal k-center (FFT) [7], [8], CLARA [9], CLARANS [10], Fuzzy K-Means [11], Fuzzy K-Modes [12], K-Modes [13], Squeezer [14], K-prototypes [15] and COOLCAT [16], etc.
Hierarchical Methods: Agglomerative Nesting (AGNES) [17], Divisive Analysis (DIANA)
[18], Clustering using Representatives (CURE), Balanced Iterative Reducing and Clustering using Hierarchies (BIRCH) are some of the hierarchical methods.
Grid Based: Some of the Grid based clustering methods are STING, Wave Cluster, CLIQUE
[19] and MAFIA [20].
Density Based Methods: Density based clustering methods include DBSCAN, GDBSCANS, OPTICS, DBCLASD and DENCLUE.
Model Based method: Model based methods are divided into two approaches: Statistical approach includes AutoClass method while Neural Network Approach includes Competitive learning and Self-organizing feature maps.
2. DBSCAN
(Density-Based Spatial Clustering of Applications with Noise) is a popular unsupervised learning method utilized in model building and machine learning algorithms. Before we go any further, we need to define what an "unsupervised" learning method is. Unsupervised learning methods are when there is no clear objective or outcome we are seeking to find. Instead, we are clustering the data together based on the similarity of observations.
DBSCAN is a clustering method that is used in machine learning to separate clusters of high density from clusters of low density. Given that DBSCAN is a density based clustering algorithm, it does a great job of seeking areas in the data that have a high density of observations, versus areas of the data that are not very dense with observations. DBSCAN can sort data into clusters of varying shapes as well, another strong advantage. DBSCAN works as such:
• Divides the dataset into n dimensions
• For each point in the dataset, DBSCAN forms an n dimensional shape around that data point, and then counts how many data points fall within that shape.
• DBSCAN counts this shape as a cluster. DBSCAN iteratively expands the cluster, by going through each individual point within the cluster, and counting the number of other data points nearby. Take the graphic shown in figure for an example:
3
Outlier
MinPts = 5 Fig.l
MOVING VIDEO CAMERA VIGILANCE USING DBSCAN
3. Location Detection and Tracking of Moving Targets
Many applications require information about an object's location for rescue, emergency and security purposes. The approaches that access an object's location are typically divided into two groups: active and passive localization. In the former approach, the object is associated with a mobile station (MS), such as a tag or device in a communication network. The object's location is determined by sharing data between the MS and the base stations (BSs). The Global Positioning System (GPS), cellular networks, Bluetooth and wireless sensor networks (WSNs) are used in active localization. In the latter approach, the object does not communicate with other devices. However, the object's location can be determined by using the reflected signal from the object. Radio detection and ranging (radar), sound navigation and ranging (sonar) and laser detection and ranging (LADAR) are the most common types of passive localization. These methods have both advantages and disadvantages.
However, GPS and long-range radar generate many errors during indoor localization and tracking. Cellular networks and WSNs are limited by their complicated controls and protocols. Sonar and LADAR signals are degraded by environmental interference. Therefore, ultra-wide band (UWB) radar has become an emerging technology that is appropriate for indoor localization and tracking. UWB radar has many advantages, such as a high spatial resolution, the ability to mitigate interference, through-the-wall visibility, a simple transceiver and a low cost.
4. Methodology
The 2 dimensional area, A, assumed square in shape is plotted having X and Y coordinates. A random number of moving objects, here assumed to be small robotic cars with constant movement are left in the aforementioned area. Since the area, A, assumed here is small, location tracing sensors are fitted on the boundary of A. A video camera, C, is also planted which is used to position on the selected target. The ideal position of the camera should be on top, middle of A [21].
The DBSCAN algorithm then determines the cluster of robotic cars with maximum density. The algorithm also returns the center of the cluster which is one of the robotic cars. All the cars have built in emitters that generate a specific signal when they are selected as the center of the densest cluster shown in figure 1, called the core. Once the car is selected as the center of the densest cluster, it emits a signal that is received by the location tracing sensors [22]. As soon as the sensors receive the signal, they generate the X and Y coordinates of the car that emitted the signal. The coordinates are fed to the camera and the movement of the cluster gets recorded. This process is continuous and if the cluster changes then the process is repeated for the new cluster, center of focus being the new selected center of the densest cluster. The moving camera continuously positions its lens on the moving densest cluster and if the density of the cluster reduces then the new densest cluster is located by the DBSCAN algorithm and the camera starts focusing on the new most densely populated cluster.
The system demonstrated above can also find its application in larger areas. As pointed out earlier that the same process with slight modification can be applied to traffic monitoring and even wild life for framing videos on moving animals in groups, etc. With the aforementioned process a traffic accumulation can be reported or even a traffic jam for the traffic controllers.
In order to find the coordinates in bigger areas, we need the geo-locations in the form of the X and Y coordinates of the moving objects and the video camera will be fed with the coordinates as broadcasted by the satellites instead of location tracing sensors [23], [24].
MOVING VIDEO CAMERA VIGILANCE USING DBSCAN
Fig. 2
5. Conclusion
In this paper the author has proposed a model to provide continuous moving camera recording for the most densely populated group of objects. Here, the author has used an unsupervised learning algorithm of the artificial intelligence, called DBSCAN to find out the most intensively crowded orientation of the objects under vigilance. The DBSCAN algorithm reports the densest point, called the core of a population. The crowd is depicted by robotic cars having a facility to emit radio signals. Once a robotic car is selected as the core, it emits radio signals. This signal is received by the sensor installed for this purpose. This sensor calculates the X and Y coordinates of the core robotic car and sends them to the positioning system of the camera. With coordinates at hand, the camera focuses its lens on the selected X and Y coordinates. In this manner, the automatic moving camera is able to keep track of the core. With time, the core is changed and so is the camera's focus. It focuses on the new car selected as the core. This installation facilitates a system where the camera always focuses on the densest part of the moving objects. As a future research, this concept can be applied in controlling the traffic, where the radio signals can be replaced by the geo-location finders.
References
[1] K.Kameshwanran, & Malarvizhi, K. Survey on Clustering Techniques in Data Mining. International Journal of Computer Science and Information Technologies , 5 .2 (2014), 22722276.
[2] Junaid, S., & Bhosle, K. Overview of Clustering Techniques. International Journal of Advanced Research in Computer Science and Software Engineering , 4.11 (2014), 621-624.
[3] Wagstaff, Kiri, et al. "Constrained k-means clustering with background knowledge." Icml. Vol. 1. 2001.
[4] Hartigan, John A., and Manchek A. Wong. "Algorithm AS 136: A k-means clustering algorithm." Journal of the Royal Statistical Society. Series C (Applied Statistics) 28.1 (1979): 100-108.
[5] Park, Hae-Sang, and Chi-Hyuck Jun. "A simple and fast algorithm for K-medoids clustering." Expert systems with applications 36.2 (2009): 3336-3341.
J. Singh, G.K. Srivastava, H. Pandey RT&A, No 2 (53) MOVING VIDEO CAMERA VIGILANCE USING DBSCAN_Volume 14, June 2019
[6] Velmurugan, T., and T. Santhanam. "Computational complexity between K-means and K-medoids clustering algorithms for normal and uniform distributions of data points." Journal of computer science 6.3 (2010): 363.
[7] Panda, Mrutyunjaya, and Manas Ranjan Patra. "A Hybrid clustering approach for network intrusion detection using cobweb and FFT." Journal of Intelligent systems 18.3 (2009): 229246.
[8] Matsuoka, Hidehiro, and Hiroki Shoki. "Comparison of pre-FFT and post-FFT processing adaptive arrays for OFDM systems in the presence of co-channel interference." 14th IEEE Proceedings on Personal, Indoor and Mobile Radio Communications, 2003. PIMRC 2003. . Vol. 2. IEEE, 2003.
[9] Ng, Raymond T., and Jiawei Han. "E dent and E ective Clustering Methods for Spatial Data Mining." Proceedings of VLDB. 1994.
[10] Ng, Raymond T., and Jiawei Han. "CLARANS: A method for clustering objects for spatial data mining." IEEE Transactions on Knowledge & Data Engineering 5 (2002): 1003-1016.
[11] Gasch, Audrey P., and Michael B. Eisen. "Exploring the conditional coregulation of yeast gene expression through fuzzy k-means clustering." Genome biology 3.11 (2002): research0059-1.
[12] Huang, Zhexue, and Michael K. Ng. "A fuzzy k-modes algorithm for clustering categorical data." IEEE Transactions on Fuzzy Systems 7.4 (1999): 446-452.
[13] Chaturvedi, Anil, Paul E. Green, and J. Douglas Caroll. "K-modes clustering." Journal of classification 18.1 (2001): 35-55.
[14] He, Zengyou, Xiaofei Xu, and Shengchun Deng. "Squeezer: an efficient algorithm for clustering categorical data." Journal of Computer Science and Technology 17.5 (2002): 611624.
[15] Huang, Zhexue. "Extensions to the k-means algorithm for clustering large data sets with categorical values." Data mining and knowledge discovery 2.3 (1998): 283-304.
[16] Barbará, Daniel, Yi Li, and Julia Couto. "COOLCAT: an entropy-based algorithm for categorical clustering." Proceedings of the eleventh international conference on Information and knowledge management. ACM, 2002.
[17] Struyf, Anja, Mia Hubert, and Peter Rousseeuw. "Clustering in an object-oriented environment." Journal of Statistical Software 1.4 (1997): 1-30.
[18] Datta, Susmita, and Somnath Datta. "Comparisons and validation of statistical clustering techniques for microarray gene expression data." Bioinformatics 19.4 (2003): 459-466.
[19] De Amorim, Saul G., Jean-Pierre Barthélemy, and Celso C. Ribeiro. "Clustering and clique partitioning: simulated annealing and tabu search approaches." Journal of Classification 9.1 (1992): 17-41.
[20] Goil, Sanjay, Harsha Nagesh, and Alok Choudhary. "MAFIA: Efficient and scalable subspace clustering for very large data sets." Proceedings of the 5th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Vol. 443. ACM, 1999.
[21] Ahiska, Yavuz. "Multiple-view processing in wide-angle video camera." U.S. Patent No. 7,450,165. 11 Nov. 2008.
[22] Diener, Neil R., David S. Kloper, and Anthony T. Collins. "Server and multiple sensor system for monitoring activity in a shared radio frequency band." U.S. Patent No. 7,184,777. 27 Feb. 2007.
[23] Belisle, Timothy, and Thomas Belisle. "Geo-location system, method and apparatus." U.S. Patent Application No. 11/381,097.
[24] Williams, Darin Scot. "Car-finder method and aparatus." U.S. Patent Application No. 12/157,889.
[25]. Pandey, H., Kumar, S., & Darbari, M. (2018). Crowd Sourcing Rules in Agile Software Engineering to Improve Efficiency using Ontological Framework. Reliability: Theory & Applications, 13(2 (49)).
J. Singh, G.K. Srivastava, H. Pandey RT&A, No 2 (53)
MOVING VIDEO CAMERA VIGILANCE USING DBSCAN_Volume 14, June 2019
[26]. Pandey, H. (2018). Comparison of usage of crowdsourcing in traditional and agile software development methodologies on the basis of effectiveness. Reliability: Theory & Applications, 13(3 (50)).
[27] Singh. J and Pandey. H "Learning and Consensus in Multi-Agent Systems", International Journal of Research in Electronics and Computer Engineering (UGC approved Journal), Volume-6, Issue-4, PP. 1437-1442, ISSN: 2348-2281 (Online), Oct.-Dec. 2018.
[28] Pandey, H., & Singh, V. K. (2015). A fuzzy logic based recommender system for E-learning system with multi-agent framework. International Journal of Computer Applications, 122(17), 0975-8887.
[29] Pandey, H., Singh, V. K., & Verma, N. K. LR Rotation rule for creating Minimal NFA. International Journal of Applied Information Systems (IJAIS-ISSN: 2249-0868 Foundation of Computer Science FCS, New York, USA.
[30] Pandey, H., Singh, V. K., & Pandey, A. (2015). A new NFA reduction algorithm for state minimization problem. Int'l Journal of Applied Information System (IJAIS), 8(3), 27-30.
[31] Pandey. H, Darbari, M., Singh, V. K., & Kumar, G. (2014). Coalescence of Evolutionary Multi-Objective Decision Making approach and Genetic Programming for Selection of Software Quality Parameter. Complexity, 7(11).
