Year-2016
http://repository.iiitd.edu.in/xmlui/handle/123456789/395
2024-01-30T06:04:54ZREVERT: runtime verification for real-time systems
http://repository.iiitd.edu.in/xmlui/handle/123456789/524
REVERT: runtime verification for real-time systems
Kochanthara, Sangeeth; Purandare, Rahul (Advisor)
Real-time systems are becoming more complex and open, thus increasing their development and verification costs. Although several static verification tools have been proposed over the last decades, they suffer from scalability and precision problems. As a result, the tools fail to cover all the necessary safety properties for realistic real-time applications involving a large number of components and tasks. Runtime verification is a formal technique that verifies properties during system execution with the support of monitors. The monitors are generated from formal languages using correct-by-construction generation methods. Runtime verification can thus be used as a complement or replacement for static verification approaches. The current state-of-the-art tools either do not have notion of time, or suffer from the potential blowup of states at run-time. This thesis proposes Revert, a framework developed with a focus on the verification of functional and non-functional properties with timing constraints. The contribution of this work is threefold: (i) a domain-specific specification language allowing the definition of requirements for real-time applications; (ii) a novel mechanism to generate monitors, with state-space and time guarantees, capable of identifying and reacting to timing properties defined with the proposed specification language. (iii) a tool that automatically transforms specifications written in Revert to monitors specified as complete timed deterministic finite automata in xml format.
2016-11-01T00:00:00ZMining top-K high utility itemsets in streaming data: a comparative study
http://repository.iiitd.edu.in/xmlui/handle/123456789/523
Mining top-K high utility itemsets in streaming data: a comparative study
Sharma, Veronica; Goyal, Vikram (Advisor)
High Utility Itemset Mining (HUIM) has gained significant progress in recent years.
The HUIM refers to the method of finding most relevant itemsets from a database and it finds its applications in the domain of senosor data analytics, ad-click data analytics and retail stores. The HUIM allows to associate notion of utility with each item which was not possible in the case of frequent pattern mining (FPM). In FPM, only presence or absence of an item is considered in a transaction itemset and hence the approach does not allow provide exibility when different items have different importance. The focus of pattern mining work has been limited to mainly static databases. However, with the increase in data and need of timely information requires existing methods to be either scaled to or adapted to the streaming environment. The key concerns for streaming data are high throughput computations with minimum time and space constraints. In this thesis, we implemented and compared the top-k streaming version of state-of-the- art algorithms T-HUDS(High Utility Itemset Mining over Data Stream), FHM(Faster High Utility Itemset Mining),EFIM(Efficient High Utility Itemset Mining) on various databases. Our experimental results show that Stream-FHM and Stream-EFIM out- performs tree based T-HUDS algorithm. The Stream-FHM results are better for sparse databases and Stream-EFIM results better for dense databases.
2016-10-25T00:00:00ZAnalysing space utilization using indoor localization
http://repository.iiitd.edu.in/xmlui/handle/123456789/520
Analysing space utilization using indoor localization
Aneja, Joy; Singh, Pushpendra (Advisor)
The wide deployment of IEEE 802.11 Wireless LAN in offices, colleges and other organizations makes it a prime solution for indoor localization. Wireless access point awareness and its small cell size helps in localizing network devices. It is a cost efficient solution as it works with the existing infrastructure and does not require any specialized hardware deployments. Wide deployment of indoor localization and without any direct involvement of users enables it for various applications and services like space utilization, power conservation etc. This work analyzes SNMP traps sent by the access points to the NMS and determines the symbolic location of the network devices and exposes this real-time information through a robust, exible and loosely coupled platform. This platform is enabled to analyze space utilization which is important in higher education institutions as a balance has to be maintained between minimizing costs and meeting pedagogical and research needs of professors, and the learning and support needs of students. Further, this work also deploys some applications like personalized timelines, time spent and others.
2016-01-01T00:00:00ZRGB-D face recognition in surveillance videos
http://repository.iiitd.edu.in/xmlui/handle/123456789/440
RGB-D face recognition in surveillance videos
Chowdhury, Anurag; Vatsa, Mayank (Advisor)
Singh, Richa (Advisor)
Biometric analysis of surveillance videos carries inherent challenges in form of variations in pose, distance, illumination and expression. To address these variations, different methodologies are proposed, including utilizing temporal and 3D information. With the introduction of consumer level depth capturing devices such as Microsoft Kinect, research has been performed in utilizing low cost RGB-D depth data for characterizing and matching faces.
Face detection being the foremost task in face biometric pipeline has a cascading effect on the performance of any face recognition system that follows. Face detection algorithms generally work best for frontal face images with good illumination and low standoff distance. Developing a face detection system robust to the variates of a surveillance scenario is a highly challenging task. Recognition of the detected faces in surveillance scenarios is a challenging task owing to high variance in pose, illumination, expression and resolution. Also, the quality of depth data in RGB-D videos deteriorates with increase in standoff distance, thus adding to the challenges of RGB-D face recognition.
This research introduces the KaspAROV RGB-D video face database which provides face videos and images from Kinect device for over 100 subjects. The database encompasses challenges such as pose, distance, and illumination. Further, a novel face detection system for RGB-D videos taken in unconstrained scenario is proposed. The proposed system makes use of human body detection in color images and fuses it with the corresponding depth map to provide a robust solution for face detection at a distance in RGB-D videos. For recognizing the detected faces we introduce a RGB-D face recognition algorithm which can also work with only RGB probe images in absence of depth data in probe images. The proposed algorithm generates a shared representation from RGB images which contains discriminative information from both the RGB and depth images. This representation is much more discriminative than the RGB images as it gives substantially higher identification accuracy than a conventional fusion based RGB-D recognition pipeline.
2016-09-20T09:27:34Z