stochastic3

Stochastic Definition:

Artificial Intelligence

"Stochastic, from the Greek "stochos" or "goal", means of, relating to, or characterized by conjecture; conjectural; random.

 The antonym is astochastic."

http://en.wikipedia.org/wiki/Stochastic

http://en.wikipedia.org/wiki/Stochastic

astochastic: mit.edu

Artificial Intelligence Laboratory Cambridge. MA

The following is a quote:

"Video Surveillance of Interactions Yuri IvanovMIT Media Laboratory Cambridge, MA 02139yivanov@media.mit.edu

Chris StaufferMIT Artificial Intelligence Laboratory Cambridge, MA 02139stauffer@ai.mit.edu

Aaron BobickMIT Media Laboratory Cambridge, MA 02139bobick@media.mit.edu

W. E. L. GrimsonMIT Artificial Intelligence Laboratory Cambridge, MA 02139welg@ai.mit.edu

Abstract This paper describes an automatic surveillance system, which performs labeling of events and interactions in an outdoor environment. The system is designed to monitor activities in an open parking lot. It consists of three components - an adaptive tracker, an event generator, which maps object tracks onto a set of pre-determined discrete events,and a stochastic parser. The system performs segmentation

and labeling of surveillance video of a parking lot and identifies person-vehicle interactions, such as pick-up and drop-off. The system presented in this paper is developed jointly by MIT Media Lab and MIT Artificial Intelli-gence Lab.

1 Introduction

Research in visual surveillance is quickly approachingthe area of complex activities, which are framed by extended context. As more methods of identifying simple movementsbecome available, the importance of the contextual methods increases. In such approaches, activities and movements arenot only recognized at the moment of detection, but their interpretation and labeling is affected by the temporally ex-tended context in which the events take place (e.g., see [2]). For example, in our application, while observing objects of different classes, cars and people, the detection of the class of an object may be uncertain. However, if when participat-ing in an interaction the object behaves like a car, entering interactions with other objects in a way characteristic to acar, then belief about its class label is reinforced.

The monitoring system we describe in this paper is an ex-ample of an end-to-end implementation, which is adaptive to the physical features of the monitored environment and

exhibits certain contextual awareness. Adaptation to the en-vironment is achieved by a tracker based on Adaptive Background Mixture s ([18]). It robustly tracks separateobjects in environments with significant lighting variation, repetitive motions, and long-term scene changes. The track-ing sequences it produces are probabilistically classified and mapped into a set of pre-determined discrete events. Theseevents correspond to objects of different types performing different actions.

Contextual information in the system is propagated via astochastic parsing mechanism. Stochastic parsing handles

noisy and uncertain classifications of the primitive eventsby integrating them into a coherent interpretation. Parallel input streams and consistency checking allow for detectionof high level interactions between multiple objects. The system demonstrates integration of low level visual infor-mation with high level structural knowledge, which serve as contextual constraints. The system is capable of main-taining concurrent interpretations when multiple activities are taking place simultaneously. Furthermore, the systemallows for interpretation of activities involving multiple objects, such as interactions between cars and people during PICK-UP and DROP-OFF.The remainder of this paper is organized as follows: section 2 introduces the problem domain and gives a briefoverview of previous and ongoing research in building automated surveillance systems. Section 3 gives an overviewof the system and details of its implementation. The section describes the system components, a tracker (section 3.1), anevent generator (section 3.2) and the parser (section 3.3) presenting the level of detail necessary for general understand-ing 1. Results of the system working on the surveillance data are shown in section 4, which is followed by conclusions,

1Complete details on the tracker can be found in [11] and [18]. Parser is described in full in [2], [12] and [13].

presented in section 5. 2 Related Work

The work presented in this paper spans a broad rangefrom low level vision algorithms to high level techniques used for natural language understanding.

The tracking component of the system uses an adaptivebackground mixture , which is similar to that used

by Friedman and Russell ([10]). Their method attemptsto explicitly classify the pixel values into three separate, predetermined distributions corresponding to the color ofthe road, the shadows, and the cars. Unfortunately, it is not clear what behavior their system would exhibit for pixelswhich did not contain these three distributions.

Pfinder ([20]) uses a multi-class statistical for thetracked objects, but the background is a single Gaussian per pixel. After an initialization period without objects,the system reports good results indoors. Ridder et al. ([17]) ed each pixel with a Kalman filter which made theirsystem more robust to lighting changes in the scene. While this method does have a pixel-wise automatic threshold, itstill recovers slowly and does not handle bimodal backgrounds well.

Contextual labeling in our system is performed by astochastic parser, which is derived from that developed by

Stolcke in [19], as was previously described in [2]. We ex-tended standard Stochastic Context-Free Grammar (SCFG) parsing to include (1) uncertain input symbols, and (2) tem-poral interval primitives that need to be parsed in a temporally consistent manner. In order to allow for noisy input weused the robust grammar approach, similar to that of Aho and Peterson ([1]). We extended the SCFG parser to accepta multi-valued input strings which allow for correction of the substitution errors, which is similar to the work done byBunke([6]).

[8] is fully devoted to syntactic analysis of interactionsand cooperative deterministic processes. It is related to our solution, formulating problems similar to ours, which arecast in terms of cooperative grammatical systems. In contrast, we describe interactions by a single stochastic gram-mar and use a single parser in an attempt to avoid the computational complexity of the complete Cooperative Distributed(CD) Grammar Systems.

In the area of monitoring long term complex activities,Courtney ([7]) developed a system, which allows for detection activities in a closed environment. The activitiesinclude person leaving an object in a room, or taking it out of the room. Perhaps the most complete general solution isdescribed in Brill at al. ([5]), who are working on an Autonomous Video Surveillance system. Brand ([3]) showedthe results of detecting manipulations in video using a nonprobabilistic grammar. This technique is non-probabilistic"

Credit: file http://web.media.mit.edu/~yivanov/Papers/CVPRVS99/CVPRVS99.ps.