Algorithmic characterization of vehicle trajectories from image sequences by motion verbs

Abstract

Images of vehicles which move in traffic scenes recorded by a stationary camera are detected and tracked without operator intervention. The resulting vehicle trajectories are projected from the image plane onto the street plane. A suitable system internal representation of about ninety German motion verbs is then exploited in order to automatically characterize trajectory segments in terms of natural language concepts.A multiresolution approach for feature matching has been developed which is robust enough to track vehicle images across hundreds of frames despite considerable variations in size and projected velocity. Results from various experiments with image sequences from real world traffic scenes are presented.

"Moving Object Tracking of Vehicle Detection": A Concise Review

“…A new 3D model-based vehicle detection and depiction framework is based on a probabilistic boundary feature grouping, which is used for vehicle detection and tracking process [37].In this paper, the occlusion of vehicles detection process uses a 3D solid cuboid form with up to six vertices, and this cuboid is used to fit any different types and sizes of vehicle images by changing the vertices for a best fit. Therefore, vehicle detection, segmentation and tracking can be achieved efficiently due to changes in the region proportion, prototype width and height with consideration to previous images.…”

Front and Rear Vehicle Detection Using Hypothesis Generation and Verification

“…However, generating a displacement vector for each pixel is a time-consuming task and impractical for real time systems. To attack this problem, discrete methods employ the image features such as color blobs [24] or local intensity minima and maxima [25].…”

SAVE: A framework for semantic annotation of visual events

“…Approaches where representations do not take on semantic meaning include Stochastic Context Free Grammars (SCFG) [13], learning methods such as Bayesian Networks (BN) [7], and Hidden Markov Models (HMM) [2]. On the other hand, semantically significant approaches like the state machines [10], and PNF Networks [16] provide varying degrees of representation to the actions and agents involved in the events. Also, the Video Event Representation Language (VERL) was proposed in [14] where complex events are semantically represented with a hierarchical structure.…”