Large-Scale Vehicle Detection, Indexing, and Search in Urban Surveillance Videos

Abstract

We present a novel approach for visual detection and attribute-based search of vehicles in crowded surveillance scenes. Large-scale processing is addressed along two dimensions: 1) largescale indexing, where hundreds of billions of events need to be archived per month to enable effective search and 2) learning vehicle detectors with large-scale feature selection, using a feature pool containing millions of feature descriptors. Our method for vehicle detection also explicitly models occlusions and multiple vehicle types (e.g., buses, trucks, SUVs, cars), while requiring very few manual labeling. It runs quite efficiently at an average of 66 Hz on a conventional laptop computer. Once a vehicle is detected and tracked over the video, fine-grained attributes are extracted and ingested into a database to allow future search queries such as "Show me all blue trucks larger than 7 ft. length traveling at high speed northbound last Saturday, from 2 pm to 5 pm". We perform a comprehensive quantitative analysis to validate our approach, showing its usefulness in realistic urban surveillance settings.

Low-Complexity Big Video Data Recording Algorithms for Urban Surveillance Systems

“…End 7. Record the video frames to HD 5 As we already know, a typical camera generates 5.4G bytes video data per day. Hence, with a commonly 25 inputs DVR, it will need 4050G bytes space to record the videos only for 30 days, which is a tremendous volume.…”

“…For example, an application for vehicle search in crowded urban surveillance videos was proposed [5]. Another work proposed a new visual object tracking algorithm using a Bayesian Kalman filter with simplified Gaussian mixture [6].…”

Learning Detectors from Large Datasets for Object Retrieval in Video Surveillance

“…The work of Feris et al [1], [5] also splits the training data into motionlet clusters to better deal with non-linearities in the dataset. Training in each cluster is done with largescale feature selection, but only few thousands of training examples are considered.…”

“…The images in this dataset were automatically collected from more than 50 surveillance cameras, using the technique proposed in [5]. The images contain significant variation in The appearance manifold of vehicle images under varying pose and lighting is complex and highly non-linear.…”

“…Different from other multimedia retrieval tasks, where a large number of object classes are involved, and few examples may be available for many of them, surveillance is mostly concerned about vehicles and people, fo which many training images can be easily obtained. In this paper we focus on vehicle detection, leveraging a dataset of around 1 million vehicle images, automatically collected from 50+ cameras, using the technique proposed in [5]. Figure 1 shows the architecture of our proposed system.…”