The Counting App, or How to Count Vehicles in 500 Hours of Video

Lessard, Adrien; Bélisle, François; Bilodeau, Guillaume-Alexandre; Saunier, Nicolas

doi:10.1109/cvprw.2016.198

Cited by 7 publications

(5 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Table 2 shows training errors ranging from 0% for Video 004 to 12% for Video 006 and generalization errors ranging from 9.5% to 19.2% for the same videos. Given the performance reported in the background section, the results presented in this paper are promising, automatically attaining the results of commercial applications on some sites and improving the results on the use of nonoptimized parameters and manual clustering presented by Lessard et al (24).…”

Section: Experimental Results Description Of Datamentioning

confidence: 61%

“…Feature grouping and trajectory clustering depend on multiple parameters, a subset of which is critical for TMCs. Previous studies showed that applying the same tracking parameters across a large data set (more than 500 h of video) delivers widely varying results (24). Many factors (camera setup, traffic conditions, etc.)…”

Section: Black-box Optimizationmentioning

confidence: 99%

“…This would be the only step the traffic engineer would undertake to calculate TMCs for newly collected data. A computer-assisted method was developed to reassign trajectories and easily adjust the results with obvious errors (24).…”

Section: Black-box Optimizationmentioning

confidence: 99%

See 2 more Smart Citations

Optimized Video Tracking for Automated Vehicle Turning Movement Counts

Bélisle¹,

Saunier

Bilodeau

et al. 2017

Transportation Research Record: Journal of the Transportation R

Self Cite

View full text Add to dashboard Cite

This paper proposes a new method for automatically counting vehicle turning movements based on video tracking, expanding on previous work on optimization of parameters for road user trajectory extraction and on automated trajectory clustering. The counting method is composed of three main steps: an automated tracker that extracts vehicle trajectories from video data, an automated trajectory clustering algorithm, and an optimization algorithm. The proposed method was applied to obtain turning movement counts in three typical traffic engineering case studies in Canada representing industry-type conditions. These exhibited varying levels of tracking difficulty, ranging from a single-lane off-ramp to a six-movement intersection with a stop and a right-turn channel. Because of a limitation of the data set, giving flows per movement and not per lane, all sites were chosen with a single lane per movement. The 3-h morning peak period was used in the case studies. The results show an average weighted generalization error of 12% for more than 3,700 vehicles automatically analyzed for more than 8 h of video, ranging from 9.5% to 19.5%. The generalization error is on average 8.6% (and as low as 6.0% per movement) for the 3,084 uninterrupted vehicles that are in plain view of the camera. This paper describes in detail the methodology used and discusses the factors that affect counting performance and how to improve counting accuracy in further research.

show abstract

Section: Experimental Results Description Of Datamentioning

confidence: 61%

Section: Black-box Optimizationmentioning

confidence: 99%

See 1 more Smart Citation

Optimized Video Tracking for Automated Vehicle Turning Movement Counts

Bélisle¹,

Saunier

Bilodeau

et al. 2017

Transportation Research Record: Journal of the Transportation R

Self Cite

View full text Add to dashboard Cite

show abstract

“…A popular application area is traffic surveillance because of its practical use for reducing traffic jam and for assessing the security of various road configurations. However, the existing intelligent transportation systems (ITS) exhibit a low performance when faced with problems like occlusions, illumination changes, motion blur and other environmental variations [12]. To address these problems, we propose a multiple object tracker based on a robust visual object tracker.…”

Section: Introductionmentioning

confidence: 99%

Multiple Object Tracking with Kernelized Correlation Filters in Urban Mixed Traffic

Yang

Bilodeau

2017

2017 14th Conference on Computer and Robot Vision (CRV)

Self Cite

View full text Add to dashboard Cite

Recently, the Kernelized Correlation Filters tracker (KCF) achieved competitive performance and robustness in visual object tracking. On the other hand, visual trackers are not typically used in multiple object tracking. In this paper, we investigate how a robust visual tracker like KCF can improve multiple object tracking. Since KCF is a fast tracker, many KCF can be used in parallel and still result in fast tracking. We built a multiple object tracking system based on KCF and background subtraction. Background subtraction is applied to extract moving objects and get their scale and size in combination with KCF outputs, while KCF is used for data association and to handle fragmentation and occlusion problems. As a result, KCF and background subtraction help each other to take tracking decision at every frame. Sometimes KCF outputs are the most trustworthy (e.g. during occlusion), while in some other cases, it is the background subtraction outputs. To validate the effectiveness of our system, the algorithm was tested on four urban traffic videos from a standard dataset. Results show that our method is competitive with state-of-the-art trackers even if we use a much simpler data association step.

show abstract

“…In fact, occlusions, illumination changes, motion blur and other environmental variations [12] are still challenging tasks in untrimmed public video footage. Therefore, in this paper, we propose an automatic yet efficient attention region localization approach through background subtraction.…”

mentioning

confidence: 99%

Deep anomaly detection through visual attention in surveillance videos

et al. 2020

View full text Add to dashboard Cite

This paper describes a method for learning anomaly behavior in the video by finding an attention region from spatiotemporal information, in contrast to the full-frame learning. In our proposed method, a robust background subtraction (BG) for extracting motion, indicating the location of attention regions is employed. The resulting regions are finally fed into a three-dimensional Convolutional Neural Network (3D CNN). Specifically, by taking advantage of C3D (Convolution 3-dimensional), to completely exploit spatiotemporal relation, a deep convolution network is developed to distinguish normal and anomalous events. Our system is trained and tested against a large-scale UCF-Crime anomaly dataset for validating its effectiveness. This dataset contains 1900 long and untrimmed real-world surveillance videos and splits into 950 anomaly events and 950 normal events, respectively. In total, there are approximately ~ 13 million frames are learned during the training and testing phase. As shown in the experiments section, in terms of accuracy, the proposed visual attention model can obtain 99.25 accuracies. From the industrial application point of view, the extraction of this attention region can assist the security officer on focusing on the corresponding anomaly region, instead of a wider, full-framed inspection.

show abstract

The Counting App, or How to Count Vehicles in 500 Hours of Video

Cited by 7 publications

References 9 publications

Optimized Video Tracking for Automated Vehicle Turning Movement Counts

Optimized Video Tracking for Automated Vehicle Turning Movement Counts

Multiple Object Tracking with Kernelized Correlation Filters in Urban Mixed Traffic

Deep anomaly detection through visual attention in surveillance videos

Contact Info

Product

Resources

About