A decentralized privacy-sensitive video surveillance framework

Senst, Tobias; Eiselein, Volker; Badii, Atta; Einig, Mathieu; Keller, Ivo; Sikora, Thomas

doi:10.1109/icdsp.2013.6622765

Cited by 5 publications

(4 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…e back-end network is composed of three branches; each branch contains the dilation convolution with different expansion factors, and the expansion factors are 1, 2, and 4. e branch with expansion factor of 1 is used to capture the features of small-scale objects, while the other branches expand the perception range to capture the features of large-scale objects. As mentioned in literature [17], it is difficult for independent branches to learn the characteristics of different patterns, which leads to parameter redundancy. erefore, in this paper, the feature maps of each branch network are concatenated in each layer, and 1 × 1 convolution is used for cross-channel feature aggregation to strengthen the information interaction between each branch, so as to make full use of the complementarity of each branch extraction feature to make the output feature map has more expressive power and scale diversity.…”

Section: Generation Networkmentioning

confidence: 99%

“…In addition, for the ShanghaiTech and UCF_CC_50 datasets, this experiment uses the original image size for training, sets the batch size to 1, and performs data through random horizontal flips. Since the UCF-QNRF dataset is all high-resolution images (such as 9000 × 6000), this paper follows the training method proposed in [17] and crops the original image into 16 nonoverlapping subimages with a size of 224 × 224, and the batch size is 16 for training.…”

Section: Parameter Settingmentioning

confidence: 99%

“…In recent years, in order to improve the performance of crowd abnormal behavior detection caused by explosions, terrorist attacks, and other emergencies, many crowd abnormal behavior detection algorithms based on video sequences have been proposed. ese algorithms can be roughly divided into two categories: visual feature extraction method and physical feature analysis method [10][11][12][13][14][15][16][17][18][19]. e former uses vision and image processing technology to extract crowd features and then detect anomalies.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Crowd Counting and Abnormal Behavior Detection via Multiscale GAN Network Combined with Deep Optical Flow

Song¹,

Sheng²

2020

Mathematical Problems in Engineering

View full text Add to dashboard Cite

Aiming at the problem of low performance of crowd abnormal behavior detection caused by complex backgrounds and occlusions, this paper proposes a single-image crowd counting and abnormal behavior detection via multiscale GAN network. The proposed method firstly designed an embedded GAN module with a multibranch generator and a regional discriminator to initially generate crowd-density maps; and then our proposed multiscale GAN module is added to further strengthen the generalization ability of the model, which can effectively improve the accuracy and robustness of the prediction detection and counting. On the basis of single-image crowd counting, synthetic optical-flow feature descriptor is adopted to obtain the crowd motion trajectory, and the classification of abnormal behavior is finally implemented. The simulation results show that the proposed algorithm can significantly improve the accuracy and robustness of crowd counting and abnormal behavior detection in real complex scenarios compared with the existing mainstream algorithms, which is suitable for engineering applications.

show abstract

Section: Generation Networkmentioning

confidence: 99%

Section: Parameter Settingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Crowd Counting and Abnormal Behavior Detection via Multiscale GAN Network Combined with Deep Optical Flow

Song¹,

Sheng²

2020

Mathematical Problems in Engineering

View full text Add to dashboard Cite

show abstract

“…Senst et al [26] have developed architecture for providing security and privacy to support video operators which works in surveillance system. The advantage of proposed system is that it uses automated calibrated cameras it display detected events and object extraction in 3D view.…”

Section: Introductionmentioning

confidence: 99%

Simplified Video Surveillance Framework for Dynamic Object Detection under Challenging Environment

Chandra

M²

2019

IJECE

View full text Add to dashboard Cite

An effective video surveillance system is highly essential in order to ensure constructing better form of video analytics. Existing review of literatures pertaining to video analytics are found to directly implement algorithms on the top of the video file without much emphasis on following problems i.e. i) dynamic orientation of subject, ii)poor illumination condition, iii) identification and classification of subjects, and iv) faster response time. Therefore, the proposed system implements an analytical concept that uses depth-image of the video feed along with the original colored video feed to apply an algorithm for extracting significant information about the motion blob of the dynamic subjects. Implemented in MATLAB, the study outcome shows that it is capable of addressing all the above mentioned problems associated with existing research trends on video analytics by using a very simple and non-iterative process of implementation. The applicability of the proposed system in practical world is thereby proven.

show abstract