Dual-channel CNN for efficient abnormal behavior identification through crowd feature engineering

Keyphrases

• Abnormal Behavior Identification 100%
• Anomaly Detection Model 25%
• Behavior-based 25%
• Benchmark Dataset 25%
• Benchmarking Method 25%
• Collectiveness 25%
• Compound Descriptors 25%
• Compound Feature 25%
• Computational Cost 25%
• Crowd Abnormal Behaviors 25%
• Crowd Anomaly Detection 25%
• Crowd Behavior 25%
• Crowd Features 100%
• Cuboid 50%
• Detection Accuracy 25%
• Detection Efficiency 25%
• Dual-channel Convolutional Neural Network 50%
• Dynamic Variation 25%
• Feature Descriptor 100%
• Feature Engineering 100%
• Foreground Region 25%
• Gaussian Mixture Model 25%
• Image Descriptors 25%
• Interaction Force 25%
• Interaction Information 25%
• K-nearest 25%
• Live Feed 25%
• Live Streaming 50%
• Local Features 50%
• Mixed Image 25%
• Multi-channel CNN 100%
• Neighbour Analysis 25%
• Online Crowds 50%
• Online Identification 25%
• Region of Interest 25%
• Spatial Space 25%
• Spatio-temporal Features 75%
• Spatiotemporal 50%
• Temporal Universe 25%
• Time-based 25%
• Tracklet 50%

Computer Science

• Anomaly Detection 25%
• Benchmarking 25%
• Computational Cost 25%
• Convolutional Neural Network 100%
• Detection Accuracy 25%
• Experimental Result 25%
• feature descriptor 100%
• Feature Engineering 100%
• Gaussian Mixture Model 25%
• Information Interaction 25%
• local feature 50%
• Temporal Feature 75%