Vision Surveillance PAUL SCOVANNER Surveillance Main tasks Locating

Vision Surveillance PAUL SCOVANNER

Surveillance �Main tasks Locating people and objects in a scene � Background Subtraction � Object Detection Track objects as they move � Associate objects across frames Beyond Tracking

Background Subtraction �Remove the background leaving areas where movement occurs �Frame Differencing: |framet – framet-1| > Threshold

Background Subtraction �Frame Differencing Fast Simple Error prone � (Illumination changes, Edges on large objects, Amplifies sensor noise) �Background Modeling |framet – Background| > Threshold Model the colors of each pixel as a Gaussian � (mean and standard deviation)

Background Subtraction Mixture of Gaussians

Object Detection aka “is that a car or a person? ” �Aspect ratio �Object Detectors

Tracking �We can detect moving objects (If background subtraction works) �We can identify pedestrians and cars (If object detection works) �What’s left?

Tracking �Associate the detections in one frame with the next. Visual similarity Spatial location

Tracking

Multi-view Tracking �If 1 camera is good… 3 Must be better

Multi-view Tracking

Multi-view Tracking

Tracking From The Air

Tracking From The Air

Tracking From The Air

Tracking �Pedestrian Modeling Predict movements of pedestrians Prediction

Anomaly Detection �Detect emergency events

Anomaly Detection

More Than Just Tracking �Crowd instability