Video Surveillance



Surveillance is the monitoring of the behavior, activities, or other changing information, usually of people for the purpose of influencing, managing, directing, or protecting.Surveillance is very useful to governments and law enforcement to maintain social contral, recognize and monitor threats, and monitor threats, and prevent/investigate activity.

Surveillance researches performed in our laboratory can be divided into 3 categories


For outdoor video surveillance systems, the weather conditions like rain,snow,fog are one of the main factors that degrade the quality of videos, leading to degradation of the system performance.

• Purpose:

– Judge the weather condition and remove the effect of weather in video

• Syntax:

– Input : video which containing rain/snow/fog

– Output : rain/snow/fog effect removed image

• Assumption:

– Weather condition : rain, snow, fog

– Standard line exists for fog detection

– Distinct snow, rain particles

Deweather algorithm process is as follows

Deweather flow chart

• Approach:

– Weather condition detection

• Frame difference for extracting rain/snow candidate particles

• Intensity calculation for dividing off to rain and snow

– De-weathering of snow/rain

• Particle size detection

• Particle proportion calculation

• Rain effect removal ; Particle angle detection for non-rain particle removal

– De-weathering of fog

• Using dark channel prior

Raindrop Removal

Example of Raindrop Removal

Snow Removal

Related Papers:

Yu Miao, Hana Hong, Hakil Kim; “Size and Angle Filter Based Rain Removal in Video
for Outdoor Surveillance Systems”; ASCC 2011

우묘, 홍하나, 김학일;”악천후 속에서의 실외용 지능형 감시 시스템 구현을 위한 De-weathering”

영상처리및이해에관한워크샵 2011

Back to top

Fire and Smoke Detection

An automatic fire alarm system is designed to detect the unwanted presence of fire by monitoring environmental changes associated with combustion.Automatic fire alarm systems are intended to notify the building occupants to evacuate in the event of a fire or other emergency, report the event to an off-premises location in order to summon emergency services, and to prepare the structure and associated systems to control the spread of fire and smoke.

• Purpose:

– Detect fire and smoke in the environment

• Syntax:

– Input : Video image

– Output : Fire descriptor

• Center and ROI (rectangle) of the fire region

• Approach:

– Detect candidate fire region with

• Fire colored-pixel detection

• Moving flame pixel detection

• Temporal luminance variation

– Use probabilistic method to classify fire

• SVM: Support Vector Machines

– Detect smoke

• Use motion block and orientation

The result is as follows

Fire detection algorithm process is as follows

Fire detection algorithm

Smoke detection algorithm process is as follows

Smoke detection algorithm

Related Papers:

De-chang Wang, Hakil Kim, Chang-long Jin;”Adaptive Flame Detection using Randomness Testing and Robust Features”

De-chang Wang, Hakil Kim;”Night-Fire Detection using Randomness Test” 영상처리 및 이해에 관한 워크샵 2012

Dechang Wang, Hakil Kim;”Improved Fire Detection Using Mean-Shift-based Mode Detection” 영상처리 및 이해에 관한 워크샵 2011

Back to top

Biometrics surveillance

Biometric surveillance refers to technologies that measure and analyze human physical and/or behavioral characteristics for authentication, identification, or screening purposes.Examples of physical characteristics include fingerprints, DNA, and facial patterns. Examples of mostly behavioral characteristics include gait (a person’s manner of walking) or voice.

• Purpose:

– Subtracting background in dynamic natural environments and detecting foreground

– Labeling each object in foreground and tracking objects

• Syntax:

– Input : image sequence or video file

– Output : object descriptor, object trajectories

• Approach:

– Local Binary Pattern (LBP) : a gray-scale invariant texture primitive statistic

– Invariant to illumination change, swaying vegetation, camera jitter etc.

• Assumption:

– Non-moving camera

Object detection and tracking algorihm process is as follows

Detection and tracking algorithm

The result we have researched is as follows

Related papers:

Nguyen Thi Hai Binh and Hakil Kim; “Human Upper Body Detection in Videos: A Performance Evaluation”;

Conference on Information Security and Cryptology, 2011,pp178-182

Back to top