Human Action Recognition Using Trajectory-Based Spatiotemporal Descriptors

Dhamsania, Chandni; Ratanpara, Tushar V.

doi:10.1007/978-981-10-3153-3_1

Cited by 1 publication

(1 citation statement)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Video is at first layer while frames are at last layer. Longer video means high number of frames at last level [22]. Video ranking is the process of arranging multiple videos in order based on some parameters.…”

Section: Introductionmentioning

confidence: 99%

An Automatic Content Based Video Ranking from Surveillance Videos

Shah¹,

Ratanpara²

2017

IJET

View full text Add to dashboard Cite

Surveillance systems become more popular in our daily life because of rapid development of the camera industry. However, it is very difficult to track specific person and active time slots from a mass of long duration surveillance videos. Automatic surveillance video ranking is also desirable in many applications like human activity detection, traffic monitoring, public safety, crime prevention etc. This research work proposed a method to extract an active time slot of a person. An approach is also introduced for ranking multiple surveillance videos based on person activity duration. First valid frames for synopsis are identified using Viola-Jones face detection algorithm and foreground object detection method. Video synopsis is generated combining all these valid frames. Object tracking is done by computing face feature using Speed-Up Robust Feature (SURF). Face detection and face feature computation are also applied to query image. Similarity with query image is found based on a match face features. Active time slot, entry and exit time of a query person is computed with respect to match feature value between query image and video synopsis. Multiple surveillance video ranking is done based on active time slot. The proposed approach is tested on various surveillance videos.

show abstract

Section: Introductionmentioning

confidence: 99%