Method and apparatus for detecting abnormal objects in video
AssigneeIndustry Academy Cooperation Foundation of Sejong University
InventorsYong Guk KIM, Long Thinh NGUYEN
Life of this patent
Estimating the term…
Abstract
Disclosed are a method and an apparatus for detecting abnormal objects in a video. The method for detecting abnormal objects in a video reconstructs a restored batch by applying each input batch to which an inpainting pattern is applied to a trained auto-encoder model, and fuses a time domain reconstruction error using time domain restored frames output by extracting and restoring a time domain feature point by applying a spatial domain reconstruction error and a plurality of successive frames using a restored frame output by combining the reconstructed restoring batch to a trained LSTM auto-encoder model to estimate an area where an abnormal object is positioned.
Claims10
An apparatus for detecting abnormal objects in a video, the apparatus comprising:
- a batch unit generating an input batch by applying each of a plurality of inpainting patterns to a current frame of a video;
- a first prediction unit reconstructing a restored batch by applying each input batch to which the inpainting pattern is applied to a trained auto-encoder model and combining the reconstructed restored batch to output a restored frame;
- a second prediction unit extracting and then restoring a time domain feature point by applying a plurality of successive frames to an LSTM auto-encoder model to output time domain restored frames, wherein the plurality of successive frames are a plurality of successive frames including the current frame;
- and
- an abnormal object detection unit deriving a spatial domain reconstruction error by using the restored frame and the current frame, deriving a time domain reconstruction error by using the successive frames and the time domain restored frame, and estimating an area whether an abnormal object is positioned by fusing the spatial domain reconstruction error and the time domain reconstruction error.
A method for detecting abnormal objects in a video, the method comprising:
- (a) generating an input batch by applying each of a plurality of inpainting patterns to a current frame of a video;
- (b) reconstructing a restored batch by applying each input batch to which the inpainting pattern is applied to a trained auto-encoder model and combining the reconstructed restored batch to output a restored frame;
- (c) extracting and then restoring a time domain feature point by applying a plurality of successive frames to an LSTM auto-encoder model to output time domain restored frames, wherein the plurality of successive frames are a plurality of successive frames including the current frame;
- and
- (d) deriving a spatial domain reconstruction error by using the restored frame and the current frame, deriving a time domain reconstruction error by using the successive frames and the time domain restored frame, and estimating an area whether an abnormal object is positioned by fusing the spatial domain reconstruction error and the time domain reconstruction error.
Dependencies are inferred from the claim text.