Reducing an amount of storage used to store surveillance videos
In some examples, a computing device may receive a request for a video segment captured by a particular camera, where the request specifies a date, a start time, and a length of the video segment. The computing device may identify stored data associated with the video segment in a storage device based on the date, the start time, the length, and an identifier associated with the particular camera and retrieve the stored data from the storage device. The computing device may determine that the stored data includes a subset of the video frames that were sent from the particular camera and excludes a remainder of the video frames and regenerate the remainder of the video frames based on the stored data to create regenerated data. The computing device may reconstruct the reconstructed video segment by merging the stored data with the regenerated data and provide the reconstructed video segment.
1. A method comprising:
receiving a sequence of video frames for a video segment from a particular camera of a plurality of cameras;
discarding every second video frame of the sequence to create a sub sequence of video frames;
comparing adjacent video frames of the sub sequence;
determining that the adjacent video frames differ from each other by less than a predetermined amount;
discarding one of the adjacent video frames from the sub sequence to create an updated sub sequence;
storing the updated subsequence in memory;
receiving, by one or more processors, a request for the video segment, the request specifying a date, a start time, a time length of the video segment, and an identifier associated with the particular camera;
selecting, based on one or more criteria, the updated subsequence stored in the memory in response to receiving the request;
regenerating, by one or more processors executing a machine learning algorithm, the discarded every second video frame of the sequence and the discarded adjacent video frame based on the updated sub sequence of video frames to create regenerated video frames;
merging, by one or more processors, the updated sub sequence of video frames with the regenerated video frames to create a reconstructed video segment; and
providing, by one or more processors, the reconstructed video segment as the requested video segment.
2. The method of claim 1 , wherein one of the discarded every second video frame is regenerated based on adjacent video frames of the sub sequene.
3. The method of claim 1 , wherein the machine learning algorithm comprises an extreme learning algorithm.
4. The method of claim 1 , wherein regeneration and reconstruction of each one minute portion of the reconstructed video segment takes three seconds or less.
5. The method of claim 1 , wherein,
wherein time stamps of the adjacent video frames are within a predetermined time period of each other.
6. A computing device comprising:
one or more processors; and
one or more non-transitory computer readable media storing instructions executable by the one or more processors to perform operations comprising:
receiving a sequence of video frames for a video segment from a particular camera of a plurality of cameras;
discarding every second video frame of the sequence to create a sub sequence of video frames;
comparing adjacent video frames of the sub sequence;
determining that the adjacent video frames differ from each other by less than a predetermined amount;
discarding one of the adjacent video frames from the sub sequence to create an updated sub sequence;
storing the updated subsequence in memory;
receiving a request for the video, the request specifying a date, a start time, and a time length of the video segment, and an identifier associated with the particular camera;
selecting, based on one or more criteria, the updated subsequence stored in the memory in response to receiving the request;
regenerating the discarded every second video frame of the sequence and the discarded adjacent video frame based on the updated sub sequence of video frames to create regenerated video frames;
reconstructing a reconstructed video segment by merging the updated sub sequence of video frames with the regenerated video frames; and
providing the reconstructed video segment as the requested video segment.
7. The computing device of claim 6 , wherein the machine learning algorithm comprises an extreme learning algorithm.
8. The computing device of claim 6 ,
wherein time stamps of the adjacent video frames are within a predetermined time period of each other.
9. The computing device of claim 6 , wherein the operation of regenerating and reconstructing of each one minute portion of the reconstructed video segments takes three seconds or less.
10. One or more non-transitory computer readable media storing instructions executable by one or more processors to perform operations comprising:
receiving a sequence of video frames for a video segment from a particular camera of a plurality of cameras;
discarding every second video frame of the sequence to create a sub sequence of video frames;
comparing adjacent video frames of the sub sequence;
determining that the adjacent video frames differ from each other by less than a predetermined amount;
discarding one of the adjacent video frames from the sub sequence to create an updated sub sequence;
storing the updated subsequence in memory;
receiving a request for the video, the request specifying a date, a start time, and a time length of the video segment, and an identifier associated with the particular camera;
regenerating, by executing a machine learning algorithm, the discarded every second video frame of the sequence and the discarded adjacent video frame based on the updated sub sequence of video frames to create regenerated video frames;
reconstructing a reconstructed video segment by merging the updated sub sequence of video frames with the regenerated video frames; and
providing the reconstructed video segment as the requested video segment.
11. The one or more non-transitory computer readable media of claim 10 , wherein the machine learning algorithm comprises an extreme learning algorithm.
12. The one or more non-transitory computer readable media of claim 10 , wherein
time stamps of the adjacent frames are within a predetermined time period of each other.
13. The one or more non-transitory computer readable media of claim 10 , wherein regeneration and reconstruction of each one minute portion of the reconstructed video segment takes three seconds or less.