Video Event Detection via Key Frame Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods lack efficiency in quickly analyzing and summarizing surveillance footage to identify important events, and there is no straightforward way to store or label significant video scenes from a segment of video.
Innovation Solution
A method involving the extraction of portions from each frame of a video segment, comparison with a reference portion to detect events, and creation of a visual summary for rapid event retrieval and reporting, utilizing motion-based or feature-based summaries to identify specific objects or movements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If surveillance footage is analyzed frame by frame to identify important events, then event detection accuracy is improved, but analysis time increases significantly
Solution Approach 1:
The patent divides the video analysis process into multiple segments by selecting specific key frames (first frame, middle frame, last frame) rather than analyzing every frame sequentially. This segmentation approach maintains event detection accuracy while significantly reducing the total number of frames that need to be processed, thereby decreasing analysis time.
Solution Approach 2:
The patent performs preliminary analysis by comparing the first and last frames to detect changes that indicate events. This preliminary action allows the system to quickly identify potential events before conducting more detailed analysis, reducing the overall time required while maintaining detection accuracy.
2Loss of information
If all video frames are stored for later review, then complete video archive is preserved, but storage requirements increase
Solution Approach 1:
The patent extracts and stores only the most relevant information from video frames - specifically, it compares key frames (first, middle, and last) to identify events, and stores only those frames where events occur or where changes are detected. This extraction approach preserves the essential information needed for review while dramatically reducing the total storage space required compared to archiving all frames.
Solution Approach 2:
The patent discards redundant frame data by not storing every frame, but instead recovers the necessary information through selective storage of key frames and event-related frames. The comparison logic allows the system to recover complete event information from these selected frames, maintaining archive completeness while minimizing storage requirements.
3Measurement precision
If detailed analysis of every video frame is performed, then comprehensive event detection is achieved, but processing complexity increases
Solution Approach 1:
The patent simplifies processing complexity by segmenting the video into key frames (first, middle, and last) and performing comparisons only between these segmented frames. This approach achieves comprehensive event detection by capturing the essential changes throughout the video sequence while avoiding the computational complexity of analyzing every single frame in detail.
Solution Approach 2:
The patent applies partial action by performing full detailed analysis only on selected key frames rather than all frames. The comparison of first, middle, and last frames provides sufficient event detection capability without the excessive processing complexity that would result from analyzing every frame with the same level of detail.
Data Source
AI summary
There is disclosed a quick and efficient method for analyzing a segment of video, the segment of video having a plurality of frames. A reference portion is acquired from a reference frame of the plurality of frames. Plural subsequent portions are then acquired from a corresponding subsequent frame of the plurality of frames. Each subsequent portion is then compared with the reference portion, and an event is detected based upon each comparison. There is also disclosed a method of optimizing video including selectively storing, labeling, or viewing video based on the occurrence of events in the video. Furthermore, there is disclosed a method for creating a video summary of video which allows a used to scroll through and access selected parts of a video. The methods disclosed also provide advancements in the field of video surveillance analysis.