Video Object Detection Using Motion Vector Thresholding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for identifying objects in motion videos, such as the brute-force approach, require significant computing resources and time, leading to inefficient processing and high power consumption.
Innovation Solution
The implementation of a video analysis service that uses motion estimation by extracting motion information, such as motion vectors or optical flow, to selectively process frames based on motion thresholds, thereby reducing unnecessary processing and optimizing resource utilization, combined with deep neural networks for object detection and label smoothing across frames.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a brute-force approach is used to analyze each frame of video in isolation, then object detection accuracy is maintained, but computing resources and time consumption increase significantly
Solution Approach 1:
The system performs preliminary motion estimation and analysis before conducting full object detection. By pre-processing frames to identify those with significant motion, the system avoids unnecessary deep neural network processing on static frames, thereby maintaining detection accuracy while improving processing speed.
Solution Approach 2:
The video processing is segmented into two stages: a fast motion estimation stage that processes all frames, and a slower object detection stage that processes only selected frames with significant motion. This segmentation allows the system to maintain accuracy where needed while improving overall productivity by skipping redundant processing.
2Reliability
If a brute-force approach is used to analyze each frame of video in isolation, then object detection is performed on all frames, but computing resources and power consumption increase significantly
Solution Approach 1:
The system performs partial object detection by applying the deep neural network only to frames with significant motion rather than all frames. This partial action approach maintains reliability for detecting objects that move, while significantly reducing power consumption by avoiding redundant processing on static frames.
3Productivity
If motion estimation is used to selectively process frames, then computing resources are reduced, but processing complexity increases
Solution Approach 1:
The system introduces motion estimation as an intermediary process between frame input and object detection. This intermediary layer filters frames based on motion criteria, reducing the number of frames that require full object detection processing. While this adds a processing step, it uses computationally efficient motion estimation algorithms that do not significantly increase overall system complexity.
Data Source
AI summary
Technologies are disclosed for efficiently identifying objects in videos using deep neural networks and motion information. Using the disclosed technologies, the amount of time required to identify objects in videos can be greatly reduced. Motion information for a video, such as motion vectors, are extracted during the encoding or decoding of the video. The motion information is used to determine whether there is sufficient motion between frames of the video to warrant performing object detection on the frames. If there is insufficient movement from one frame to a subsequent frame, the subsequent frame will not be processed to identify objects contained therein. In this way, object detection will not be performed on video frames that have changed minimally as compared to a previous frame, thereby reducing the amount of time and the number of processing operations required to identify the objects in the video.