Video Frame Entropy Filtering for False-Positive Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Object detection algorithms based on artificial intelligence are prone to false-positive detections due to non-domain objects such as artifacts and blurred images, which distort actual information and hinder effective perception.
Innovation Solution
Implement a method to suppress potentially false-positive detections by calculating and indicating information content in video images using extended entropy metrics and similarity characteristics, and filtering out images with low content relevance or similarity to known disturbances.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If object detection algorithms process all images in a video signal, then detection coverage is maximized, but false-positive detections increase due to artifacts and blurred images
Solution Approach 1:
The patent applies preliminary action by calculating an information content metric for each image frame before performing object detection. This pre-assessment allows the system to identify and skip frames with insufficient information content (such as artifacts or blurred images) that would likely produce false-positive detections, thereby improving detection reliability without processing unnecessary frames
Solution Approach 2:
The patent introduces an intermediary mechanism - an information content metric based on entropy calculation - that acts as a mediator between the raw video signal and the object detection algorithm. This intermediary assesses image quality and determines whether a frame is suitable for detection, filtering out problematic frames before they can generate false positives
2Reliability
If object detection algorithms process all images in a video signal, then no detections are missed, but computing power is wasted on images without relevant content
Solution Approach 1:
The patent applies partial action by performing object detection only on a subset of images that meet the information content threshold. Instead of processing every frame, the system selectively processes only those frames with sufficient information content, reducing computing power consumption while maintaining detection completeness for relevant frames
Solution Approach 2:
The system performs preliminary assessment of each frame's information content using entropy calculation before committing computational resources to object detection. This pre-filtering mechanism ensures that computing power is allocated only to frames with relevant content, optimizing energy efficiency
3Reliability
If information content calculation is performed on every image section, then false-positives are reduced, but processing time increases
Solution Approach 1:
The patent extracts and processes only the essential information content metric from each image frame using entropy calculation, rather than performing comprehensive analysis on every image section. This selective extraction approach reduces processing time while maintaining the ability to identify and suppress false-positive detections
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An information content of a section of a current image of a series of images of a video signal is calculated (S101), wherein the video signal has to be fed to an algorithm for calculating and indicating detections of objects in the video signal. If the calculated information content of the section of the current image is smaller than a threshold value (S103), the calculation and indication of detections of objects for the section of at least the current image or the current image and further images of the series of images is suppressed (S105).