Drone Inspection Frame Extraction for Stable Image Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Inspection accuracy is significantly degraded when using flying objects like drones to capture images of periodically operating targets due to changing relative positions and orientations, leading to frames with significant noise and incomplete captures.
Innovation Solution
An information processing apparatus that acquires video from a flying object, uses a reference still image and similarity threshold to extract consecutive frames for analysis, performs filtering and image processing to stabilize the camera viewpoint, and adjusts the flying object's position to maintain optimal distance for accurate image capture.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If video is captured by a flying object in flight to inspect a target, then the inspection coverage and mobility are improved, but the relative position and orientation changes cause noise and incomplete captures that degrade inspection accuracy
Solution Approach 1:
The patent segments the video stream into individual frames and further extracts only the relevant frames that meet the similarity threshold criteria. This segmentation allows the system to process only the useful information (frames with high similarity to reference) while discarding noisy or incomplete frames, thereby maintaining inspection accuracy despite the mobile and changing conditions of flight-based inspection
Solution Approach 2:
The patent extracts and removes frames that do not meet the predetermined similarity threshold from the video data. By taking out only the relevant frames that satisfy the similarity condition, the system eliminates the harmful noise and incomplete captures caused by changing relative positions and orientations, thus preserving inspection accuracy while maintaining the mobility benefits
2Quantity of substance
If all frames from the video are used for analysis, then the completeness of data is improved, but the noise and incomplete captures from changing positions reduce the reliability of inspection results
Solution Approach 1:
The patent applies different quality criteria to different frames based on their similarity to the reference image. Frames with high similarity (above threshold) are selected for analysis, while frames with low similarity (below threshold) are discarded. This local quality filtering ensures that only reliable, high-quality frames are used for inspection, improving reliability while maintaining sufficient data completeness through the extracted frame set
3Measurement precision
If the flying object maintains optimal distance for accurate capture, then the image quality is improved, but the complexity of position adjustment and maintaining field of view increases
Solution Approach 1:
The patent uses the extracted frames and their similarity scores as feedback to evaluate whether the flying object is at the optimal distance and orientation for accurate capture. The system analyzes the quality of extracted frames and can feed this information back to adjust the flying object's position or to determine when repositioning is needed, thereby maintaining image quality while managing position control complexity through intelligent feedback loops
Data Source
AI summary
An information processing apparatus (10) according to the present disclosure includes an image acquisition interface (131) that acquires a video of an inspection target from a flying object (20), the video having been taken by the flying object (20) in flight, a memory (12) storing a reference still image serving as a reference for analysis and a threshold of similarity between the reference still image and frames included in the acquired video, and an image extractor (132) that calculates the similarity to the reference still image for each frame and extracts a plurality of consecutive frames as an analysis target based on the similarity and the threshold.