Object Label Persistence via Pixel Shift Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image capturing devices with limited computational power face challenges in efficiently analyzing and labeling objects across multiple images captured by a moving camera, leading to redundant processing and high computational burdens due to the need for full-image analysis and network communication.
Innovation Solution
Implementing a system with a Convolutional Neural Network (CNN) that analyzes images, caches analysis results, and uses movement sensors to detect pixel shifts, allowing for the reuse of cached analysis results for subsequent images, thereby reducing computational workload and minimizing network communication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full-image analysis is performed on each captured image, then object labeling accuracy is improved, but computational burden increases
Solution Approach 1:
The patent segments the image analysis process by dividing the image into multiple blocks or regions. Instead of analyzing the entire image, the system identifies and processes only specific blocks containing objects of interest, thereby reducing computational burden while maintaining labeling accuracy for relevant objects.
Solution Approach 2:
The patent performs preliminary actions by using movement sensors to predict the camera's next position and pre-identifying potential object blocks before the actual image capture. This preliminary block identification reduces the computational workload during the main labeling process by narrowing down the search space in advance.
2Reliability
If full-image analysis is performed on each captured image, then object detection completeness is improved, but processing time increases
Solution Approach 1:
The patent applies local quality by focusing computational resources on specific local regions (blocks) of the image that are likely to contain objects, rather than uniformly processing the entire image. This selective local processing maintains detection completeness for relevant objects while significantly reducing overall processing time.
Solution Approach 2:
By using movement sensor data to preliminarily determine which image blocks may contain objects before actual processing, the system prepares a targeted list of regions to analyze. This preliminary action ensures that no potential objects are missed while avoiding unnecessary processing of empty regions, thus reducing processing time.
3Measurement precision
If image analysis is performed without using cached results, then analysis accuracy is improved, but network communication increases
Solution Approach 1:
The patent performs preliminary analysis on the first image and caches the results, including identified objects and their characteristics. When processing subsequent images, the system preliminarily checks the cache for matching objects based on movement data, avoiding redundant network communication while maintaining analysis accuracy through intelligent result reuse.
Solution Approach 2:
The patent creates copies of analysis results from the first image and stores them in a cache. For subsequent images, instead of performing complete new analyses that would require network communication, the system copies and adapts previously validated analysis results, significantly reducing network energy consumption while preserving accuracy.
4Measurement precision
If pixel shift detection is not used, then system complexity is reduced, but object tracking accuracy decreases
Solution Approach 1:
The patent introduces pixel shift detection as an intermediary mechanism that bridges the first image analysis results and subsequent image processing. By detecting the pixel shift caused by camera movement, the system accurately transforms and matches objects across different images, maintaining high tracking accuracy without requiring complex full-image re-analysis.
Solution Approach 2:
The system performs preliminary pixel shift detection using movement sensor data before processing subsequent images. This preliminary action provides accurate transformation information that simplifies the object matching process, improving tracking accuracy while avoiding the need for complex real-time image registration algorithms.
Data Source
AI summary
A method enables object label persistence between subsequent images captured by a camera. One or more processors receive a first image, which is captured by an image sensor on a camera, and which includes a depiction of an object. The processor(s) generate a label for the object, and display the first image on a display. The processor(s) subsequently receive movement data that describes a movement of the camera after the image sensor on the camera captures the first image and before the image sensor on the camera captures a second image. The processor(s) receive the second image. The processor(s) display the second image on the display, and then detect a pixel shift between the first image and the second image as displayed on the display. The processor(s) then label the object with the label on the second image as displayed on the display.