Dual-Pitch Image Recognition for Fast Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image recognition methods require significant time for high-resolution image processing and suffer from reduced precision when using lower pixel pitches, making them inefficient for detecting predetermined image objects in captured images.
Innovation Solution
An image recognition program that employs a dual-precision approach, first detecting an object area with low precision to shorten the detection time and then performing high-precision object detection within that area to maintain precision, using pixel values at different pitches to optimize the process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image recognition is performed using all pixels of the captured image, then detection precision is maintained, but the processing time is significantly increased
Solution Approach 1:
The patent divides the image processing into two segments: first processing the entire image at low precision to identify a candidate area, then processing only that specific area at high precision. This segmentation allows the system to maintain detection precision while reducing the overall processing time by limiting high-precision operations to a small portion of the image.
Solution Approach 2:
The patent performs preliminary low-precision processing across the entire image to identify a candidate area before performing high-precision detection. This preliminary action narrows down the search space, allowing subsequent high-precision processing to focus only on the relevant region, thereby reducing total processing time while maintaining accuracy.
2Productivity
If image recognition is performed with lower precision to reduce processing time, then detection speed is improved, but object detection precision is reduced
Solution Approach 1:
The patent applies different processing qualities to different regions of the image: low precision is applied to the entire image for initial scanning, while high precision is applied specifically to the candidate area where the object is likely located. This local quality differentiation optimizes both speed and precision by allocating computational resources efficiently.
3Measurement precision
If the entire captured image is processed with high precision, then object detection accuracy is maintained, but processing efficiency is reduced
Solution Approach 1:
The patent extracts the candidate area from the entire captured image based on low-precision processing results, then applies high-precision detection only to this extracted region. This extraction approach maintains detection accuracy while improving processing efficiency by eliminating unnecessary high-precision computations in areas where objects are not present.
Data Source
AI summary
A game apparatus obtains a captured image captured by a camera. First, the game apparatus detects an object area of the captured image that includes a predetermined image object based on pixel values obtained at a first pitch across the captured image. Then, the game apparatus detects a predetermined image object from an image of the object area based on pixel values obtained at a second pitch smaller than the first pitch across the object area of the captured image.