Object Detection via Random Subimage Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object detection systems in vehicles are prone to varying classifications of the same object due to minor pixel adjustments caused by factors like debris or changes in lighting, leading to inconsistent object identification.
Innovation Solution
The system generates multiple subimages of an image by dividing it into equally sized divisions using a random number generator and inputs these subimages into a machine learning program to identify objects, using a multi-task neural network to classify objects when a threshold of matching classifications is met, thereby improving object detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional object detection systems process images directly, then processing speed is maintained, but object identification consistency deteriorates due to pixel variations from debris or lighting changes
Solution Approach 1:
The patent divides the original image into multiple subimages by segmenting the image into a grid of divisions and randomly selecting subsets. This segmentation allows the system to process multiple views of the same scene, improving object identification consistency by reducing the impact of local pixel variations caused by debris or lighting changes.
2Measurement precision
If multiple subimages are generated and processed, then object detection accuracy improves, but computational time increases
Solution Approach 1:
The patent processes multiple subimages (partial action) rather than requiring analysis of the entire original image. By randomly selecting and processing subsets of image divisions, the system achieves improved detection accuracy through multiple perspectives while reducing the total computational burden compared to analyzing the complete image with high precision.
Data Source
AI summary
A plurality of subimages of an image are generated based on output from a random number generator. Each subimage is input to a machine learning program trained to output an object classified in the subimage. When a number of subimages with a same classified object exceeds a threshold, an object in the image is identified as the classified object.