Object Detection Using Segmented Image Regions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection techniques, such as YOLOv4, face challenges in accurately detecting regions representing objects, often leading to incorrect detections of inappropriate regions or objects.
Innovation Solution
A method involving a non-transitory computer-readable storage medium storing program instructions that acquire captured image data, process it to detect a first type region indicating an object, and then use a second image data representing the object with a higher number of pixels to detect a second type region for inspection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a machine learning model like YOLOv4 is used to detect objects, then object detection can be performed, but the detection accuracy is insufficient and inappropriate regions may be detected
Solution Approach 1:
The patent divides the detection process into two separate stages: a first detection process that identifies candidate regions, and a second detection process that performs precise detection on cropped images of those regions. This segmentation allows each process to be optimized for its specific purpose, improving overall detection accuracy and reliability by preventing inappropriate regions from being misidentified.
Solution Approach 2:
The first detection process performs preliminary identification of candidate object regions in the captured image. Based on these detection results, the system crops and prepares second images for more accurate detection. This preliminary action filters out inappropriate regions early, ensuring that only promising candidates undergo the more resource-intensive second detection process.
2Measurement precision
If a single detection process is used, then the processing speed is maintained, but the detection accuracy and precision are insufficient
Solution Approach 1:
The detection system is segmented into two distinct processes: a first detection process for rapid candidate region identification, and a second detection process for precise measurement on cropped images. This segmentation enables the system to maintain high processing efficiency through the first process while achieving high precision through the second process, avoiding the need to process entire images at full resolution.
Solution Approach 2:
The system extracts and crops only the relevant candidate regions from the full captured image to create second images for the second detection process. This extraction eliminates unnecessary processing of background and irrelevant areas, significantly improving detection precision while maintaining processing efficiency by focusing computational resources only on regions of interest.
3Measurement precision
If the entire captured image is processed for detection, then all regions are covered, but the processing time and computational resources increase
Solution Approach 1:
The system extracts only the candidate object regions identified by the first detection process and crops them into separate second images for the second detection process. This extraction approach ensures that all potential objects are covered (maintaining completeness) while significantly reducing the total processing time by excluding irrelevant background regions from the computationally intensive second detection process.
Solution Approach 2:
The first detection process performs preliminary identification of candidate regions before the second detection process begins. This preliminary action creates a filtered set of regions that need detailed analysis, ensuring comprehensive coverage of all potential objects while minimizing processing time by preparing only the necessary cropped images in advance for the second process.
Data Source
AI summary
A set of program instructions, when executed by a computer, causes the computer to perform: acquiring captured image data of a captured image including a first object; acquiring first image data from the captured image data, the first image data representing the first object with a first number of pixels; detecting a first type region indicating the first object by using the first image data; acquiring second image data from the captured image data, the second image data indicating a partial image; detecting a second type region by using the second image data, the second type region indicating at least part of the first object, the second image data representing the first object with a second number of pixels, the second number of pixels being greater than the first number of pixels; and inspecting the first object by using a detection result of the second type region.