Depth Image Target Region Extraction for Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies for recognizing and processing real-world objects using computer information processing face challenges in precision and user convenience, particularly when objects need to be distinguished from others and when environmental factors like brightness affect detection and tracking.
Innovation Solution
An information processing device and method that utilize a depth image acquiring unit, a target region setting unit, and a target detecting unit to virtually set a three-dimensional target region and extract images of objects within it, allowing for precise detection and tracking without the need for pre-registration or external markers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a marker is added to the object for detection, then the object can be easily recognized and tracked, but the user experiences inconvenience in handling and the risk of breakage increases
Solution Approach 1:
The invention extracts the detection function from the physical object by using a separate imaging device to capture images of the object. The object itself remains marker-free and unchanged, while the imaging device and processing system perform the detection and recognition functions externally.
Solution Approach 2:
The invention introduces an intermediary processing system that includes an imaging device, image processing unit, and database. This intermediary system mediates between the physical object and the detection requirements, allowing objects to be recognized without physical markers while maintaining high detection precision through image analysis.
2Ease of operation
If feature-based object recognition is used, then no markers are needed on objects, but detection precision changes depending on image capturing environment such as brightness
Solution Approach 1:
The invention performs preliminary actions by capturing images of objects under various lighting conditions and storing them in a database before actual detection occurs. This pre-captured image library allows the system to compare against known object appearances regardless of current lighting conditions, maintaining detection precision while requiring no markers on objects.
Solution Approach 2:
The invention changes the parameter being measured from direct visual features sensitive to lighting conditions to a comparison-based approach using stored reference images. By comparing captured images against a database of pre-captured images, the system achieves robust detection that is insensitive to brightness and environmental variations.
3Measurement precision
If pre-registration of object shapes is required for recognition, then registered objects can be accurately identified, but unregistered objects are excluded from detection targets
Solution Approach 1:
The imaging device and image processing system serve multiple functions: they can detect both pre-registered objects and newly introduced objects, accommodate objects of various shapes and sizes, and adapt to different detection scenarios without requiring separate systems or pre-registration procedures.
Solution Approach 2:
The detection system is dynamic and adaptable - it can adjust to new objects as they are introduced to the play area without requiring pre-registration. The system continuously captures images and compares them against the database, allowing it to dynamically recognize both known and novel objects, expanding its detection scope while maintaining accuracy.
Data Source
AI summary
On a play field 20 set in real space, a target region 102 having the play field 20 as a bottom surface and having a predetermined height h is set virtually. A target region image that expresses a distance of the target region 102 from an imaging surface as a pixel value on an image plane is generated, and compared with a depth image corresponding to an actual captured image 100; thus, only images of blocks 3c and 3b inside the target region 102 are extracted as detection and tracking targets.