Object Recognition Using 3D Position and 2D Reference Image Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognition technologies face challenges in accurately identifying the type of objects with variable shapes, such as humans or robots, using combined two-dimensional and three-dimensional image data, as they rely heavily on object shape recognition rather than predefined reference images.
Innovation Solution
An object recognition device that captures both two-dimensional and three-dimensional images, extracts areas with specific pixel value ranges, and searches for registered reference images within these areas to determine object types, allowing for reliable recognition even with variable shapes by using pre-defined marks and adjusting reference image scales and orientations based on height and inclination information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If object recognition relies on shape analysis from combined 2D and 3D images, then positional information can be acquired, but recognition accuracy deteriorates for objects with variable shapes
Solution Approach 1:
The patent segments the object recognition process into two independent parts: (1) extracting object position and posture from 3D image data, and (2) recognizing object type by searching for reference images in the 2D image. This segmentation allows each part to be optimized independently, resolving the contradiction between handling variable shapes and maintaining recognition accuracy.
Solution Approach 2:
The patent extracts the object type recognition function from the overall recognition system and implements it as a separate reference image search process. By taking out the recognition determination from the combined 2D-3D analysis, the system can accurately identify objects with variable shapes without being affected by shape variations.
2Measurement precision
If reference images are searched in the entire two-dimensional image, then object type recognition is possible, but processing time increases
Solution Approach 1:
The patent makes the reference image search area dynamic by adjusting it based on the object's position and posture detected from 3D image data. The search area is dynamically positioned and sized according to the extracted object characteristics, which reduces the search scope while maintaining recognition accuracy, thus reducing processing time.
Solution Approach 2:
The patent performs preliminary extraction of object position and posture from 3D image data before conducting the reference image search in 2D images. This preliminary action defines a restricted search area, preventing the need to search the entire 2D image and thereby reducing processing time while maintaining recognition accuracy.
3Reliability
If multiple cameras are used to capture both two-dimensional and three-dimensional images, then recognition reliability improves, but device complexity increases
Solution Approach 1:
The patent makes the imaging unit multi-functional by configuring it to capture both 2D images and 3D images (using techniques like structured light or stereo vision). This universal imaging capability allows a single device to provide both positional information from 3D data and recognition information from 2D data, improving reliability without proportionally increasing device complexity.
Solution Approach 2:
The patent merges the 2D imaging function and 3D imaging function into a unified imaging system that captures both types of data simultaneously or in sequence. By combining these functions in one integrated system rather than separate systems, the patent improves recognition reliability while minimizing the increase in device complexity.
Data Source
AI summary
The object recognition device includes an imaging unit that captures images of a predetermined monitoring area to acquire a three-dimensional image and a two-dimensional image, an object extraction unit that extracts an area having pixels whose pixel values are within a predetermined range from the three-dimensional image acquired by the imaging unit, an image searching unit that searches the two-dimensional image, acquired by the imaging unit, for a reference image registered in advance according to the type of an object, and a determination unit that determines the type of the object depending on whether or not the reference image searched for by the image searching unit exists in the area extracted by the object extraction unit.