Object Recognition System Using Stereo Parallax and Pattern Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognition technologies using stereo cameras struggle to accurately detect objects in varying installation environments and distances, especially when the camera installation position and angle differ, leading to difficulties in calculating parallaxes and recognizing objects due to changes in shape and illumination.
Innovation Solution
An object recognition system comprising a recognition unit, a recognition reliability calculation unit, and a combining unit that utilizes multiple methods for object recognition, including image and three-dimensional recognition, to combine recognition reliability and results, enabling accurate detection of specified objects regardless of the installation environment or object position.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If stereo camera is used for object recognition, then three-dimensional information can be obtained for accurate recognition, but the system cannot handle objects at varying distances or in different installation environments due to parallax calculation limitations
Solution Approach 1:
The patent divides object recognition into two segments: close-range objects recognized by stereo camera parallax analysis and distant objects recognized by pattern recognition. This segmentation allows each method to operate within its optimal range, resolving the contradiction between measurement precision and adaptability to varying distances and installation environments.
Solution Approach 2:
The patent dynamically switches between stereo camera-based recognition and pattern recognition based on object distance and installation environment conditions. This dynamic adaptation enables the system to maintain high recognition accuracy across varying distances and camera configurations, overcoming the static limitations of pure stereo vision systems.
2Ease of manufacture
If pattern recognition is used for object detection, then existing monitoring cameras can be utilized with lower cost, but recognition accuracy decreases when object appearance differs greatly from sample data due to illumination conditions
Solution Approach 1:
The patent introduces three-dimensional information from stereo camera parallax as an intermediary to enhance pattern recognition. The 3D depth data serves as additional features that help distinguish objects under varying illumination conditions, improving recognition reliability while still utilizing existing monitoring cameras for cost-effective deployment.
Solution Approach 2:
The patent combines two different recognition approaches (stereo vision and pattern recognition) into a composite recognition system. This composite approach leverages the strengths of both methods: the 3D spatial information from stereo cameras and the illumination-robustness of pattern recognition, achieving reliable object detection across diverse conditions while maintaining cost efficiency.
3Measurement precision
If multiple recognition methods are combined, then recognition accuracy can be improved across different distances and environments, but device complexity increases
Solution Approach 1:
The patent segments the recognition task into distinct distance-based zones, assigning different recognition methods to each zone. This segmentation simplifies the overall system architecture by creating clear decision boundaries, reducing the complexity of coordinating multiple recognition methods while maintaining high accuracy across different distances and environments.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system achieves high accuracy in detecting specified objects by combining recognition reliability and results from multiple methods, effectively addressing the limitations of existing technologies in varying installation environments and distances.
Implementation Method 1
a distance in a real space from the stereo camera to the object is calculated for each pixel of the image by using a parallax calculated by comparing a pair of left and right camera images captured by the stereo camera
Data Source
AI summary
An object recognition device includes a recognition unit, a recognition reliability calculation unit, and a combining unit. A recognition unit recognizes an object by a plurality of functions, based on information which is obtained by measuring an object by a plurality of measurement devices. A recognition reliability calculation unit calculates recognition reliability of recognition results that are obtained by recognizing an object by a recognition unit, for each function. A combining unit combines recognition reliability of the object and recognition results, detects a specified object, and outputs detection results of the specified object.