Stereo Camera Object Recognition via Situation Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional moving object recognition systems face challenges in accurately detecting and tracking objects, particularly due to sensitivity to lighting conditions, object position changes, and variations in view, leading to non-recognition or erroneous recognition, which can be hazardous in applications like automatic braking systems.
Innovation Solution
The proposed image processing device employs a solid object recognition system that recognizes current and future situations using a stereo camera unit and information processing unit, adjusting recognition methods and thresholds based on situational predictions to improve object recognition accuracy, and continuously tracks recognized objects by varying algorithm weights and selecting appropriate recognition methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a moving object is directly recognized from the disparity image obtained by the stereo camera device, then the distance measurement can be achieved, but the recognition accuracy for the moving object is low
Solution Approach 1:
The recognition process is segmented into two distinct stages: first recognizing the situation (environmental context, road conditions, weather) from the disparity image, then using that situation recognition result to guide the target object recognition. This segmentation allows each stage to optimize for its specific task, improving overall accuracy.
Solution Approach 2:
The situation recognition is performed as a preliminary action before target object recognition. By first understanding the environmental context, lighting conditions, and scene characteristics, the system prepares appropriate recognition parameters and thresholds that are optimized for the current situation, thereby improving subsequent target recognition accuracy.
2Device complexity
If the recognition method uses fixed thresholds and algorithms, then the processing is simple, but the recognition accuracy decreases under changing conditions such as brightness variations and shape changes
Solution Approach 1:
The recognition system dynamically adjusts its parameters and algorithm selection based on the recognized situation. Different recognition methods are selected for different situations (e.g., different lighting conditions, weather, road types), allowing the system to adapt to changing environmental conditions rather than using fixed thresholds and algorithms.
Solution Approach 2:
The system changes recognition parameters such as thresholds, algorithm weights, and processing methods based on the situation recognition results. For example, lighting conditions may trigger adjustments to brightness thresholds, while object shape changes may alter the recognition algorithm selection, maintaining high accuracy across varying conditions.
Data Source
AI summary
An image processing device includes circuitry to acquire a range image, recognize an external situation indicating a situation of an imaging area corresponding to the range image, recognize a recognition target to be tracked from the range image by a recognition method based on the recognized external situation, remove the recognition target from the range image to generate a new range image, and detect an object in the new range image from which the recognition target has been removed.