Stereo Camera Depth Estimation Using Object-Aware Disparity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Stereo camera depth estimation in vehicles is computationally expensive and often inaccurate due to the challenge of calculating depth from surrounding pixels that may not relate to the same object, making it unsuitable for real-time applications in dynamic weather conditions.
Innovation Solution
The method improves stereo estimation by classifying pixels based on object types before depth estimation, optimizing disparity calculations using semantic information and sample pixels to reduce computational expenses, and prioritizing disparities for closer objects, thereby enhancing accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If stereo camera depth estimation is performed using traditional methods, then depth information can be obtained, but the computational cost is high and accuracy is reduced due to calculating depth from surrounding pixels that may not relate to the same object
Solution Approach 1:
The patent segments the image processing task by identifying and focusing on specific objects of interest rather than processing all surrounding pixels. The system divides the computational workload into object detection, object classification, and depth estimation for identified objects only, thereby reducing overall computational complexity while maintaining or improving depth estimation accuracy.
Solution Approach 2:
The patent applies local quality by using object type information to guide disparity calculations. Different object types receive different processing priorities and methods, with the system allocating computational resources selectively based on object importance and characteristics rather than applying uniform processing to all pixels.
2Productivity
If traditional stereo depth estimation is used, then depth can be calculated, but the process is too computationally expensive for real-time applications in dynamic weather conditions
Solution Approach 1:
The patent performs preliminary object detection and classification before depth estimation. By identifying and categorizing objects in advance, the system prepares the data structure and prioritization needed for efficient depth calculation, enabling real-time processing by avoiding redundant computations during the actual depth estimation phase.
Solution Approach 2:
The patent changes the processing parameters dynamically based on object type and scene conditions. The system adjusts disparity calculation methods, search ranges, and processing priorities according to the detected objects and weather conditions, optimizing computational energy consumption while maintaining real-time performance.
Data Source
AI summary
A method comprises: receiving, at a processor, a first image from a first camera from a stereo camera pair and a second image from a second camera from the stereo camera pair. The method also includes determining, at the processor using a machine learning model, a first set of objects in the first image. The processor determines an object type. The processor identifies a second set of objects in the second image associated with the first plurality of objects. The method also includes calculating, at the processor, a set of disparity values between the first image and the second image based on (1) an object from the first set of objects, (2) an object from the second set of objects and associated with the object from the first set of objects, and (3) an object type of the object from the first set of objects.