Single Camera Depth Estimation Using IMU Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for estimating depth from images, such as stereo vision and plenoptic cameras, require multiple cameras or devices, are computationally expensive, and struggle with accuracy and reliability, especially in real-time applications with single-camera setups.
Innovation Solution
A method that uses a single conventional camera with integrated accelerometer and gyroscope data to capture sequential images, rectify them, and generate a dense depth map by analyzing the movement and vibrations during image acquisition, effectively creating a virtual stereo or plenoptic system for depth estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If stereo vision techniques are used to obtain depth information, then depth estimation capability is improved, but device complexity increases due to the need for multiple cameras
Solution Approach 1:
The patent segments the depth estimation problem into two independent parts: (1) capturing multiple views using a single camera at different positions, and (2) processing these views to compute depth. This allows avoiding multiple cameras while achieving stereo vision效果
Solution Approach 2:
The patent introduces an intermediary computational process (correspondence algorithms and triangulation) that transforms single-camera sequential images into depth information, replacing the need for physical multiple cameras
2Measurement precision
If stereo vision techniques are used to obtain depth information, then depth map accuracy is improved, but computational cost increases due to correspondence algorithms
Solution Approach 1:
The patent performs preliminary actions by capturing multiple views sequentially before depth computation, and uses preliminary image processing steps (rectification, feature matching) to simplify the subsequent triangulation process, making the overall computation more efficient
3Speed
If ToF cameras are used to obtain depth information, then real-time depth measurement is improved, but power dissipation and bulkiness increase
Solution Approach 1:
The patent creates a virtual copy of the stereo vision system using software processing on a single camera, avoiding the need for physical ToF hardware components (IR transmitters, specialized sensors) that consume high power and increase bulk
Solution Approach 2:
The patent replaces the mechanical/optical ToF system with a computational approach using standard camera imaging and image processing algorithms, eliminating the need for specialized hardware
4Device complexity
If single camera sequential imaging is used to obtain depth information, then device complexity is reduced, but measurement precision deteriorates due to limited baseline
Solution Approach 1:
The patent makes the single-camera system dynamic by moving it to multiple positions sequentially, creating a virtual large baseline through temporal separation rather than spatial separation, thereby maintaining depth estimation accuracy with simplified hardware
Data Source
AI summary
A method for obtaining depth information from a scene is disclosed wherein the method has the steps of: a) acquiring a plurality of images of the scene by means of at least one camera during a time of shot wherein the plurality of images offer at least two different views of the scene; b) for each of the images of step a), simultaneously acquiring data about the position of the images referred to a six-axis reference system; c) selecting from the images of step b) at least two images; d) rectifying the images selected on step c) thereby generating a set of rectified images; and e) generating a depth map from the rectified images. Additionally devices for carrying out the method are disclosed.