Movable Cabin Camera Positioning for Low-Cost 3D Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection devices in vehicle cabins face challenges in accurately determining the size of detected targets using 2D cameras and incurring high costs due to the use of 3D sensing methods or high-performance processors.
Innovation Solution
An object detection device that uses a single image sensor and a sensor moving device to move the sensor part rectilinearly or rotationally, combined with a controller to detect the object's position by processing data before and after the sensor's movement, thereby reducing the need for multiple sensors and high-performance hardware.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a 3D camera is used to obtain 3D coordinates of objects in the vehicle cabin, then the accuracy of position detection is improved, but the device cost increases
Solution Approach 1:
The patent applies the dynamics principle by making the 2D camera movable instead of stationary. The camera is moved to different positions and angles to capture multiple images of the same object, enabling 3D coordinate calculation through multi-view geometry. This dynamic approach allows a simple 2D camera to achieve 3D detection capabilities that would otherwise require expensive 3D cameras.
Solution Approach 2:
The patent uses the copying principle by capturing multiple copies (images) of the same object from different camera positions. These multiple 2D image copies are then processed together to reconstruct 3D coordinates, effectively replacing the need for a single expensive 3D camera with multiple inexpensive 2D camera captures.
2Device complexity
If a 2D camera with distance estimation algorithm is used, then the device cost is reduced, but the accuracy of determining target size based on distance deteriorates
Solution Approach 1:
The patent moves the 2D camera to multiple known positions around the object, capturing images from different distances and angles. By knowing the exact camera positions and using multi-view geometry, the system can accurately calculate both distance and target size without relying on less accurate monocular depth estimation algorithms.
Solution Approach 2:
The patent transitions from 2D image analysis to 3D spatial reasoning by incorporating camera position information as an additional dimension. The multiple 2D images taken from different spatial positions are combined with known camera coordinates to compute accurate 3D target dimensions, effectively adding the spatial dimension back into the measurement process.
3Measurement precision
If a structured light camera or TOF camera is used for 3D sensing, then the position detection accuracy is improved, but the material cost increases
Solution Approach 1:
The patent creates multiple 2D image copies from different camera positions and combines them to achieve 3D detection. This approach replaces expensive specialized 3D sensing hardware (structured light or TOF cameras) with multiple standard 2D camera captures, achieving similar accuracy at lower material cost.
Solution Approach 2:
The patent replaces complex optical-mechanical 3D sensing systems (structured light projection and detection) with a simpler computational approach using multiple 2D images. Instead of using mechanical/optical systems to directly measure depth, the solution uses geometric computation from multiple viewpoint images to derive 3D information.
Data Source
AI summary
An object detection device is capable of accurately detecting a position of an object that is a detection target inside a vehicle cabin by using a single image sensor. The object detection device also includes: a sensor part configured to photograph the object existing inside the vehicle cabin; a sensor moving device configured to move a position of the sensor part; and a controller configured to detect the position of the object by combining data on the object photographed before movement of the sensor part and data on the object photographed after the movement of the sensor part.