Camera-Based Object Distance Estimation With Bounding-Box Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing collision avoidance systems for personal use, such as for skiers, cyclists, and motorcyclists, face challenges due to errors in distance estimation caused by camera tilt and object rotation, which are not corrected by existing methods, and require bulky, energy-intensive sensors like LIDAR or RADAR, making them impractical for portable use.
Innovation Solution
A system using a single camera with a processing device that applies correction criteria based on pre-stored parameters for bounding box dimensions to correct tilt and rotation errors, allowing accurate distance estimation without additional sensors or machine learning, using a convolutional neural network for object recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If LIDAR or RADAR sensors are used for distance measurement, then measurement precision is improved, but device complexity and energy consumption increase
Solution Approach 1:
The patent replaces complex electromagnetic sensors (LIDAR/RADAR) with a simple camera-based optical system. The camera captures images and the processing device calculates distance using image analysis algorithms, substituting mechanical/electromagnetic measurement systems with an optical-mechanical system that is lighter, cheaper, and more suitable for personal use devices.
Solution Approach 2:
The system creates a 2D image copy of the 3D scene and extracts distance information from this copy through image processing. Instead of directly measuring distance with specialized sensors, the system captures a visual representation and derives spatial information through computational analysis of the image data.
2Measurement precision
If multiple cameras are used for parallax analysis, then measurement precision is improved, but device complexity and volume increase
Solution Approach 1:
The patent segments the distance measurement task into two parts: a single camera captures the image, and the processing device performs computational segmentation of the image data to extract object boundaries and calculate distance. This replaces the need for multiple physical cameras with a single camera plus computational processing.
Solution Approach 2:
The system transitions from 3D spatial measurement (requiring multiple cameras in space) to 2D image plane analysis followed by 3D distance calculation. By capturing a 2D projection and using computational geometry to infer the third dimension, the system achieves depth measurement without adding spatial volume.
3Ease of operation
If bounding box size is used for distance estimation, then ease of operation is improved, but measurement precision deteriorates due to camera tilt and object rotation
Solution Approach 1:
The system uses feedback from the detected bounding box characteristics (position, size, orientation) to iteratively refine distance estimates. The processing device analyzes the bounding box data, detects errors due to tilt or rotation, and applies corrections to improve measurement accuracy while maintaining the simplicity of the initial bounding box approach.
Solution Approach 2:
The patent changes the parameters used for distance calculation from simple bounding box dimensions to corrected parameters that account for camera tilt angle and object rotation angle. By introducing these additional parameters and their correction factors, the system maintains ease of operation while significantly improving measurement precision.
Data Source
AI summary
A system for measuring a distance of an object. The system includes a camera and a processing device connected to the camera. The camera captures an image that is received by the processing device. The processing device runs an image recognition algorithm that, for each object recognized in the acquired image, identifies a type to which the object belongs, and defines a bounding box that encloses the object. The processing device also determines a reference dimension of the bounding box. The computing device corrects the reference size of the bounding box of the recognized object by applying a correction criterion selected according to the value of at least one pre-stored parameter associated with the type of the recognized object and at least one feature of the bounding box of the recognized object. Then, the processing device calculates the distance to the object according to the reference dimension.