Single-Camera Object Distance Estimation with Tilt 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 addressed by existing technologies that rely on multiple sensors or machine learning, and are often bulky, energy-intensive, or emit harmful radiation.
Innovation Solution
A system using a single camera and computer vision algorithms to estimate distance by correcting errors through pre-stored parameters for bounding box dimensions, applying tilt and rotation corrections based on object type and image analysis, without requiring additional sensors or machine learning, ensuring low computational cost and real-time operation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple sensors (RADAR, LIDAR) are used for distance estimation, then measurement precision is improved, but device complexity and weight increase
Solution Approach 1:
The patent extracts only the essential information needed for distance estimation from the camera image - specifically the bounding box dimensions and object type classification - while discarding unnecessary visual data. This allows single-camera systems to achieve automotive-level distance estimation accuracy without requiring complex multi-sensor setups, directly resolving the contradiction between measurement precision and device complexity
Solution Approach 2:
The patent replaces physical sensing mechanisms (RADAR, LIDAR) with a computational vision system that uses camera images and pre-stored geometric parameters. By substituting mechanical sensors with algorithmic processing of visual data, the system achieves similar or superior distance estimation precision while dramatically reducing device complexity and weight
2Measurement precision
If machine learning algorithms are used for distance estimation, then measurement precision is improved, but use of energy increases
Solution Approach 1:
The patent performs all complex computational operations during an offline training phase, where the system learns and stores the relationship between bounding box dimensions, object types, and distance parameters. During actual operation, the system only needs to retrieve pre-computed lookup tables and apply simple geometric formulas, reducing real-time energy consumption while maintaining high measurement precision
Solution Approach 2:
The patent replaces expensive, energy-intensive machine learning models with lightweight lookup tables and simple geometric calculations. The complex pattern recognition is encapsulated in pre-stored parameter tables that can be quickly queried during operation, significantly reducing the energy required for distance estimation while achieving comparable or superior accuracy
3Device complexity
If camera tilt and object rotation are not corrected, then device complexity is reduced, but measurement precision deteriorates
Solution Approach 1:
The patent dynamically adjusts the interpretation of bounding box parameters based on detected object type and estimated camera orientation. By changing how the same visual data is processed according to contextual parameters, the system corrects for tilt and rotation effects without adding complex hardware, maintaining simplicity while improving measurement precision through adaptive parameter adjustment
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
The present invention relates to a system for measuring a distance of an object (01-03). The system includes a single camera (2) and a processing device (3) connected to the camera (2). The camera (2) captures an image that is received by the processing device (3). The processing device (3) runs an image recognition algorithm that, for each object recognised in the acquired image identifies a type to which the object belongs, and defines a bounding box that encloses the object. In addition, the processing device (3) determines a reference dimension of the bounding box. In particular, the portable device corrects the reference dimension of the bounding box of the recognised object by applying a correction criterion selected according to the value of at least one pre-stored parameter associated with the type of the recognised object and at least one feature of the bounding box of the recognised object. Finally, the processing device (3) calculates the distance to the object according to the reference dimension thus corrected and at least one comparison dimension stored in the system.