Movable object and control method thereof
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Movable objects face challenges in accurately determining their location amidst changing environmental factors, such as varying illumination, without relying on feature points that may be obscured or difficult to extract.
Innovation Solution
A method and system that utilize omnidirectional image acquisition, conversion to panoramic images, and signal vectorization, combined with machine learning algorithms like deep neural networks, to generate maps and determine location based on environmental signals and images, enabling the object to recognize its position despite environmental changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature points are used for location determination, then location accuracy can be achieved in stable environments, but location recognition fails when environmental factors change or feature points are obscured
Solution Approach 1:
The patent changes the parameters used for location determination from feature point coordinates to signal strength values and panoramic image characteristics. By transforming the input parameters from traditional visual features to radio signal strengths and panoramic visual data, the system achieves reliable location recognition in environments where traditional feature points are obscured or environmental conditions change.
Solution Approach 2:
The patent replaces the mechanical/visual system of feature point extraction and matching with a hybrid system combining signal processing and machine learning. Instead of relying on visual feature detection algorithms, the system uses signal strength vectorization and deep neural networks to determine location, substituting traditional computer vision methods with a more robust signal-based approach.
2Measurement precision
If traditional image processing methods are used, then processing speed is sufficient for simple environments, but accuracy decreases in complex dynamic environments
Solution Approach 1:
The patent adds a new dimension to image processing by converting 2D images into panoramic representations that capture 360-degree environmental information. This dimensional transformation from standard 2D image processing to panoramic spatial mapping enables the system to recognize locations more accurately in complex environments by utilizing omnidirectional visual data.
Solution Approach 2:
The patent performs preliminary actions by pre-processing images into panoramic format and pre-vectorizing signal strength data before location determination. By preparing the data in advance through panoramic conversion and signal vectorization, the system reduces computational complexity during real-time location recognition while maintaining high accuracy in dynamic environments.
3Loss of information
If omnidirectional images are converted to panoramic images, then location information is enhanced, but processing time and computational load increase
Solution Approach 1:
The patent performs panoramic conversion and signal vectorization as preliminary actions during system initialization or map building phase. By pre-processing the omnidirectional images into panoramic format and storing the processed data, the system avoids repeated heavy computation during real-time location determination, thus reducing processing time while maintaining complete environmental information.
Solution Approach 2:
The patent creates panoramic image copies and signal strength vector representations that can be stored and reused for multiple location determination queries. By generating these processed representations in advance and storing them in a database, the system minimizes repeated computational operations while preserving complete environmental information for accurate location recognition.
Data Source
AI summary
Disclosed herein are a movable object and a movable object control method. The movable object control method may include acquiring an image of a movable object's surroundings, acquiring a signal having strength changing depending on a location of the movable object, generating a map on the basis of the signal and the image of the surroundings of the movable object, and applying the map to an algorithm to acquire a learned algorithm.