Autonomous Target Localization Using Visibility-Guided Search Regions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional search strategies for unmanned vehicles, such as drones, incur excessive movement and sensing costs when sequentially searching a large area for a target object, as they often navigate through regions with poor visibility due to environmental conditions or suboptimal distances.
Innovation Solution
A system utilizing two machine learning models to determine the visibility of a target object in images captured by an unmanned vehicle, where the first model identifies visible regions and the second model navigates the vehicle to optimal sub-regions with higher probability of target presence, adjusting distance and sensing modes based on environmental conditions to maximize visibility and minimize costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If conventional sequential search strategies are used to cover a large search area, then the entire area is searched, but movement cost and sensing cost increase excessively
Solution Approach 1:
The patent divides the search area into multiple sub-regions and uses machine learning models to evaluate each sub-region independently. The search space is segmented into grid cells that are processed in parallel, allowing the system to focus computational resources on high-probability areas rather than sequentially scanning the entire area.
Solution Approach 2:
The patent performs preliminary analysis by capturing images of the entire search area first, then uses machine learning models to predict target probabilities for each sub-region before actual search navigation. This preliminary evaluation allows the system to plan an optimized search path that avoids low-probability areas, reducing both movement and sensing costs.
2Area of stationary object
If the unmanned vehicle navigates to regions with poor visibility, then more areas are covered, but target detection accuracy decreases
Solution Approach 1:
The patent dynamically adjusts the unmanned vehicle's distance from the search area based on environmental conditions and target probability predictions. The system uses multiple sensing modes with different distances, selecting the optimal distance for each sub-region to maximize visibility and detection accuracy while maintaining comprehensive coverage.
Solution Approach 2:
The patent changes key parameters including sensing distance, sensing mode, and search priority based on environmental conditions and machine learning predictions. The system adjusts these parameters dynamically for each sub-region to optimize both coverage and detection accuracy, avoiding fixed search patterns that would compromise either metric.
3Loss of energy
If the unmanned vehicle focuses on high-probability regions only, then movement cost is reduced, but target detection reliability may decrease
Solution Approach 1:
The patent implements feedback mechanisms where machine learning models continuously update target probability predictions based on observed images and environmental conditions. The system uses this feedback to refine search priorities and adjust navigation decisions in real-time, ensuring that high-probability regions are targeted while maintaining the ability to discover new targets in previously low-probability areas.
Solution Approach 2:
The patent dynamically changes search parameters including distance and sensing mode based on environmental conditions and target probability predictions. This allows the system to maintain reliability by adapting to changing conditions while focusing resources on high-probability regions, rather than using fixed parameters that would compromise either efficiency or reliability.
4Measurement precision
If multiple sensing modes are used to maximize visibility, then detection accuracy improves, but sensing cost increases
Solution Approach 1:
The patent applies different sensing modes locally to different sub-regions based on their specific characteristics and target probability predictions. Rather than using a single sensing mode for the entire search area or switching modes continuously, the system selects appropriate sensing modes for each sub-region, optimizing detection accuracy where needed while reducing sensing cost in low-priority areas.
Data Source
AI summary
A method for localizing a target object may include applying, to a first image captured by an apparatus, a first machine learning model trained to determine whether the target object is visible in the first image depicting a first region of a search area. In response to the target object being absent from the first image, a second machine learning model may be applied to identify, based on the target object as being visible and/or nonvisible in one or more portions of the first image, a second region of the search area corresponding to a sub-region within the first region of the search area having a highest probability of including the target object. A command may be sent to the apparatus to navigate the apparatus to the second region of the search area in order to capture a second image of the second region of the search area.