Sonar Display Object-Type Correlation via AI Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The interpretation of sonar data is often challenging due to its complexity and requires significant knowledge and experience, making it difficult for novice users to distinguish between objects and identify their types, especially when multiple objects are represented in the data.
Innovation Solution
The system employs artificial intelligence to refine models for determining object characteristics using historical comparisons between sonar data and additional data from sources like radar, sensors, and maps, improving the accuracy of object-type estimation and providing easy-to-understand displays that emphasize important objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If sonar data is presented with representations of multiple objects, then the quantity of information provided is improved, but the ease of operation deteriorates due to difficulty in distinguishing between objects and identifying their types
Solution Approach 1:
The patent applies color coding to different object types in sonar displays, where each object category (fish, structure, marine mammal, etc.) is assigned a distinct color. This allows users to quickly distinguish between multiple objects and identify their types without requiring extensive interpretation experience, directly resolving the contradiction between providing complete object information and maintaining ease of operation.
Solution Approach 2:
The patent segments the complex sonar data by separating objects into distinct categories and presenting them with unique visual identifiers. By dividing the information into manageable segments (different object types with different colors/symbols), the system maintains complete object information while making it easier for users to process and interpret multiple objects simultaneously.
2Device complexity
If traditional sonar display methods are used, then the device complexity is minimized, but the measurement precision deteriorates in terms of object-type identification accuracy
Solution Approach 1:
The patent introduces an intermediary processing layer between the sonar transducer and the display that automatically classifies objects using machine learning models. This intermediary component analyzes sonar return data, compares it against trained models, and assigns object types with high accuracy. The display itself remains relatively simple, showing pre-classified objects with their identified types, thus achieving high measurement precision without significantly increasing device complexity.
Solution Approach 2:
The patent replaces manual object identification (mechanical/expert system) with automated machine learning-based classification. Instead of relying on user expertise to interpret sonar data, the system uses trained AI models to automatically identify object types, achieving superior measurement precision while keeping the display interface simple and user-friendly.
3Measurement precision
If machine learning models are used to determine object characteristics, then the measurement precision is improved, but the device complexity increases due to model training and data processing requirements
Solution Approach 1:
The patent performs preliminary action by training machine learning models in advance using historical sonar data and object characteristics. The trained models are then deployed to the sonar system, where they automatically classify new objects without requiring real-time complex computations. This preliminary training phase separates the complex model development from the operational phase, achieving high measurement precision while keeping the operational device complexity manageable.
Solution Approach 2:
The patent uses copying by creating simplified representations of complex machine learning models that can be deployed on embedded systems. Instead of implementing full-scale AI systems, the patent uses pre-trained model copies or simplified versions that maintain adequate accuracy while reducing computational requirements and device complexity for implementation.
Data Source
AI summary
A system for analysis of sonar data is provided comprising sonar transducer assembl(ies), processor(s), and a memory. The memory includes computer program code that is configured to, when executed, cause processor(s) to receive sonar data, where an object is represented within sonar data, and additional data from a data source other than the sonar transducer assembl(ies). The processor(s) further determine object characteristic(s) of the object using sonar data and additional data, and determine an estimated object-type for the object represented within sonar data using the object characteristic(s). The processor(s) further generate a sonar image based on sonar data, cause display of the sonar image, and cause provision of an indication of the estimated object-type so that the indication of the estimated object-type is correlated to the object representation in the sonar image.