Automatic Water Surface Labelling via Sensor Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current water surface detection systems rely on manual image labelling, which is time-consuming, expensive, and error-prone, necessitating a method for automatic image labelling to generate large datasets for training machine learning algorithms efficiently.
Innovation Solution
A method that receives image data and water surface extension data, matching it to the spatial relationship between the camera and the water surface area, allowing for automatic labelling of the water surface within the image, thereby facilitating the creation of a large dataset for machine learning algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual image labelling is used to generate training datasets, then labelling accuracy can be maintained, but the process becomes time-consuming and expensive
Solution Approach 1:
The patent uses sensor data (LiDAR, radar, or other imaging devices) to create accurate copies or representations of the water surface area, which are then projected onto the image to generate labels. This copying approach maintains high labelling accuracy while automating the process, eliminating the need for time-consuming manual pixel-by-pixel annotation.
Solution Approach 2:
The patent introduces sensor data as an intermediary between the physical water surface and the image labels. The sensor data serves as a mediator that provides accurate spatial information about the water surface, which is then used to automatically generate labels for the image pixels corresponding to the water surface area.
2Reliability
If manual image labelling is used to generate training datasets, then label quality can be ensured, but the cost increases significantly
Solution Approach 1:
The patent replaces expensive manual labelling with automated copying of water surface boundaries from sensor data. By copying the water surface area information from LiDAR, radar, or other imaging sensors and projecting it onto the image, the system ensures label quality while dramatically reducing the cost of dataset generation.
Solution Approach 2:
The system performs self-service labelling by automatically generating labels through the integration of sensor data and image data. The automated process eliminates the need for human annotators, thereby ensuring consistent label quality while reducing costs associated with manual labour.
3Reliability
If large quantities of labelled images are generated manually, then machine learning algorithm performance can be improved, but the productivity remains low
Solution Approach 1:
The patent enables rapid generation of large quantities of labelled images by automatically copying water surface area information from sensor data to multiple images. This automated copying process can be applied to entire datasets simultaneously, dramatically increasing the productivity of dataset generation while maintaining the quality needed for improving machine learning algorithm performance.
Solution Approach 2:
The patent performs preliminary action by pre-acquiring sensor data that contains water surface area information before the labelling process. This pre-collected sensor data can then be rapidly projected onto multiple images, enabling fast batch processing and significantly increasing the productivity of training dataset generation.
Data Source
AI summary
A method of labelling a water surface within an image is provided. The method includes: receiving image data of the image generated by a camera, the image including at least one water surface; receiving water surface extension data, the water surface extension data being representative of an area over which the water surface extends in the real world; matching the water surface extension data to the image data based on a spatial relationship between the camera and the area of the water surface; and labelling the water surface in the image based on the matched water surface extension data.