Aerial Map Object Detection with Ground-Level Shadow Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing aerial imagery often struggles to accurately distinguish between map objects and their shadows, leading to unreliable map data due to similar sizes and shapes, especially in high-shadow environments, which complicates applications like automated driving and vehicle routing.
Innovation Solution
A system and method that utilizes aerial imagery in conjunction with ground-level image data from proximate cameras to identify and confirm the presence and dimensions of map objects by analyzing shadows, employing techniques such as machine learning and artificial intelligence to differentiate between objects and shadows.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If aerial imagery is used to identify map objects, then coverage area and efficiency are improved, but accuracy deteriorates due to inability to distinguish objects from shadows
Solution Approach 1:
The patent combines aerial imagery with ground-level images from proximate cameras to create a composite identification system. By merging data from multiple sources (aerial view and ground-level views), the system achieves both wide coverage and high accuracy in distinguishing map objects from shadows.
Solution Approach 2:
The patent introduces ground-level images as an intermediary data source to resolve the ambiguity in aerial imagery. These ground-level images serve as a mediator that provides additional perspective information, enabling accurate distinction between objects and shadows while maintaining the efficiency of aerial survey coverage.
2Measurement precision
If ground-level cameras are added to verify map objects, then identification accuracy is improved, but system complexity increases
Solution Approach 1:
The patent segments the imaging system into two functional components: aerial imagery for broad coverage and ground-level cameras for verification. This segmentation allows each component to perform its specialized function efficiently, reducing overall system complexity while maintaining high accuracy through coordinated operation.
Solution Approach 2:
The patent applies ground-level camera verification selectively rather than universally. By using ground-level cameras only when needed to verify specific map objects or resolve ambiguities, the system achieves high accuracy without the full complexity of continuous multi-camera operation.
3Reliability
If shadow areas are excluded from aerial image analysis, then false positives are reduced, but information loss increases
Solution Approach 1:
The patent inverts the traditional approach by not excluding shadow areas but rather using ground-level images to illuminate and analyze these previously problematic regions. This inversion allows shadow areas to be fully utilized as information sources while their ambiguity is resolved through complementary ground-level perspectives.
Solution Approach 2:
The patent changes the analysis parameters for shadow areas by incorporating ground-level image data. Instead of treating shadow regions as low-confidence or excluded zones, the system transforms them into high-value information sources by analyzing them from multiple perspectives with different lighting and angle parameters.
Data Source
AI summary
A method, apparatus, user interface, and computer program product are provided to receive an aerial image of a geographic region, identify at least one potential map object within the aerial image, determine at least one shadow area associated with the at least one potential map object within the aerial image, identify at least one camera proximate to the location of the at least one potential map object, capture image data of the area proximate to the least one potential map object via the at least one camera, and generate location data for the at least one map object or a road attribute based at least in part on the aerial image and image data of the area proximate to the least one potential map object or road attribute.