Fish Stock Assessment | ||||||||
Ground truth for solar panels bounding boxes (Left) and prediction using deep learning-based object detection model (Right) using Worldview-3 images (0.3 m)over southern Germany (Source: Maxar). | ||||||||
Product Category | ||||||||
|
| |||||||
Financial Domains | ||||||||
| ||||||||
User requirements | ||||||||
UN37: Projection of risk to portfolio assets into future | ||||||||
Description | ||||||||
VHR satellite imagery can assess the condition of solar panels, detect anomalies or defects, and evaluate the overall energy generation of the installation. By comparing historical data, it becomes possible to identify changes in performance over time and address maintenance or operational issues promptly. In addition, satellite imagery can be used to evaluate land use changes that affect the performance of solar panels such as shadows from tall buildings and vegetation cover. | ||||||||
Spatial Coverage Target | ||||||||
Asset level | ||||||||
Data Throughput | ||||||||
Rapid tasking Data availability |
|
| ||||||
PRODUCT SPECIFICATIONS | ||||||||
Main processing steps | After the acquisition and preprocessing of the optical VHR (< 0.5 m), deep learning-based object detection algorithms such as YOLO can be used to detect solar panels. First, the dataset of the VHR images would be divided into training, validation, and test datasets. Training and validation will be used to train and optimize the deep learning model, which would be used then for inference to detect solar panels in the test data (our interest). Subsequently, temporal image pairing and image registration would be applied to analyse changes in the solar panels. Then, change detection techniques should be applied to the detected solar panels to identify changes over time. In terms of monitoring vegetation cover over the solar panels, vegetation indices can be used with change detection techniques. | |||||||
Input data sources | Optical: VHR based on the availability like Pleiades 1A/1B & NEO, WorldView2&3, and SPOT6/7 Radar: N.A Supporting data: Solar panel datasets for deep learning models (if any) | |||||||
Accessibility | VHR imagery: commercially available on demand from EO service providers. | |||||||
Spatial resolution | Optical VHR: ≤ 0.5 m | |||||||
Frequency (Temporal resolution) | Optical VHR: Daily | |||||||
Latency | Daily | |||||||
Geographical scale coverage | Globally | |||||||
Delivery/ output format | Data type: Raster File format: GeoTIFF | |||||||
Accuracies | Thematic accuracy: 80-90% Spatial accuracy: 1.5-2 pixels of input data | |||||||
Constraints and limitations |
| |||||||
User's level of knowledge and skills to extract information and perform further analysis on the EO products. | Skills: Essential Knowledge: Essential |
P12: Monitoring Solar Panel Installations | |
Maturity score | |
Mean: 2.00 | STD: 0.82 |
Constraints and limitations · Cloud presence. · Panels integrated into complex rooftop configurations can be harder to identify due to varying angles and orientations. | |
Relevant user needs UN37: Projection of risk to portfolio assets into the future. | |
R&D gaps · The availability and size of solar panels dataset to train the deep learning model. · Higher costs as balancing higher spatial resolution (to detect small panels) with broader coverage (to monitor larger installations) can be challenging due to cost constraints. · The resolution of thermal sensors is insufficient at the solar panel level. · Price models for commercial EO data. | |
Potential improvements drivers · Provide more training datasets. · Higher-resolution thermal sensors. | |
Utilisation level review | |
Utilisation score | |
Mean: 3.00 | STD: 0.89 |
No utilisation Unawareness of the existence of this EO product. Low utilisation Medium utilisation Unawareness of the existence of the best available commercial EO product with better specifications. High utilisation | |
Critical gaps related to relevant user needs | |
|
This page has no comments.