Estimating Vessel Detection Features
For vessels detected in Sentinel-2 Optical Imagery, Skylight additionally estimates length and heading for the vessel based on the image. This is particularly helpful for analyzing “dark” vessels where no additional information is available via AIS.
Length Estimation
Here is the Confusion Matrix for our model that predicts vessel length. “Actual Category” is the length as reported in the AIS data correlated with the detection, while “Predicted Category” is the length the model determined from the image itself. Examples of how to interpret this matrix:
- 73% of vessels with actual length between 100-150m are predicted with some length value falling between 100-150m
- 89% of the vessels predicted to be between 10–20 m are actually within one bucket on either side (0–30 m)
Predicted Category |
|||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Total | |||||||||||
| Actual Category | 495 | ||||||||||
| 4997 | |||||||||||
| 4669 | |||||||||||
| 5935 | |||||||||||
| 4912 | |||||||||||
| 4899 | |||||||||||
| 7631 | |||||||||||
| 10733 | |||||||||||
| 11279 | |||||||||||
| Total | 0 | 1487 | 5837 | 7327 | 5886 | 4860 | 7919 | 9898 | 12336 | 55550 | |
Note: This model is regressing the length (predicting continuous numerical values) and the confusion matrix here is produced by putting the ground truth and predicted length into buckets. This may also explain part of why the model performance is slightly worse when looking at correct category, but better when looking one bucket off. Small (meter-scale) errors often push a prediction into the adjacent bin, so exact-bucket accuracy falls while “within one bucket” accuracy rises.
Heading Estimation
- 79% of heading estimates are within 10 degrees of the ground truth heading
- The arrow on the top right of vessel image chips indicates the direction of the predicted heading (example right)
Speed Estimation
Here is the confusion matrix for our model that predicts speed. All values are in knots. “Actual Category” is the speed as reported by AIS data correlated with the detection, while “Predicted Category” is the speed the model determined from the image itself. Examples of how to interpret this matrix:
- 77% of vessels are correctly categorized
- If we assume 0-2 is “stationary” and >2 is “moving” (i.e. reduce this to a 2x2 matrix), then 94% of vessels are correctly classified as stationary vs moving
Predicted Category |
||||||
|---|---|---|---|---|---|---|
| Total | ||||||
| Actual Category | 3,217 | |||||
| 9,431 | ||||||
| 14,247 | ||||||
| 23,988 | ||||||
| Total | 12 | 12791 | 15606 | 22474 | 50,883 | |
Type Estimation
Here is the confusion matrix for ship type classification. “Actual Category” is the type as reported by AIS data correlated with the detection, while “Predicted Category” is the type the model determined from the image itself. Examples of how to interpret this matrix:
- The overall accuracy is 78%
- In terms of distinguishing fishing vessels from other vessel types, the recall is 70% and precision is 82%
Predicted Category |
||||||||
|---|---|---|---|---|---|---|---|---|
| Total | ||||||||
| Actual Category | 21,983 | |||||||
| 6,185 | ||||||||
| 3,230 | ||||||||
| 5,117 | ||||||||
| 5,975 | ||||||||
| 10,717 | ||||||||
| Total | 22,304 | 7,253 | 2,688 | 5,047 | 5,239 | 10,676 | 53,207 | |
Was this article helpful?