⚡ electrical engineering

Improving the Estimation of Ship Length via ISAR

This paper presents the ISAR AutoTrack (IAT) algorithm, which leverages adaptive motion compensation velocity and autofocus techniques to accurately estimate ship aspect angles and lengths within 10 percent error, thereby reducing radar resource requirements and maintaining robustness against ship maneuvers.

Original authors: John R. Bennett

Published 2026-03-03

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: John R. Bennett

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: The "Blurry Photo" Problem

Imagine you are flying a plane over the ocean at night, trying to identify a ship below using a special radar camera. This camera doesn't take a normal photo; it creates an image based on how radio waves bounce off the ship. This is called ISAR (Inverse Synthetic Aperture Radar).

The main goal is to figure out how long the ship is. Why does this matter? Because a 200-foot fishing boat and a 1,000-foot aircraft carrier look very different. If you know the length, you know what the ship is.

The Problem:
To measure the length, the radar needs two things:

The "Raw" Length: How much space the ship takes up in the radar's view (Range Extent).
The Angle: Is the ship sailing straight toward you, straight away, or sideways?

Think of it like looking at a ruler. If you look at the ruler head-on (from the end), it looks like a tiny dot. If you look at it from the side, it looks long. If you guess the angle wrong, your calculation of the ruler's length will be wildly off.

Traditionally, the radar plane tries to guess the ship's angle by tracking it with a "searchlight" radar before zooming in. But this is like trying to guess a car's speed by watching it from a mile away through a foggy window. The guess is often wrong, leading to big errors in the length calculation.

The Solution: The "Auto-Track" (IAT)

The author, John Bennett, developed a new method called ISAR AutoTrack (IAT).

The Analogy: The Auto-Focus Camera
You know how a camera lens "hunts" back and forth to find the perfect focus? It does this by measuring how blurry the image is and adjusting until it's sharp.

Bennett realized that the radar system already does something similar when it takes the picture. To keep the ship's image sharp, the radar computer has to constantly adjust for the ship's movement and the plane's movement. It calculates a "motion compensation" value to cancel out the wobble.

The "Aha!" Moment:
Bennett realized that the errors in these adjustments contain the secret to the ship's true speed and angle.

If the radar thinks the ship is moving slower than it actually is, the computer has to "push" the image forward to keep it sharp.
If the radar thinks the ship is moving sideways when it's actually going straight, the computer has to "twist" the image.

By analyzing these tiny "pushes and twists" (the adaptive motion compensation data) that the radar already calculates to make a clear picture, the IAT algorithm can reverse-engineer the ship's true speed and angle. It's like a detective looking at the footprints left behind by a thief to figure out exactly how fast they were running, rather than just guessing from a distance.

How It Works (The Two Methods)

The paper describes two ways to solve the math puzzle:

The Search Method: The computer tries thousands of guesses (e.g., "Is the ship going 10 knots at a 30-degree angle? What about 12 knots at 32 degrees?") and picks the one that makes the radar's "push and twist" data look the most consistent.
The Analytical Method: The computer uses a direct math formula to calculate the answer instantly.

Both methods use a "smoothing" filter to ignore the noise of ocean waves. Imagine trying to hear a whisper in a storm; the filter helps the computer ignore the crashing waves so it can hear the ship's true movement.

The Results: From "Wild Guess" to "Pinpoint Accuracy"

The paper tested this on real data from commercial ships (like container ships and tankers).

Before (The Tracker): The radar's standard guess for the ship's angle was often off by 26 degrees. This is like trying to measure a ruler while looking at it from a weird angle; the length calculation was off by nearly 35%.
After (The IAT): The new method reduced the angle error to about 8 degrees. This brought the length calculation error down to roughly 16%, and in many cases, even closer to the goal of 10%.

The "Magic" of the Method:
The IAT doesn't need to watch the ship for 15 minutes to get a good guess. It works during the short time the radar is actually taking the picture (the "ISAR window"). This means it works even if the ship suddenly changes direction (maneuvers), which would confuse the old tracking methods.

The Hurdles (Why it's not perfect yet)

Even with this smart new tool, there are still some "foggy" situations:

The Mirror Effect (Multipath): If a ship is heading straight toward the radar, its large deckhouse can act like a mirror. The radar signal bounces off the deck, hits a crane, bounces back to the deck, and then to the radar. This creates a "ghost" image that makes the ship look longer than it is.
The Shadow Effect: If the ship is heading toward the radar, the front of the ship can cast a "radar shadow" over the back, hiding the stern. This makes the ship look shorter.
The Wave Problem: If the ocean is very rough, the ship bobs up and down violently. This confuses the "push and twist" data, making it hard to tell the difference between the ship moving and the waves moving it.

The Future: What's Next?

The author suggests that to make this system perfect, we need to:

Get Better Bearings: Use the radar to slightly wiggle its antenna to find the exact direction of the ship (like turning your head to see a bird better).
Fly Sideways: When looking at a target, the plane should fly slightly sideways (not just straight ahead) to get a better "side view" of the ship's movement.
Combine Forces: Mix this new math with 3D imaging techniques to filter out the "ghosts" caused by waves and mirrors.

Summary in One Sentence

This paper introduces a clever "auto-focus" trick that uses the radar's own internal adjustments to figure out exactly how a ship is moving, allowing it to measure the ship's length with much greater accuracy than before, even in rough seas.

1. Problem Statement

Accurately estimating the Length Overall (LOA) of ships is a critical parameter for maritime surveillance and Automatic Target Recognition (ATR). The standard method for calculating LOA using Inverse Synthetic Aperture Radar (ISAR) relies on the "Thin Ship Approximation":
$LOA = \frac{RE}{\cos(\text{Aspect})}$
Where $RE$ is the Range Extent (difference between the farthest and nearest scatterers) and $\text{Aspect}$ is the angle between the ship's heading and the radar line-of-sight.

The Core Challenge:

Tracker Inaccuracy: Traditional methods rely on large-scale search radars (trackers) to estimate the ship's course and speed to derive the aspect angle. However, these trackers often suffer from significant errors (RMS errors of ~26 degrees in course/aspect) due to multipath scattering, ship maneuvering, and the difficulty of associating detections over long periods.
Error Amplification: Because LOA calculation involves the secant of the aspect angle, small errors in aspect estimation lead to massive errors in LOA, particularly at higher aspect angles (e.g., >45°).
Operational Limitations: Existing ATR systems often restrict operations to low aspect angles (<45°) to avoid these errors, effectively ignoring nearly half of the expected operational scenarios. Furthermore, reliance on trackers makes the system vulnerable to "Dark Ships" (vessels with AIS off) and ship maneuvers during the tracking window.

2. Methodology: The ISAR AutoTrack (IAT) Algorithm

The paper proposes the ISAR AutoTrack (IAT) algorithm, which repurposes classical SAR (Synthetic Aperture Radar) autofocus techniques to improve aspect angle estimation directly from the ISAR processing stream.

Core Principles:
The IAT utilizes Adaptive Motion Compensation (Mocomp) data generated during the coherent ISAR imaging phase. It operates on two fundamental principles derived from the Map-Drift autofocus algorithm:

Mean Mocomp Velocity: The average value of the adaptive motion compensation velocity represents the error in the in-range component of the ship's velocity.
Mocomp Velocity Trend (Acceleration): The linear time trend (time derivative) of the adaptive motion compensation velocity represents the error in the cross-range component of the ship's velocity.

Algorithmic Approach:
The algorithm assumes the aircraft's motion is known precisely (via GPS/INS) and that the tracker provides a "first guess" of the ship's course and speed. It then refines these estimates by minimizing the difference between the deterministic motion compensation (calculated from the tracker guess) and the actual adaptive motion compensation derived from the ISAR data.

The IAT offers two implementation methods:

Search Method: Iteratively searches a grid of possible ship courses and speeds to find the combination that minimizes the variance of the motion compensation residual.
Analytical Method: Solves the motion compensation equation analytically to derive the course and speed directly.

Data Pre-processing:
To ensure accuracy, the algorithm includes an intelligent smoothing process to remove:

System errors and random noise (via median and three-point filters).
Ocean wave signatures (via an adaptive narrow-band filter using a "chapeau" function to estimate and remove wave-induced phase oscillations).

3. Key Contributions

Repurposing Autofocus: The novel application of SAR Map-Drift principles to maritime ship tracking, utilizing the adaptive Mocomp data usually discarded after image focusing.
Independence from AIS/Tracker: The algorithm can operate without Automatic Identification System (AIS) data or highly accurate initial tracker estimates, making it effective against "Dark Ships."
Robustness to Maneuvers: Since IAT operates within the short ISAR dwell time (tens of seconds), it is unaffected by ship maneuvers that occur over longer tracking windows.
Level-0 to Level-2 Implementation: The paper defines a roadmap for implementation:
- Level-0: Software-only update using existing Mocomp data.
- Level-1: Integration of adaptive beam pointing or multi-aperture techniques to correct initial bearing errors.
- Level-2: Estimation of bearing time-derivatives to resolve "blind spots" at head-on or tail-on azimuths.

4. Results and Validation

The author validated the IAT using two distinct datasets:

A. MSR2 Data (Maritime Surveillance Radar #2):

Dataset: 22 cases of commercial ships (container ships, tankers) with X-band radar data.
Tracker Performance: RMS Heading Error = 26.56°; RMS Speed Error = 4.02 m/s.
IAT Performance: RMS Heading Error reduced to 8.04°; RMS Speed Error reduced to 1.00 m/s.
LOA Accuracy:
- Tracker-based LOA Error: 35.3% (88.4 m).
- IAT-based LOA Error: 16.0% (40.1 m).
- Note: The remaining error is attributed to Range Extent (RE) issues like shadowing and multipath, not aspect angle errors.

B. SCATR Data (Small Craft ATR Program):

Dataset: US Coast Guard cutters (Tybee and Point Stuart) imaged by a shore-based radar.
Challenge: The stationary radar lacks the cross-range platform velocity needed for the linear term in the Map-Drift equation, creating a "partial blind spot."
Solution: The author developed an extrapolation algorithm to simulate the data as if collected by a moving aircraft (50 m/s cross-range velocity) at a longer range (10x).
Result: After extrapolation, the IAT achieved an LOA RMS error of 9.5%, meeting the paper's 10% goal.

Simulation Results:
Simulations confirmed that IAT maintains accuracy across aspect angles from 5° to 85°. Errors in cross-range position (bearing) were identified as the primary remaining source of error, suggesting that improved bearing estimation (Level-1) is the next critical step.

5. Significance and Future Recommendations

Significance:

Operational Efficiency: IAT allows for a major reduction in radar resources dedicated to long-term tracking, as it can generate accurate aspect estimates from short ISAR dwells.
Expanded Operational Envelope: By improving aspect angle accuracy, the algorithm enables reliable ship length estimation at high aspect angles (up to 85°), doubling the effective operational coverage compared to traditional tracker-dependent methods.
ATR Enhancement: Accurate aspect angles are essential for 3D-ISAR algorithms to distinguish between "Profile" and "Plan" views, which are critical for ship classification.

Recommendations for Future Development:

Adaptive Beam Pointing: Implement algorithms to vary the antenna azimuth during ISAR collection to find the bearing of maximum return power, reducing cross-range position errors.
Multi-Aperture Integration: Use phase comparisons from multiple apertures to refine bearing estimates.
Blind Spot Mitigation: For circular-scan radars looking directly ahead or behind, estimate the time derivative of the bearing to calculate cross-range velocity directly.
Data Collection: Conduct airborne flight tests with AIS recording to validate the algorithm across a wider variety of ship types and sea states, specifically targeting the "Level-1" implementation.

In conclusion, the paper demonstrates that by leveraging the adaptive motion compensation data inherent in ISAR processing, it is possible to significantly outperform traditional tracker-based methods in estimating ship aspect angles and, consequently, ship length, achieving a target accuracy of within 10% for a wide range of operational conditions.