Improving the Estimation of Ship Length via ISAR

A method for estimating the aspect angle of ships at sea from an ISAR is developed. The ISAR AutoTrack (IAT) algorithm uses the information from the adaptive motion compensation velocity to improve the tracker estimation of the ship aspect angle and thus to improve the estimation of ship length. The IAT is based on classical methods of autofocus for synthetic aperture radar. The average mocomp velocity yields the error in the in-range component of the ship velocity; the linear time trend of the velocity determines the cross-range component of the ship velocity. The IAT has two methods for implementing the algorithm, the Search and Analytical methods. Both methods benefit from an intelligent smoothing process that removes system errors, random noise, and ocean waves. The goal of the IAT is to measure ship length to within 10 percent over all azimuth angles and ranges relative to the aircraft and for (unsigned) aspect angles from 5 to 85 degrees. Using the IAT allows a major reduction in the radar resources dedicated to tracking; and since the IAT creates its estimates during the ISAR time window it is unaffected by ship maneuvers. Recommendations for further development and testing of the IAT are presented.

💡 Research Summary

The paper introduces the ISAR AutoTrack (IAT) algorithm, a novel approach for estimating ship length (Length Overall, LOA) from airborne ISAR imagery with an accuracy goal of within 10 % across a wide range of aspect angles (5°–85°). Traditional ISAR processing estimates LOA by measuring the range extent (RE) of a ship and multiplying by the secant of the aspect angle (LOA = RE·sec (AspectU)). However, errors in the aspect‑angle estimate can dominate the LOA error, especially at low aspect angles where RE measurement is also difficult.

IAT tackles this problem by exploiting the full motion‑compensation (MOCOMP) information that modern ISAR systems already compute. The average MOCOMP velocity provides the error in the range‑direction component of ship motion, while the linear time trend of the velocity yields the cross‑range component. By jointly estimating these two components, IAT can infer the ship’s unsigned aspect angle (AspectU) without relying on a conventional tracker’s estimate.

Two implementation pathways are described:

1. Search Method – a hypothesis‑testing approach that scans the admissible aspect‑angle interval, evaluates an autofocus quality metric for each candidate, and selects the angle that yields the sharpest focused image.

2. Analytical Method – a closed‑form solution derived from the linear relationship between the MOCOMP parameters and the image rotation rates (aspect and tilt).

Both pathways incorporate an “intelligent smoothing” stage that removes systematic system biases, random thermal noise, and wave‑induced jitter. The smoothing combines temporal averaging with a Kalman‑filter‑like predictor, which is especially effective at low aspect angles where RE estimation suffers from shadowing and multipath scattering.

The authors present a theoretical error model showing that, if the aspect‑angle error were a constant 3 °, the fractional LOA error would follow a tan(AspectU) curve, reaching about 5 % at 45°. Simulations that include realistic cross‑range beam‑width effects, however, produce a much flatter error curve, indicating that the cross‑range component becomes easier to estimate at higher aspect angles, thereby reducing LOA error.

Experimental validation uses 43 “MSR2” cases collected with a single‑aperture airborne ISAR system. Ground‑truth aspect angles are derived from AIS/GPS ship heading reports, while RE is measured from high‑resolution ISAR images. The mean ship length in the dataset is 250 m; IAT achieves an RMS LOA error of 20 m (≈8 %). This performance is comparable to earlier work that relied on external AIS/GPS data, but the key advantage is that IAT can operate autonomously when such data are unavailable (e.g., “dark ships”).

The paper also discusses limitations: (i) multi‑scatterer ships (large superstructures) can cause RE under‑ or over‑estimation due to shadowing or multipath; (ii) synchronization delays between the tracking mode and the ISAR mode can corrupt the MOCOMP parameters; (iii) rapid course changes or high speeds violate the linear‑trend assumption; and (iv) the current formulation assumes a single antenna and single frequency, limiting direct applicability to emerging multi‑beam or MIMO 3‑D ISAR systems.

Future work suggested includes: integrating real‑time AIS/GPS for initial angle seeding, employing deep‑learning‑based denoising to handle non‑linear wave and multipath effects, extending the algorithm to multi‑aperture/MIMO platforms to directly estimate 3‑D rotation rates (thus removing the need for a mean aspect‑angle estimate), and implementing hardware‑accelerated versions (FPGA/GPU) for real‑time deployment on manned or unmanned aircraft.

Overall, IAT represents a significant step toward more robust maritime surveillance: by fusing motion‑compensation data with autofocus techniques, it delivers ship‑length estimates that meet operational accuracy requirements while reducing reliance on external navigation aids and decreasing radar resource consumption.