| dc.description.abstract | A moving target will change the coefficients of the synthetic aperture radar (SAR) chirp signals. Vice versa, the velocity could be inferred from the SAR images or signals. This paper proposes two algorithms for estimating a moving target’s velocity in the SAR image or signal. Topic-1 describes the velocity estimation of a moving target in the SAR signals; Topic-2 describes that in the SAR images.
Topic-1 describes a new approach for estimating the Doppler coefficients from a finite number of noisy discrete-time observations, which are functions of the speed variation of target/radar. The approach adopts the finite difference method to estimate the chirp signal coefficients. It is motivated with the concept of the HAF. But the finite difference method directly calculates the phase of the signal. The finite difference method, with respect to the HAF, replaces the correlation operations with addition and average operations. That reduces the computation load. A closed-form expression is derived that describes the relationship between the phase differences and the chirp signal parameters. The difference method could derive the phase differences, but cause the deterioration of the signal variance. The finite difference method is a good way to reduce the noise deviations, but the measurable spans will become smaller. The measurable span could be improved by adopting a phase unwrapping method, as proposed in Topic-1. Unwrapping the phase differences could recover the chirp signal coefficients from bias estimates. The maximum measurable span of the coefficients will be significantly larger. The statistical analysis for the finite difference estimation proves that the variance could attain the Cramer-Rao lower bounds in higher SNR. In conclusion, our algorithm can reduce the computational complexity and remove the effect of the signal amplitude variation.
A moving ship on SAR image produces multiscale wake with a characteristic linear V-shaped pattern. Detection of the wake can provide substantial information about the ship, such as its size, direction and speed of movement. In general, ship-generated wakes in SAR images are associated with high sea clutter, which causes the deterioration of detection performance. Topic-2 presents a hybrid method that combines the wavelet technique and the Radon transforms technique to detect the ship wake. The wavelet technique is first applied to generate a set of multiscale images. An orthogonal basis function is adopted so that three high-pass images in horizontal, vertical and diagonal direction are generated for each resolution scale. Then a spatial correlator is applied among the moduli of different scale, where modulus images are formed from three high-pass images. The output of the correlation process is shown to be highly representative at ship wake edges. Comparisons with other methods indicate the superior performance of the present approach in that not only are the wakes detected but also the V-shaped pattern is well pre-served. The second stage of the method involves the application of the Radon transform technique to estimate the V-opening angle from the detected ship wakes. Compared with a direct Radon transform, the greater effectiveness of the proposed scheme is demonstrated in a terms of efficiency as well as reliability for ship wake detection in noisy backgrounds. | en_US |