The paper presents the aerodynamic characterization of a low-speed unmanned aerial vehicle, with cropped delta planform and rectangular cross section, at and around high angles-of-attack using flight test methods. Since the linear models used for identification from flight data at low and moderate angles of attack become unsuitable for accurate parameter estimation at high angles of attack, a non-linear aerodynamic model has to be considered. Therefore, the Kirchhoff's flow separation model was used to incorporate the non-linearity in the aerodynamic model in terms of flow separation point and stall characteristic parameters. The Maximum Likelihood (ML) and Neural Gauss-Newton (NGN) methods were used to perform the parameter estimation on one set of low angle-of-attack and one set of near-stall flight data. It is evident from the estimates that the NGN method, which does not involve solving equations of motion, performs on a par with the classical ML method. This may be attributed to the reason that NGN method uses a neural network which has been trained by performing point to point mapping of the measured flight data. This feature of NGN method enhances its application over a wider envelope of high angles of attack flight data.