Parabolic antennas enable long-distance, high-precision signal transmission and reception, making them widely used in the field of communications. In recent years, parabolic antennas have gained significant attention in space situational awareness applications due to advances in extraction techniques for geometric parameters(such as antenna pointing direction) from Inverse Synthetic Aperture Radar(ISAR) images. However, accurately and efficiently estimating the attitude of parabolic antennas remains a major challenge. Addressing the insufficient extraction of parabolic characteristics in current approaches, this paper proposes ISARAngleNet, a network that integrates scattering point enhancement with Transformer architecture. The method incorporates specialized scattering point enhancement modules within ResNet(Residual Network) residual blocks to extract robust antenna features, while introducing a Transformer model to overcome ResNet's limitations in global feature modeling. Additionally, we develop a loss function that combines dual-path attitude regression with scattering point smoothness constraints, enabling the network to better capture complex relationships between angular and scattering features. The anechoic chamber experiments demonstrate that, compared to the standard ResNet network, the proposed method achieves improvements of 32.85% and 24.5% in RMSE(Root Mean Square Error) and MAE(Mean Absolute Error), respectively, on the parabolic antenna datasets, thereby validating its superiority.