This paper proposes a method based on target and shadow regions for Synthetic Aperture Radar (SAR) target recognition under complex conditions. Zernike moment features are extracted from the target and shadow regions segmented in SAR images to describe geometric shape distributions of targets. Both the target and shadow regions can be used to analyze the shape of the target，which have good correlation. Therefore，joint sparse representation is used to comprehensively characterize the two Zernike moment feature vectors. Based on outputs from joint sparse representation，the reconstruction errors for the target and shadow regions achieved by different training classes are calculated and the target category is determined based on the principle of the minimum error. The integration of target and shadow regions can more comprehensively reflect the geometric shape information of the target in SAR images，which is helpful for discriminating different categories. Based on MSTAR dataset samples，one standard operating condition and three extended operating conditions including configuration difference，pitch angle difference and noise interference are set up for experimental analysis and comparison validation. The results show the performance superiority of the proposed method.