Abstract: Manual measurement on mango fruit dimensions is often confined to the time inefficiency and low accuracy. Specifically, the human judgment and simplified definitions of fruit shape can impact the morphology classification and subsequent packaging. Consequently, it is highly required for the dedicated platform to determine the morphological indicators for the accurate identification of mango fruit shapes. Such a platform can be expected to streamline the sizing process for the shape identification. Advanced techniques, such as three-dimensional (3D) scanning, have been successfully applied into the morphological analysis of other fruit types, demonstrating their potential to the high accuracy. Additionally, machine learning can also be integrated to significantly improve the classification of fruit shapes, in order to effectively mitigate the errors associated with human assessment. In this study, a 3D scanning platform was implemented to rapidly capture the geometric data of mango fruits. An accurate 3D model was constructed to facilitate the measurement on the morphological indices and shape identification. 75 mango varieties were selected with three replicates for each variety. 10 quantitative parameters were then measured, including volume, longitudinal and transverse diameters, as well as fruit shape indices using 3D scanning. A comparison was then made on the scanning and manual measurement. The results indicated that the precision of scanning was surpassed the manual ones, indicating the superiority of the former. Subsequently, eigenvalue, principal component, correlation and cluster analysis were conducted to categorize the mango varieties, according to their distinct fruit shapes. There were the rich genetic diversity and variability in the mango germplasm resources. Notably, the fruit shape index was taken as the critical indicators, together with longitudinal and transverse diameter. The comprehensive evaluation of fruit morphology was performed for the subsequent application of machine learning. The input parameters were then optimized to predict the modeling. A back-propagation (BP) neural network model was developed to classify the fruit shape, according to the SPSSPRO online data analysis platform. The architecture of model consisted of 100 hidden layers, with a learning rate set at 0.1 and a total of 1000 iterations. Furthermore, an success rate was achieved in the 94.12% accuracy, when testing on 68 samples and training with 157 samples. Moreover, the training set exhibited a precision and recall rate of 0.941, with a remarkably rapid recognition time of only 0.451 s. In conclusion, 3D point cloud data was utilized to provide the rapid, precise, and lossless morphological information for mango fruits. 3D point cloud maps were generated to further enhance the overall accuracy, such as fruit color information. Compared with the conventional, this approach can offer numerous advantages: rapid sampling, high accuracy, reduced susceptibility to external factors, and a non-contact measurement. A significant framework can also advance the field of fruit morphology analysis for future application in fruit classification and assessment.