Aluminum alloy materials have been widely used in many industrial fields due to their advantages of easy processing, corrosion resistance and recyclability, and has become the second largest metal material after steel. In the face of the scarcity of mineral resources and the large number of aluminum products reaching the service life, the recycling of aluminum alloy is particularly important. Recycled aluminum is of great significance to the sustainable development of economy, environment and energy. At present, it is difficult to classify scrap aluminium alloys efficiently because of their various types and shapes, which leads to the degradation of high-quality aluminium alloys and the direct casting of aluminium ingots. Aluminum for aerospace is mainly made of 2xxx and 7xxx aluminum alloys. Due to the special use environment, aviation aluminum products have good quality and high value, and degraded use will cause huge waste. This paper automatically classifies aluminum alloys of aluminum 3xxx, 7xxx and A356 into a research target. Based on laser induced breakdown spectroscopy technology, an automated classification detection experimental platform was built. Image recognition is used to locate the dynamic sample, which is accurately captured by laser induced breakdown spectroscopy (LIBS). For the single-pulse LIBS spectral signal, the multi-dimensional Gaussian probability density distribution discriminant function of three series of aluminum alloys is established. Completed high-efficiency, high-precision continuous classification detection of 2xxx, 7xxx and A356 aluminum alloys. The experimental results show that the recognition time of the material in the 1.2 m·s^-1 transmission process is 18 ms, the laser excitation control deviation is less than 20.83 ms, and the minimum size of the test sample is 25 mm. For the three series of aluminum alloy samples with a height difference of less than 3 mm, the average prediction classification accuracy of the multi-dimensional Gaussian probability density distribution method can reach 99.15%, and the average modeling time only takes 7 ms. Compared with the widely used support vector machine (SVM) classification method, the prediction accuracy is equivalent, and the modeling time is increased by order of magnitude. The generalization ability of the classification prediction is good and the modeling efficiency is high. This study validated the effectiveness of automated rapid classification and detection of aluminum alloys based on laser-induced breakdown spectroscopy. It provides a theoretical and technical basis for the establishment of a fully automated scrap metal sorting system.