Legal judgment prediction (LJP) is an important task in legal AI that aims to predict outcomes based on the factual description of a case. Most current approaches are data-intensive and focus on introducing external knowledge to improve judgment prediction accuracy. However, they do not address the legal data imbalance problem well, and there is still a large gap between the prediction performance of low-frequency legal events and the overall average performance. Therefore, we propose a method that integrates meta-learning and contrastive learning for Chinese legal judgment prediction (MJP). The approach leverages transferred knowledge from the source category to discern the target category, achieved through constructing support sets and query sets. Additionally, it aligns legal text representations with optimizing metrics and trains the model to effectively distinguish between similar legal factual descriptions. In this way, the model is able to quickly learn effective feature representations from a small amount of labeled data and generalize to new categories. Experimental results show that our proposed approach effectively improves the performance of LJP, not only achieving better than the current model results on two real datasets but also surpassing the performance of other models under low resource settings.