As a unique trading rule in China's A-share market, the price-limit system significantly influences price discovery and investor behavior. To explore and leverage this information, motivated by the theoretical mechanisms of market sentiment amplification and price reversal effects, we employ image recognition techniques. We innovatively propose two methods: the Line Convolutional Neural Network (LCNN) and the Point Convolutional Neural Network (PCNN). These methods visualize stock data—including trading volume, prices, and closing limit status from 5,353 A-shares since 2010—into trend charts that explicitly incorporate either price-limit state lines or prompt points for prediction. Empirical results demonstrate that, compared to traditional factor models and methods using structured data, our models incorporating visualized price-limit information exhibit significant advantages in both prediction accuracy and portfolio returns. The predictive performance is stronger when stocks exhibit characteristics such as high turnover, high volatility, high transaction friction, low beta, and irrational market participants. This research validates the unique value of "visualizing" key institutional information for quantitative prediction and provides empirical evidence for optimizing China's distinctive market regulations, such as its price-limit and short-selling systems, ultimately promoting fairer pricing, market efficiency, and sustainable development in the Chinese financial market.