目的 研究基于粒子群算法优化 BP 神经网络对显示器色彩空间转换的预测准确性的方法。方法 主要通过数据归一化处理、改进最大限制速度、惯性常数和适应度函数来优化 BP 神经网络的权值和阈值,以缩小其分布范围,再用 BP 神经网络法进行色差预测。 结果 改进粒子群算法优化 BP神经网络预测模型,测试 20 次得到色块平均色差为 2 . 8526 ,最小平均色差为 2 . 0453 。 结论 该方法大大降低了 BP 神经网络预测模型陷入局部极小值的可能性,对显示器色彩空间转换具有较好的非线性拟合能力和更高的预测准确性。

Objective To study the method for prediction accuracy of color space conversion of monitor in BP neural network optimized based on particle swarm optimization. Methods The weights and threshold of the BP neural network were mainly optimized through the improvement of data normalization and maximum limiting speed, inertia constant and the fitness function, in order to reduce their distribution range, and predict the color difference through the BP neural network method. Results Through improved particle swarm optimization algorithm of BP neural network prediction model, after 20 times of tests, the average color difference reached 2. 8526, and the minimum average color difference reached 2. 0453. Conclusion The results showed that the method could greatly reduce the probability of BP neural network prediction model falling into local minima, resulting in good non-linear fitting capability and higher prediction accuracy in color space conversion.