通过多尺度变换实现了反演范围的自适应调整, 使其更接近真实范围。分别采用反演范围固定算法与自适应算法对200～600 nm单峰和200～900 nm双峰分布颗粒的模拟相关函数进行了反演, 结果表明：自适应算法的结果更接近理论分布, 抗干扰能力更强。相对于固定算法, 单峰分布颗粒最多可缩小峰值误差4.73%,缩小峰宽误差185 nm。双峰分布颗粒在0～0.001噪声水平时, 峰值误差分别小于11.33%,12.45%, 峰宽误差分别小于35,160 nm, 而固定算法在噪声水平大于0.000 1时, 难以得到合理的反演结果。反演范围自适应调整方法能够有效优化粒径反演结果。

By the multi-scale transform inverse range can be self-adaptively adjusted so that inverse range is closer to the true range. The simulative autocorrelation functions (ACF) of unimodal distribution particles of 200~600 nm and bimodal distribution particles of 200~900 nm were respectively inversed by the fixed and the self-adaptive algorithm of inverse range. Comparing to the inverse results of fixed algorithm, that of the self-adaptive algorithm is much closer to theory distribution and its tolerance of noises is stronger, for unimodal distribution particles, it can at most reduce peak value error 4.73% and peak width error 185 nm, for bimodal distribution particles, when noise levels are 0～0.001, its peak value errors are less than11.33%,12.45% and its peak width errors are less than 35,160 nm, however, when the noise level is greater than 0.0001, fixed algorithm can't get the reasonable results. Therefore, this self-adaptive adjustment method of inverse range can effectively optimize the inverse results．