Besides the advantages such as the fast response speed, non-contact measurement and so on, the multiangle dynamic light scattering (MDLS) can give better particle-size distribution (PSD) because it can get more information about scattering light compared to single-angle light scattering technolgy, which can reflect the characteristic parameters of particles, and the improvement of the accuracy of MDLS depends largely on appropriate PSD inversion methods. Based on Phillips-Twomey (PT) algorithm, the Recursion Nonnegative Phillips-Twomey (RNNPT) algorithm is proposed, which uses the MR-L-curve to estimate the regularization parameter, recursion algorithm to obtain the weighting coefficients and the addition of the nonnegative constraint. Through the exact determination of the weighting coefficients, RNNPT algorithm can improve the accuracy of MDLS PSD inversion algorithm. When the light intensity autocorrelation function noise range is from 0.01% to 1%, the results show that the inverted PSDs of the two unimodal distribution and the bimodal distribution are comparatively desired with the noise level under 0.1%. In contrast to recursion Phillips-Twomey (RPT) and the existing recursion Tikhonov (RT) algorithm, the calculation results show that the weighting coefficient ratio obtained from RNNPT algorithm is minimum and the inverted PSD is closest to the theoretical one.