Neutral points are one of important distribution characteristics of the atmospheric polarization pattern, which have important value on polarotactic navigation, atmospheric environmental monitoring, remote sensing, et al.. Based on the analysis of the degree of polarization′s and the neural line′s distribution characteristics, two recognition methods of neutral points are proposed. The experimental results show that, both of these methods can recognize the neutral points from the real atmospheric polarization pattern automatically and precisely. The recognition method based on the intersection features of the neutral lines has better algorithm performance. Meanwhile, the recognition method based on the distribution characteristics of the degree of polarization has better algorithm robustness, and has much better algorithm performance than the other method at the midday. By the integrated application of two methods, the accuracy of the neutral points recognition can be further improved.