Abstract: The visual perception module in the original RatSLAM system is largely influenced by environmental features and illuminations, leading to reduced mapping accuracy and stability. To combat this issue, we propose an improved RatSLAM system based on a fast incremental visual information processing method. First, the binary search tree is used to store and retrieve the images. Then, the dynamic island mechanism is employed to group the images. Finally, the sequence match mechanism is employed to identify the previously visited places. The proposed method performs visual place recognition in real-time with high accuracy. The experimental results indicate that the proposed algorithm has an accuracy of over 99% and a recall rate of over 80%. The average processing time is less than 50 ms. Thus, the SLAM system incorporating the proposed visual processing method is superior to the existing RatSLAM-related systems in terms of loop-closure detection accuracy, processing time, and robustness of mapping, further confirming the robustness and efficiency of the proposed fast incremental visual information processing method.