RatSLAM is a navigation algorithm that simulates the environmental perception mechanism of rodents, aiming to achieve localization and mapping. To address the issue of errors in RatSLAM caused by complex environments, we propose an optimized RatSLAM brain-like navigation model based on SURF feature matching. This method utilizes SURF feature matching algorithm to obtain the direction and position information of the vehicle in the environment, optimize the visual odometry based on the original RatSLAM. By combining head direction cells and place cells through a continuous attractor neural network, the current pose of the vehicle is jointly represented. Using the pose and time information obtained from these cells, the current position in the coordinate system is calculated through path integration, and a topological experiential map is constructed. Additionally, in local scenes, we perform loop closure detection and correction of the current trajectory by detecting pre-set place cell nodes using the SURF feature matching algorithm. Experimental results demonstrate that our proposed method exhibits better localization capabilities across different datasets.