As engineering scenes increasingly require higher precision, real-time performance and robustness from perception systems, traditional approaches are becoming inadequate for handling complex and dynamic environments. To tackle this issue, we present an enhanced multi-sensor collaborative mapping algorithm based on improvements to the FAST-LIVO framework. The initial step in this process is the optimization of the feature point extraction method. Furthermore, the Lucas-Kanade (LK) feature point optical flow method is introduced to replace the direct method, and the RANSAC algorithm is applied to optimize feature matching and eliminate mismatched points. Moreover, the error-state iterated Kalman filter (ESIKF) is constructed on the basis of reprojection error in order to enhance fusion accuracy. A novel visual map management scheme has been developed to enhance system operational efficiency. In order to validate the practical effectiveness of the algorithm, a hardware platform was established and a series of tests were conducted. The experimental results demonstrate that the mean absolute trajectory error of the proposed algorithm in a typical industrial park environment was reduced by 39%, with the map construction exhibiting greater consistency and continuity.