This paper proposes a visual simultaneous localization and mapping (SLAM) loop closure detection algorithm based on an improved line band descriptor (LBD) and data-dependent measures with point and line features to improve the accuracy and recall rate associated with visual SLAM loop detection. First, each band descriptor was subjected to an internal comparison operation for solving the problem of low matching accuracy that can be attributed to the existing binary conversion operation of LBD, wherein only the sizes of the individual band descriptors were compared and their internal attributes were ignored. Then, we used a method based on data-dependent to estimate the image similarity considering the influence of the visual word frequency distribution information on similarity. Finally, the results obtained via verification using public data sets show that the proposed algorithm can achieve a high recall rate with 100% accuracy.