In response to the issues of ghost regions in traditional ViBe (visual background extractor) algorithm during motion target detection, as well as the potential for false positives or misses in the presence of dynamic interference in complex video frames, an improved ViBe algorithm is proposed. Firstly, improve the initial background model using the five-frame differencing method, and then employ the weighted local binary patterns (WLBP) algorithm to extract texture features from the frame difference images for the suppression of ghost regions. Secondly, during the foreground detection process, adaptively calculate the radius threshold using a dynamic adjustment of the update rate based on the scene’s complexity, enhancing resistance to interference. Lastly, perform morphological operations and connectivity analysis on the extracted foreground objects, fill in the detected foreground object gaps to address deficiencies in motion object detection. The experimental results demonstrate that the improved algorithm, as compared to the traditional ViBe algorithm, enhances detection accuracy across various test scenarios such as subway stations, boats, highways, and railways. Specifically, the accuracy improved by 11.6, 20, 11.3, and 10.8% in these respective scenarios. The recall rate also showed significant improvements, with increases of 18.8, 22.3, 11.8, and 12.3%. The comprehensive evaluation metric, F-M value, increased by 0.134, 0.196, 0.111, and 0.117 in these scenarios. The proposed method effectively reduces the occurrence of ghost regions, achieving lower false positive and false negative rates, and significantly improves accuracy compared to the traditional ViBe algorithm.