Vision-based sensors have been widely used in reality capture and the corresponding scene understanding tasks such as object detection. Given the increasing complexity of built environments, geometric features from the raw scanning data can become too vague for effective object detection. One example challenge is stacked object recognition, i.e., the segmentation, detection, and recognition of objects being stacked together with similar geometric features or occlusions. Previous methods propose to use high-resolution sensors to capture more detailed geometry information to highlight the boundaries between adjacent objects, which increase the deployment cost and computing needs. This paper proposes a novel data augmentation and voting method for stacked object detection with only low-cost sparse sensors. Several locomotion strategies were used to focus on filling the gaps of the sparse light detection and ranging (LiDAR) sensor. A modified LiDAR odometry and mapping (LOAM) method was used to register and augment raw point cloud data from multiple scans in real time. Then a voxel-based density voting method was applied to centralize the points in enhanced scan for a more accurate clustering. Finally, the clustered points were grouped and applied to generate three-dimensional (3D) bounding boxes for object boundary identification. A pilot test was performed to show the improved results of the proposed methods. A series of benchmarking studies were also performed to identify the minimum acceptable density level for the proposed method.