Laser scanning technologies have significantly improved the efficiency of spatial data collection to deliver the needed as-built information of job sites. However, many imageries contain unneeded dense data that waste time for data collection and processing, or missing data that are in need. Targeted laser scan data collection is pivotal to avoid such problems. For example, engineers could avoid densely sampling simple geometries (e.g. flat walls) for saving time to focus on edges and openings. Scan planning algorithms can produce data collection plans based on targeted objects marked by users. Unfortunately, manually defining objects and their needed level of detail (LOD) is impractical in dynamic environments, such as construction sites. This research proposes an approach that identifies visually complex regions through discontinuity analysis in rapidly captured sparse imageries for guiding the imaging planning. First, we fuse 3D point cloud data with sparse laser-scan data. A visual complexity analysis algorithm then detects locations in sparse imageries that contain discontinuous 2D and 3D patterns (e.g., color change) for identifying parts deserving detailed laser scan. A frequency analysis for each location would then estimate the LOD necessary for assessing each targeted region. Finally, a sensor-planning algorithm generates laser scan plans based on visually complex regions and LOD requirements.