Abstract: The clustering problem of dense point targets is an important research topic in the planning of imaging satellite Earth observation tasks. The satellite performs a startup observation and scans a rectangular strip on the ground, which can simultaneously observe multiple ground targets. Due to the minimum single start-up time and energy constraints of satellites, the length of the bands formed by the satellite camera's push scan is not fixed, which makes the task clustering problem more complex. A variable length strip task clustering method is proposed for the clustering problem of dense point targets. Firstly, analyze the distribution of task points and establish successor node rules. Based on this rule, define a directed acyclic graph and use a depth first search method based on finite state machines to traverse the edges of the graph. Cluster all target points into as long overlapping bands as possible, and then select non overlapping bands that meet the constraints. Finally, the feasibility and effectiveness of the proposed method were verified through several simulation cases.