The manual generation of instructions for industrial assembly processes is associated with significant effort, especially concerning processes with a high variance of low quantity products. Companies are therefore often hesitant to create detailed assembly instructions and rely on the implicit knowledge of employees instead. Due to demographic change, capturing this implicit knowledge is becoming increasingly relevant to prevent a loss of know-how. Contributing to the effort, this paper presents a methodology that, building on previous works of the authors, addresses optical movement detection for automatic assembly instruction generation (AIG). For this purpose, dense optical flow and conventional background subtraction were applied in our AIG framework and evaluated in three assembly use cases, showcasing optical flow as most effective. By enabling the extraction of implicit knowledge at low training time, our findings provide a foundation for future advances in automatic AIG, alongside practical solutions for industries relying on manual assembly processes.