Landslides are natural hazards that can cause severe damage and loss of life. Optical cameras are a low-cost and high-resolution alternative among many monitoring systems, as their size and capabilities can vary, allowing for flexible implementation and location. Computer vision is a branch of artificial intelligence that can analyze and understand optical images, using techniques such as optical flow, image correlation and machine learning. The application of such techniques can estimate the motion vectors, displacement fields, providing valuable information for landslide detection, monitoring and prediction. However, computer vision also faces some challenges such as illumination changes, occlusions, image quality, and computational complexity. In this work, a computer vision approach based on Lucas-Kanade optical dense flow was applied to estimate the motion vectors between consecutive images obtained during landslide simulations in a laboratory environment. The approach is applied to two experiments that vary in their illumination and setup parameters to test its applicability. We also discuss the application of this methodology to images from Sentinel-2 satellite optical sensors for landslide monitoring in real-world scenarios.