Adaptive three-degree-of-freedom (pan-tilt-linear displacement) stereo method and device for depth estimation

Abstract

The present invention relates to a system comprising a three-degree-of-freedom robotic stereo camera designed to capture synchronized images from two lenses of different focal lengths, with angular movement (pan, tilt) and variable linear displacement between sensors. The system incorporates a motorized mechanism that allows dynamic modification of the stereo baseline, as well as a control unit configured to optimize this separation based on a depth error estimate associated with a region of interest detected in the scene. In particular, the invention integrates a depth accuracy estimation model that allows the calculation, before or during stereo capture, of an optimal baseline value based on geometric parameters such as estimated distance to the target, optical characteristics of the system, and disparity resolution.Based on this estimate, the control unit physically adjusts the sensor spacing to optimize the accuracy of the three-dimensional measurement for each specific scenario. The three degrees of freedom are integrated into an architecture that allows: a) Orienting the camera in any direction using pan and tilt movements; b) Dynamically adjusting the stereo baseline by modifying the physical spacing between the sensors; and c) Adapting the stereo geometry in real time through an error estimation-based optimization process, allowing for improved depth accuracy across different distance ranges.In preferred embodiments, the system can perform closed-loop iterative adaptive control to refine the baseline to a predetermined accuracy threshold, as well as restrict disparity calculations to geometric overlap regions associated with the region of interest, reducing computational load and increasing the robustness of the analysis. The invention is designed to integrate into the BioRed ecosystem, which comprises a distributed network of smart devices, digital platforms, and scientific analysis spaces that operate in an integrated manner to monitor, process, and share environmental and biological information in real time. Its modular, scalable, and open architecture allows for the participation of communities, institutions, and individual users.