Ontological models of the StereoBM and StereoSGBM algorithms are used to construct depth maps from stereo images obtained with two cameras. They are implemented on a controlling personal computer and a stereo camera, consisting of software-implemented algorithms for constructing depth maps from stereo images: local (StereoBM) and semi-global matching (StereoSGBM). The images from the camera are converted to grayscale to accelerate computations, after which similarity metrics (SAD, SSD, and NCC) are applied to determine disparity (the pixel shift on the right image relative to the left). A limitation of the model is the sensitivity of the StereoBM algorithm to noise and lack of texture, which is compensated by applying StereoSGBM with global optimization. Experimental results based on MSE showed that the StereoSGBM algorithm provides more accurate depth map construction due to global optimization, but requires significantly more computational resources (execution time increases by 4.5 times compared to StereoBM). Despite its high speed, the StereoBM algorithm is less preferable due to high noise levels in the depth map.