Depth Estimation from Video using Computer Vision and Machine Learning with Hyperparameter Optimization

This research work focuses on a widespread interpretation for the operation of depth estimation using machine learning (ML) algorithms and computer vision. Depth estimation, also known as abstraction, refers to the processes and systems used to describe the geometric features of a situation. Computer vision is a branch of artificial intelligence that teaches advanced analytics and comprehends images. Based on the principle of disparity an algorithm was presented for the application of computer vision through which moving objects and human was detected using calibrated stereo camera. A correlation technique was used to identify the disparity as well as to detect, extract and match features of any given inputs. The experiment was performed on MATLAB environment by dint of a function of computer vision. Using this algorithm, detection and identification of a moving object was done and the depth of it was measured. The disparity map and reconstruction of the 3D image was presented as well as the distance of each person from the camera was calculated. Afterwards, a 3D mapping of that moving object was created using reconstruction. Moreover, monocular 3-D reconstruction was performed to provide nearly accurate measurement.