Disparity map acquisition method and apparatus, device, control system and storage medium

A disparity map acquisition apparatus, comprising:

• at least one memory configured to store program code;
• and
• at least one processor configured to read the program code and operate as instructed by the program code, the program code including:
• feature extraction code configured to cause at least one of the at least one processor to respectively perform, for each layer of M cascaded feature extraction layers, feature extraction on left-view images and right-view images of a captured object, to obtain a left-view feature map set and a right-view feature map set of each feature extraction layer, M being a positive integer greater than or equal to 2;
• initial disparity map obtainment code configured to cause at least one of the at least one processor to obtain an initial disparity map based on the left-view feature map set and the right-view feature map set extracted by an Mth feature extraction layer;
• and
• disparity map refinement code configured to cause at least one of the at least one processor to, iteratively refine, in an order from the (M−1)th feature extraction layer to a first feature extraction layer in sequence, a disparity map through the left-view feature map set and the right-view feature map set extracted by each feature extraction layer until a final disparity map is obtained based on an iteratively refined disparity map of the first feature extraction layer, an iteratively refined disparity map of an ith feature extraction layer being obtained by refining, by using a residual of a matching cost for matching the left-view feature map set and the right-view feature map set extracted by the ith feature extraction layer, a matching cost after iterative refinement of an (i+1)th feature extraction layer, the disparity map used by the (M−1)th feature extraction layer for iterative refining being the initial disparity map, and i being a positive integer greater than zero and less than M.

A non-transitory computer readable storage medium, having computer instructions stored thereon, the computer instructions, when run by at least one processor, causing at least one of the at least one processor to perform:

• respectively performing, for each layer of M cascaded feature extraction layers, feature extraction on left-view images and right-view images of a captured object, to obtain a left-view feature map set and a right-view feature map set of each feature extraction layer, M being a positive integer greater than or equal to 2;
• obtaining an initial disparity map based on the left-view feature map set and the right-view feature map set extracted by an Mth feature extraction layer;
• and
• iteratively refining, in an order from an (M−1)th feature extraction layer to a first feature extraction layer in sequence, a disparity map through the left-view feature map set and the right-view feature map set extracted by each feature extraction layer until a final disparity map is obtained based on an iteratively refined disparity map of the first feature extraction layer, an iteratively refined disparity map of an ith feature extraction layer being obtained by refining, by using a residual of a matching cost for matching the left-view feature map set and the right-view feature map set extracted by the ith feature extraction layer, a matching cost after iterative refinement of an (i+1)th feature extraction layer, the disparity map used by the (M−1)th feature extraction layer for iterative refining being the initial disparity map, and i being a positive integer greater than zero and less than M.