Object detection utilizing geometric information fused with image data

A system for automatically identifying portions of an image of a physical scene as depicting physical objects of predefined types in the physical scene, the system comprising:

• a digital camera capturing two-dimensional data from the physical scene, the two-dimensional data being in the form of the image of the physical scene;
• a lidar device capturing three-dimensional data from the physical scene;
• and
• a computing device comprising:
• one or more processing units;
• and
• computer-readable media comprising computer-executable instructions, which, when executed by the one or more processing units, cause the computing device to:
• generate correspondences between subsets of the two-dimensional data and subsets of the three-dimensional data;
• identify, based on a first subset of the two-dimensional data, a first portion of the image of the physical scene as depicting a physical object of a first predefined type;
• select a first subset of the three-dimensional data that corresponds to the first subset of the two-dimensional data based on the generated correspondences;
• and
• determine, based on the first subset of the three-dimensional data, that the identified first portion of the image does not actually depict a physical object of the first predefined type.

A method of automatically identifying portions of an image of a physical scene as depicting physical objects of predefined types in the physical scene, the method comprising the steps of:

• generating a set of image features from two-dimensional image data that was obtained by two-dimensionally imaging the physical scene;
• generating a set of geometric features from three-dimensional image data that was obtained by three-dimensionally measuring the physical scene coincidentally with the two-dimensional imaging of the physical scene;
• generating correspondences between image features in the generated set of image features and geometric features in the generated set of geometric features;
• identifying, based on a first subset of the image features, a first portion of the image of the physical scene as depicting a physical object of a first predefined type;
• selecting a first subset of the geometric features that corresponds to the first subset of the image features based on the generated correspondences;
• and
• determining, based on the first subset of the geometric features, that the identified first portion of the image does not actually depict a physical object of the first predefined type.