Generating an object map for use in a vehicle navigation system

A non-transitory computer-readable storage medium storing executable instructions that, when executed by a hardware processor, cause the hardware processor to perform steps comprising:

• capturing, by a set of cameras mounted on a rear side of a construction vehicle, a set of images of an area behind the construction vehicle;
• generating, by the construction vehicle, a disparity image of the area behind the construction vehicle by combining the set of images, each pixel of the disparity image representing a distance between the construction vehicle and a portion of the area behind the construction vehicle;
• generating, by the construction vehicle, a 3D point cloud comprising a 3D model of the area behind the construction vehicle;
• identifying, by the construction vehicle, a ground plane corresponding to a ground surface of the area behind the construction vehicle by performing a random sample consensus (RANSAC) operation on the 3D point cloud;
• generating, by the construction vehicle, a virtual plane parallel to the identified ground plane at a height above the ground plane selected based on a combination of:
• 1) a clearance height under the construction vehicle,
• 2) a pitch angle determined based on a height of a rear portion of the construction vehicle and a location of the rear portion of the construction vehicle relative to a wheel of the construction vehicle, and
• 3) an orientation, location, and height of the identified ground plane;
• generating, by the construction vehicle, an object map by masking pixels of the disparity image corresponding to objects in the area behind the construction vehicle that are below a threshold size and less than a height of the virtual plane;
• and
• reverse navigating, by the construction vehicle, within the area behind the construction vehicle using the generated object map.

A navigation system for a construction vehicle comprising:

• a hardware processor;
• and
• a non-transitory computer-readable storage medium storing executable instructions that, when executed, cause the hardware processor to perform steps comprising:
• capturing, by a set of cameras mounted on a rear side of a construction vehicle, a set of images of an area behind the construction vehicle;
• generating, by the construction vehicle, a disparity image of the area behind the construction vehicle by combining the set of images, each pixel of the disparity image representing a distance between the construction vehicle and a portion of the area behind the construction vehicle;
• generating, by the construction vehicle, a 3D point cloud comprising a 3D model of the area behind the construction vehicle;
• identifying, by the construction vehicle, a ground plane corresponding to a ground surface of the area behind the construction vehicle by performing a random sample consensus (RANSAC) operation on the 3D point cloud;
• generating, by the construction vehicle, a virtual plane parallel to the identified ground plane at a height above the ground plane selected based on a combination of:
• 1) a clearance height under the construction vehicle,
• 2) a pitch angle determined based on a height of a rear portion of the construction vehicle and a location of the rear portion of the construction vehicle relative to a wheel of the construction vehicle, and
• 3) an orientation, location, and height of the identified ground plane;
• generating, by the construction vehicle, an object map by masking pixels of the disparity image corresponding to objects in the area behind the construction vehicle that are below a threshold size and less than a height of the virtual plane;
• and
• navigating, by the construction vehicle, within the area behind the construction vehicle using the generated object map.