Efficient birds-eye view (BEV) generation in vehicular systems

An apparatus, comprising:

• a memory storing processor-readable code;
• and
• at least one processor coupled to the memory, the at least one processor configured to execute the processor-readable code to cause the at least one processor to perform operations including:
• receiving two or more image frames from two or more image sensors arranged to capture an environment around a vehicle;
• determining corresponding frustum cells for locations within the environment, wherein each frustum cell corresponds to at least one pixel within at least one of the two or more image frames;
• and
• determining a top-down representation of the environment around the vehicle by mapping vertically aligned frustum cells to the same location within the top-down representation, wherein determining the top-down representation comprises, for each location within the top-down representation;
• determining a combined depth matrix by combining individual depth vectors for vertically aligned pixels within the two or more image frames;
• determining a combined context matrix by combining individual context vectors for vertically aligned pixels within the two or more image frames;
• and
• determining at least a portion of the top-down representation by multiplying the combined depth matrix and the combined context matrix.

A vehicle, comprising:

• two or more image capture devices;
• a memory storing processor-readable code;
• and
• at least one processor coupled to the memory and the two or more image capture devices, the at least one processor configured to execute the processor-readable code to cause the at least one processor to perform operations including:
• receiving two or more image frames from the two or more image capture devices;
• determining corresponding frustum cells for locations within an environment surrounding the vehicle, wherein each frustum cell corresponds to at least one pixel within at least one of the two or more image frames;
• and
• determining a top-down representation of the environment around the vehicle by mapping vertically aligned frustum cells to the same location within the top-down representation, wherein determining the top-down representation comprises, for each location within the top-down representation:
• determining a combined depth matrix by combining individual depth vectors for vertically aligned pixels within the two or more image frames;
• determining a combined context matrix by combining individual context vectors for vertically aligned pixels within the two or more image frames;
• and
• determining at least a portion of the top-down representation by multiplying the combined depth matrix and the combined context matrix.

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

• receiving two or more image frames from two or more image sensors arranged to capture an environment around a vehicle;
• determining corresponding frustum cells for locations within the environment, wherein each frustum cell corresponds to at least one pixel within at least one of the two or more image frames;
• and
• determining a top-down representation of the environment around the vehicle by mapping vertically aligned frustum cells to the same location within the top-down representation, wherein determining the top-down representation comprises, for each location within the top-down representation:
• determining a combined depth matrix by combining individual depth vectors for vertically aligned pixels within the two or more image frames;
• determining a combined context matrix by combining individual context vectors for vertically aligned pixels within the two or more image frames;
• and
• determining at least a portion of the top-down representation by multiplying the combined depth matrix and the combined context matrix.