Visual detection of hands on steering wheel

A processing apparatus comprising:

• a processor configured to perform a method including:
• determining, based on an image capturing an operating element of a vehicle and a driver of the vehicle, a distance between the operating element and a hand of the driver;
• and
• in response to the distance meeting a criterion that is indicative of a possibility that the driver operates the operating element:
• determining, based on a part of the image capturing the operating element, classification information of the operating element;
• determining, based on a part of the image capturing the driver, classification information of a pose of the driver, wherein determining classification information of the pose of the driver includes:
• classifying, using a classification network for detecting a body pose of the driver, the driver captured in the image to detect the body pose of the driver;
• and
• determining, based on the classifying, the classification information indicating whether the body pose of the driver is to allow the driver to operate the operating element;
• and
• determining, based on the distance, the classification information of the operating element and the classification information of the pose of the driver, whether the driver operates the operating element.

A non-transitory computer-readable medium comprising instructions including:

• determining, based on an image capturing an operating element of a vehicle and a driver of the vehicle, a distance between the operating element and a hand of the driver;
• and
• in response to the distance meeting a criterion that is indicative of a possibility that the driver operates the operating element:
• determining, based on a part of the image capturing the operating element, classification information of the operating element;
• determining, based on a part of the image capturing the driver, classification information of a pose of the driver, wherein determining classification information of the pose of the driver includes:
• classifying, using a classification network for detecting a body pose of the driver, the driver captured in the image to detect the body pose of the driver;
• and
• determining, based on the classifying, the classification information indicating whether the body pose of the driver is to allow the driver to operate the operating element;
• and
• determining, based on the distance, the classification information of the operating element and the classification information of the pose of the driver, whether the driver operates the operating element.