Visual perception method, apparatus, device, and medium based on an autonomous vehicle

1. A visual perception method based on an autonomous vehicle, comprising:

obtaining a first visual perception image collected by the autonomous vehicle;

inputting the first visual perception image into a first neural network model, and recognizing multi-channel feature information of at least one target recognition object to be recognized, to eliminate redundant feature information in the first visual perception image; and

inputting the multi-channel feature information of each of the at least one target recognition object to be recognized into a corresponding sub-neural network model of multiple sub-neural network models in a second neural network model, to obtain the at least one target recognition object; wherein each of the multiple sub-neural network models corresponds to a class of target recognition object.

2. The method according to claim 1 , wherein the method further comprises:

obtaining the first neural network model.

3. The method according to claim 2 , wherein the obtaining the first neural network model comprises:

training a first preset neural network model according to a second visual perception image and multi-channel feature information of at least one training sample object, to generate the first neural network model; wherein the second visual perception image comprises the at least one training sample object.

4. The method according to claim 1 , further comprises:

obtaining the second neural network model.

5. The method according to claim 4 , wherein the obtaining the second neural network model comprises:

training multiple sub-preset neural network models in a second preset neural network model according to at least one training sample object and multi-channel feature information of the at least one training sample object, to generate the second neural network model; wherein each of the multiple sub-preset neural network models corresponds to a class of training sample object.

6. The method according to claim 1 , wherein the obtaining the first visual perception image collected by the autonomous vehicle comprises:

obtaining the first visual perception image collected by an image collection apparatus in the autonomous vehicle.

7. The method according to claim 1 , wherein the target recognition object comprises at least one of the followings: a target substance and target semantics.

8. A visual perception apparatus based on an autonomous vehicle, comprising:

a processor; and

a computer-readable medium for storing program codes, which, when executed by the processor, the processor is caused to:

obtain a first visual perception image collected by the autonomous vehicle;

input the first visual perception image into a first neural network model, and recognize multi-channel feature information of at least one target recognition object to be recognized, to eliminate redundant feature information in the first visual perception image; and

input the multi-channel feature information of each of the at least one target recognition object to be recognized into a corresponding sub-neural network model of multiple sub-neural network models in a second neural network model, to obtain at least one target recognition object; wherein each of the multiple sub-neural network models corresponds to a class of target recognition object.

9. The apparatus according to claim 8 , wherein the processor is further caused to:

obtain the first neural network model.

10. The apparatus according to claim 9 , wherein the processor is further caused to:

train a first preset neural network model according to a second visual perception image and multi-channel feature information of at least one training sample object, to generate the first neural network model; wherein the second visual perception image comprises the at least one training sample object.

11. The apparatus according to claim 8 , wherein the processor is further caused to:

obtain the second neural network model.

12. The apparatus according to claim 11 , wherein the processor is further caused to:

train multiple sub-preset neural network models in a second preset neural network model according to at least one training sample object and multi-channel feature information of the at least one training sample object, to generate the second neural network model; wherein each of the multiple sub-preset neural network models corresponds to a class of training sample object.

13. The apparatus according to claim 8 , wherein the processor is further caused to:

obtain the first visual perception image collected by an image collection apparatus in the autonomous vehicle.

14. The apparatus according to claim 8 , wherein the target recognition object comprises at least one of the followings: a target substance and target semantics.

15. A non-transitory storage medium, comprising a non-transitory readable storage medium and a computer instruction, the computer instruction is stored in the readable storage medium; the computer instruction is configured to implement a visual perception method comprising the steps of:

obtaining a first visual perception image collected by the autonomous vehicle;

inputting the first visual perception image into a first neural network model, and recognizing multi-channel feature information of at least one target recognition object to be recognized, to eliminate redundant feature information in the first visual perception image; and

inputting the multi-channel feature information of each of the at least one target recognition object to be recognized into a corresponding sub-neural network model of multiple sub-neural network models in a second neural network model, to obtain the at least one target recognition object; wherein each of the multiple sub-neural network models corresponds to a class of target recognition object.