CN-115984795-B - Image sensing method, computer device, computer-readable storage medium, and vehicle

Abstract

The invention relates to the technical field of automatic driving, in particular to an image sensing method, computer equipment, a computer readable storage medium and a vehicle, and aims to solve the problem of improving image sensing accuracy. The training method of the image sensing model comprises the steps of obtaining a teacher model capable of estimating the image depth of the image frame, and enabling the teacher model to guide the image sensing model to conduct pre-training of image depth estimation by using the image frame, and conducting final training of image target sensing on the image sensing model which is completed to conduct pre-training according to the image frame and image target marking information of the image frame. By the image sensing method, not only can the accuracy of image sensing be improved, but also the robustness of image sensing can be improved.

Inventors

• KANG ZIJIAN

Assignees

• 安徽蔚来智驾科技有限公司

Dates

Publication Date

20260508

Application Date

20221230

Claims (9)

1. 1. A method of image perception, the method comprising: Acquiring an image frame acquired by a vehicle; Image target perception is carried out on the image frame by adopting an image perception model; the image perception model is obtained through training in the following mode: Acquiring a teacher model capable of estimating image depth of an image frame; The image sensing method comprises the steps of respectively constructing a feature extraction network, a target sensing network and a depth estimation network to form an image sensing model, wherein the feature extraction network is used for extracting image features of an image frame, the target sensing network is used for sensing an image target according to the image features, and the depth estimation network is used for estimating the depth of each pixel point position on the image frame according to the image features; A knowledge distillation method is adopted, so that a teacher model guides the image perception model to use an image frame to conduct pre-training of image depth estimation, and the depth estimation network is removed after the pre-training is completed; And performing final training of image target perception on the image perception model subjected to pre-training according to the image frame and the image target annotation information thereof.
2. 2. The image sensing method according to claim 1, wherein the step of using a knowledge distillation method to enable a teacher model to guide the image sensing model to perform pre-training of image depth estimation using image frames specifically comprises: Estimating the depth and the confidence level of each pixel point position on the image frame by adopting a teacher model; acquiring pixel point positions of which the confidence coefficient of the depth is larger than a preset confidence coefficient threshold value; acquiring a high confidence region on the image frame according to the pixel point position; And (3) adopting a knowledge distillation method to enable the teacher model to guide the image perception model to conduct pre-training of image depth estimation by using a high confidence region on the image frame.
3. 3. The image sensing method according to claim 2, wherein the step of acquiring the high confidence region on the image frame according to the pixel position comprises: generating an image mask of a high confidence region according to the pixel point positions; and acquiring a high confidence region on the image frame according to the image mask.
4. 4. The image sensing method according to claim 2, wherein the step of estimating the depth and the confidence level of each pixel point position on the image frame by using the teacher model comprises: If the teacher model is obtained by performing image depth estimation training according to the image frame and the non-dense depth annotation information thereof, estimating the depth of each pixel point position on the image frame by adopting the teacher model, and estimating the probability of each pixel point on the image frame being scanned by the vehicle radar; And respectively determining the confidence of the depth at each pixel point position according to the probability.
5. 5. The image sensing method according to claim 2, wherein the step of estimating the depth and the confidence level of each pixel point position on the image frame by using the teacher model comprises: If the teacher model is obtained by performing image depth estimation training according to the image frame and dense depth annotation information thereof, estimating the depth of each pixel point position on the image frame by adopting the teacher model; estimating uncertainty of depth at each pixel point position based on an uncertainty estimation method; And respectively determining the confidence of the depth at each pixel point position according to the uncertainty.
6. 6. The image perception method according to claim 1, further comprising constructing the feature extraction network and the target perception network by: Constructing a feature pyramid network to form the feature extraction network; And constructing a plurality of target perception networks, wherein each target perception network is respectively used for perceiving different types of image targets according to the image features extracted by the feature pyramid network.
7. 7. A computer device comprising a processor and a storage means, the storage means being adapted to store a plurality of program code, characterized in that the program code is adapted to be loaded and executed by the processor to perform the image perception method of any one of claims 1 to 6.
8. 8. A computer readable storage medium, in which a plurality of program codes are stored, characterized in that the program codes are adapted to be loaded and executed by a processor to perform the image perception method of any one of claims 1 to 6.
9. 9. A vehicle, characterized in that it comprises the computer device of claim 7.

Description

Image sensing method, computer device, computer-readable storage medium, and vehicle Technical Field The invention relates to the technical field of automatic driving, in particular to an image sensing method, computer equipment, a computer readable storage medium and a vehicle. Background When the automatic driving control is performed on the vehicle, an image around the vehicle is generally collected through a visual sensor, and then the image is perceived through a perception model so as to identify information of target objects such as lane lines, traffic signs, pedestrians, obstacles and the like around the vehicle. However, the conventional image sensing method at present cannot generally obtain more accurate depth information when a target object is identified, so that the accuracy of image sensing is reduced. Accordingly, there is a need in the art for a new solution to the above-mentioned problems. Disclosure of Invention The present invention has been made to overcome the above-mentioned drawbacks, and provides an image sensing method, a computer device, a computer-readable storage medium, and a vehicle that solve or at least partially solve the technical problem of how to improve the accuracy of image sensing. In a first aspect, there is provided an image perception method, the method comprising: Acquiring an image frame acquired by a vehicle; Image target perception is carried out on the image frame by adopting an image perception model; the image perception model is obtained through training in the following mode: Acquiring a teacher model capable of estimating image depth of an image frame; a knowledge distillation method is adopted, so that a teacher model guides the image perception model to use an image frame to conduct pre-training of image depth estimation; And performing final training of image target perception on the image perception model subjected to pre-training according to the image frame and the image target annotation information thereof. In one technical scheme of the image sensing method, the step of using the knowledge distillation method to enable the teacher model to guide the image sensing model to conduct pre-training of image depth estimation by using image frames specifically comprises the following steps: Estimating the depth and the confidence level of each pixel point position on the image frame by adopting a teacher model; acquiring pixel point positions of which the confidence coefficient of the depth is larger than a preset confidence coefficient threshold value; acquiring a high confidence region on the image frame according to the pixel point position; And (3) adopting a knowledge distillation method to enable the teacher model to guide the image perception model to conduct pre-training of image depth estimation by using a high confidence region on the image frame. In one technical scheme of the image sensing method, the step of acquiring the high confidence region on the image frame according to the pixel point position specifically includes: generating an image mask of a high confidence region according to the pixel point positions; and acquiring a high confidence region on the image frame according to the image mask. In one technical scheme of the image sensing method, the step of estimating the depth and the confidence of the depth at each pixel point position on the image frame by using the teacher model specifically includes: If the teacher model is obtained by performing image depth estimation training according to the image frame and the non-dense depth annotation information thereof, estimating the depth of each pixel point position on the image frame by adopting the teacher model, and estimating the probability of each pixel point on the image frame being scanned by the vehicle radar; And respectively determining the confidence of the depth at each pixel point position according to the probability. In one technical scheme of the image sensing method, the step of estimating the depth and the confidence of the depth at each pixel point position on the image frame by using the teacher model specifically includes: If the teacher model is obtained by performing image depth estimation training according to the image frame and dense depth annotation information thereof, estimating the depth of each pixel point position on the image frame by adopting the teacher model; estimating uncertainty of depth at each pixel point position based on an uncertainty estimation method; And respectively determining the confidence of the depth at each pixel point position according to the uncertainty. In one technical solution of the above image sensing method, before the step of "training the teacher model to guide the image sensing model to perform image depth estimation using image frames by using the knowledge distillation method", the method further includes constructing an image sensing model by: Respectively constructing a feature extraction network, a target perception