Method, electronic device, and computer program product for road monitoring

Patent metadata

Field	Value
Publication number	US-12450912-B2
Application number	US-202318144912-A
Country	US
Kind code	B2
Filing date	May 9, 2023
Priority date	Apr 14, 2023
Publication date	Oct 21, 2025
Grant date	Oct 21, 2025

Abstract

Embodiments of the present disclosure provide a method, an electronic device, and a computer program product for monitoring. The method includes receiving, at a second node, a target image from a first node, where the target image is determined as being of a first category based on performing target detection on a road monitoring image obtained by the first node, and the second node is closer to a cloud end than the first node. The method further includes determining, at the second node, a second category of the target image, where the second category is a subcategory of the first category; and in response to the second category being a preset category, sending a warning corresponding to the second category to a terminal device associated with the road monitoring image. According to the method of the embodiments of the present disclosure, the accuracy of road monitoring warnings can be improved.

First claim

What is claimed is:

1. A method for road monitoring, comprising:
receiving, from a first node, by a second node, a target image from the first node, wherein the target image is determined as being of a first category based on performing target detection on a road monitoring image obtained by the first node, and the second node is closer to a cloud end than the first node;
determining, at the second node, a second category of the target image, wherein the second category is a subcategory of the first category; and
in response to the second category being a preset category, sending a warning corresponding to the second category to a terminal device associated with the road monitoring image;
wherein determining, at the second node, the second category of the target image comprises:
extracting a target feature of the target image;
determining similarities between the target feature and preset features in a first feature library; and
determining the second category of the target image according to the similarities and categories associated with the preset features;
wherein determining the second category of the target image comprises:
choosing a first group of preset features meeting a preset condition according to values of the similarities; and
determining the second category based on categories associated with respective preset features in the first group of preset features.

2. The method according to claim 1 , wherein the target image is a cropped part of an image area containing an object of the first category in the road monitoring image based on the first node.

3. The method according to claim 1 , wherein determining the second category of the target image further comprises:
determining, according to a category associated with each preset feature in the first group of preset features, a proportion of each category in all categories associated with the first group of preset features; and
determining a category with the greatest proportion as the second category of the target image.

4. The method according to claim 1 , further including:
in response to determining that each of the similarities is less than a preset similarity threshold, sending the target image to the cloud end;
receiving, at the second node, the second category of the target image from the cloud end, wherein the second category is determined by the cloud end according to a target feature in the target image, preset features in a second feature library, and categories associated with the preset features in the second feature library, and the first feature library is a subset of the second feature library.

5. The method according to claim 4 , wherein the target feature of the target image is obtained by a fine-tuned classification model at the cloud end, and the fine-tuned classification model performs fine-tuning on a trained classification model at the cloud end based on a newly added monitoring image and a newly added category corresponding to the newly added monitoring image.

6. The method according to claim 5 , wherein the second feature library is obtained at the cloud end by acquiring a newly added preset feature corresponding to the newly added category using the fine-tuned classification model based on the newly added monitoring image and adding the newly added preset feature and the newly added category to the second feature library in the cloud end.

7. The method according to claim 5 , further comprising:
receiving, at the second node, the fine-tuned classification model from the cloud end to update a trained classification model at the second node.

8. The method according to claim 1 , wherein extracting the target feature of the target image comprises:
fine-tuning a trained classification model at the second node according to source monitoring images of the preset features in the first feature library of the second node and categories associated with the preset features to obtain a fine-tuned classification model, wherein the source monitoring images are images associated with a geographical area corresponding to the second node; and
acquiring a target feature in the target image using the fine-tuned classification model based on the target image.

9. The method according to claim 1 , wherein the first node is a roadside edge node.

10. An electronic device, comprising:
at least one processor; and
memory coupled to the at least one processor and having instructions stored therein, wherein the instructions, when executed by the at least one processor, cause the electronic device to perform operations comprising:
receiving, from a first node, by a second node, a target image from the first node, wherein the target image is determined as being of a first category based on performing target detection on a road monitoring image obtained by the first node, the second node is closer to a cloud end than the first node, and the first node is a roadside edge node;
determining, at the second node, a second category of the target image, wherein the second category is a subcategory of the first category; and
in response to the second category being a preset category, sending a warning corresponding to the second category to a terminal device associated with the road monitoring image;
wherein determining, at the second node, the second category of the target image comprises:
extracting a target feature of the target image;
determining similarities between the target feature and preset features in a first feature library; and
determining the second category of the target image according to the similarities and categories associated with the preset features;
wherein determining the second category of the target image comprises:
choosing a first group of preset features meeting a preset condition according to values of the similarities; and
determining the second category based on categories associated with respective preset features in the first group of preset features.

11. The electronic device according to claim 10 , wherein the target image is a cropped part of an image area containing an object of the first category in the road monitoring image based on the first node.

12. The electronic device according to claim 10 , wherein determining the second category of the target image further comprises:
determining, according to a category associated with each preset feature in the first group of preset features, a proportion of each category in all categories associated with the first group of preset features; and
determining a category with the greatest proportion as the second category of the target image.

13. The electronic device according to claim 10 , wherein the operations further comprise:
in response to determining that each of the similarities is less than a preset similarity threshold, sending the target image to the cloud end;
receiving, at the second node, the second category of the target image from the cloud end, wherein the second category is determined by the cloud end according to a target feature in the target image, preset features in a second feature library, and categories associated with the preset features in the second feature library, and the first feature library is a subset of the second feature library.

14. The electronic device according to claim 13 , wherein the target feature of the target image is obtained by a fine-tuned classification model at the cloud end, and the fine-tuned classification model performs fine-tuning on a trained classification model at the cloud end based on a newly added monitoring image and a newly added category corresponding to the newly added monitoring image.

Assignees

• Dell Products Lp

Inventors

• He Bin
• Wu Wenlei
• Ni Jiacheng
• Jia Zhen

Classifications

• G06V10/761

  Proximity, similarity or dissimilarity measures · CPC title

• G08G1/0104

  Measuring and analyzing of parameters relative to traffic conditions · CPC title

• G06V10/7715

  Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods · CPC title

• G06V2201/07

  Target detection · CPC title

• G06V10/764

  using classification, e.g. of video objects · CPC title