# US 12,183,059 B2 · Ai-based object classification method and apparatus, and medical imaging device and storage medium

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,183,059 B2
- Document ID: US12183059B2
- Status: Granted
- Kind: B2
- Application number: 17/686,950
- Filed: 2022-03-04
- Published: 2022-06-16
- Granted: 2024-12-31
- Earliest claimed: 2020-02-17
- Assignees: Tencent Technology (Shenzhen) Company Limited
- Applicants: Tencent Technology (Shenzhen) Company Limited
- Inventors: Liang Wang; Jiarui Sun; Rongbo Shen; Cheng Jiang; Yanchun Zhu; Jianhua Yao
- Web page: https://patentgrep.com/patent/US12183059B2

## CPC

- G06V 10/273: {removing elements interfering with the pattern to be recognised}
- G06V 10/764: using classification, e.g. of video objects
- G06T 3/4038: Image mosaicing, e.g. composing plane images from plane sub-images
- G06V 10/26: Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V 10/454: {Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]}
- G06V 10/82: using neural networks
- G06V 2201/032: of protuberances, polyps nodules, etc.
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL

## Abstract

An AI-based object classification method and apparatus, a computer-readable storage medium, and a computer device. The method includes: obtaining a target image to be processed, the target image including a target detection object; separating a target detection object image of the target detection object from the target image; inputting the target detection object image into a feature object prediction model to obtain a feature object segmentation image of a feature object in the target detection object image; obtaining quantitative feature information of the target detection object according to the target detection object image and the feature object segmentation image; and classifying the target detection object image according to the quantitative feature information to obtain category information of the target detection object in the target image.

## Claims

### Claim 1

An Al-based object classification method, performed by a computer device, the method comprising:
obtaining a target image to be processed, the target image comprising a target detection object;
separating a target detection object image of the target detection object from the target image;
inputting the target detection object image into a feature object prediction model to obtain a feature object segmentation image of a feature object in the target detection object image;
obtaining quantitative feature information of the target detection object according to the target detection object image and the feature object segmentation image; and
classifying the target detection object image according to the quantitative feature information to obtain category information of the target detection object in the target image,
wherein the obtaining quantitative feature information comprises:
determining a region where the feature object is located and an area value of the region where the feature object is located from the feature object segmentation image;
performing contour extraction on the region where the feature object is located;
calculating a perimeter of a contour of the region where the feature object is located, the perimeter of the contour being represented by a quantity of pixels of the contour of the region where the feature object is located, to calculate an area value of a circle with an equal perimeter; and
obtaining a ratio of the area value of the circle with the equal perimeter to the area value of the region where the feature object is located to obtain the quantitative feature information of the feature object in circularity.

### Claim 2 (depends on 1)

The Al-based object classification method according to claim 1, wherein the separating comprises:
inputting the target image into a target detection object segmentation model to obtain a first target detection object prediction image corresponding to the target image, and obtaining a region where the target detection object is located in the target image according to the first target detection object prediction image; and
segmenting the target image according to the region where the target detection object is located to obtain the target detection object image.

### Claim 3 (depends on 2)

The AI-based object classification method according to claim 2, wherein the inputting the target image into a target detection object segmentation model to obtain a first target detection object prediction image corresponding to the target image, and obtaining a region where the target detection object is located in the target image according to the first target detection object prediction image comprises:
scaling the target image to obtain a scaled image;
inputting the scaled image into the target detection object segmentation model to obtain a second target detection object prediction image of the scaled image;
obtaining region information of a region where a target detection object is located in the scaled image according to the second target detection object prediction image; and
obtaining the region where the target detection object is located in the target image according to the region information.

### Claim 4 (depends on 2)

The AI-based object classification method according to claim 2, wherein the inputting the target image into a target detection object segmentation model to obtain a first target detection object prediction image corresponding to the target image comprises:
cutting the target image into a plurality of target sub-images to be processed according to a cutting rule;
inputting the target sub-images into the target detection object segmentation model to obtain target detection object prediction sub-images corresponding to the target sub-images; and
stitching the target detection object prediction sub-images according to the cutting rule to obtain the first target detection object prediction image corresponding to the target image.

### Claim 5 (depends on 2)

The AI-based object classification method according to claim 2, wherein the target detection object segmentation model comprises an encoder and a decoder; and
the inputting the target image into a target detection object segmentation model to obtain a first target detection object prediction image corresponding to the target image comprises:
inputting the target image into the encoder of the target detection object segmentation model, and encoding the target image by the encoder to obtain image feature information of the target image; and
inputting the image feature information into the decoder of the target detection object segmentation model, and performing a decoding operation on the image feature information by the decoder to obtain the first target detection object prediction image corresponding to the target image.

### Claim 6 (depends on 1)

The AI-based object classification method according to claim 1, wherein the feature object prediction model comprises a heat map prediction network; and the inputting comprises:
inputting the target detection object image into the heat map prediction network to obtain a feature object heat map corresponding to the target detection object image;
obtaining a heat value of each pixel of the feature object in the target detection object image according to the feature object heat map; and
determining a region where the feature object is located from the feature object heat map according to the heat value to obtain the feature object segmentation image.

### Claim 7

An Al-based object classification apparatus, comprising:
at least one memory configured to store program code; and
at least one processor configured to read the program code and operate as instructed by the program code, the program code comprising:
image obtaining code configured to cause the at least one processor to obtain a target image to be processed, the target image comprising a target detection object;
image segmentation code configured to cause the at least one processor to separate a target detection object image of the target detection object from the target image;
feature image obtaining code configured to cause the at least one processor to input the target detection object image into a feature object prediction model to obtain a feature object segmentation image of a feature object in the target detection object image;
feature information obtaining code configured to cause the at least one processor to obtain quantitative feature information of the target detection object according to the target detection object image and the feature object segmentation image; and
object classification code configured to cause the at least one processor to classify the target detection object image according to the quantitative feature information to obtain category information of the target detection object in the target image,
wherein the feature information obtaining code is further configured to cause the at least one processor to:
determine a region where the feature object is located and an area value of the region where the feature object is located from the feature object segmentation image;
perform contour extraction on the region where the feature object is located;
calculate a perimeter of a contour of the region where the feature object is located, the perimeter of the contour being represented by a quantity of pixels of the contour of the region where the feature object is located, to calculate an area value of a circle with an equal perimeter; and
obtain a ratio of the area value of the circle with the equal perimeter to the area value of the region where the feature object is located to obtain the quantitative feature information of the feature object in circularity.

### Claim 8 (depends on 7)

The Al-based object classification apparatus according to claim 7, wherein the image segmentation code comprises:
object region determining code configured to cause the at least one processor to input the target image into a target detection object segmentation model to obtain a first target detection object prediction image corresponding to the target image, and obtain a region where the target detection object is located in the target image according to the first target detection object prediction image; and
object region segmentation code configured to cause the at least one processor to segment the target image according to the region where the target detection object is located to obtain the target detection object image.

### Claim 9 (depends on 8)

The Al-based object classification apparatus according to claim 8, wherein the image segmentation code further comprises:
image scaling code configured to cause the at least one processor to scale the target image to obtain a scaled image; and
the object region determining code is further configured to cause the at least one processor to:
input the scaled image into the target detection object segmentation model to obtain a second target detection object prediction image of the scaled image;
obtain region information of a region where a target detection object is located in the scaled image according to the second target detection object prediction image; and
obtain the region where the target detection object is located in the target image according to the region information.

### Claim 10 (depends on 8)

The Al-based object classification apparatus according to claim 8, wherein the object region determining code is further configured to cause the at least one processor to:
cut the target image into a plurality of target sub-images to be processed according to a cutting rule;
input the target sub-images into the target detection object segmentation model to obtain target detection object prediction sub-images corresponding to the target sub-images; and
stitch the target detection object prediction sub-images according to the cutting rule to obtain the first target detection object prediction image corresponding to the target image.

### Claim 11 (depends on 8)

The Al-based object classification apparatus according to claim 8, wherein the target detection object segmentation model comprises an encoder and a decoder; and
the object region determining code is further configured to cause the at least one processor to:
input the target image into the encoder of the target detection object segmentation model, and encode the target image by the encoder to obtain image feature information of the target image; and
input the image feature information into the decoder of the target detection object segmentation model, and
perform a decoding operation on the image feature information by the decoder to obtain the first target detection object prediction image corresponding to the target image.

### Claim 12 (depends on 7)

The Al-based object classification apparatus according to claim 7, wherein the feature object prediction model comprises a heat map prediction network; and the feature image obtaining code is further configured to cause the at least one processor to:
input the target detection object image into the heat map prediction network to obtain a feature object heat map corresponding to the target detection object image;
obtain a heat value of each pixel of the feature object in the target detection object image according to the feature object heat map; and
determine a region where the feature object is located from the feature object heat map according to the heat value to obtain the feature object segmentation image.

### Claim 13

A non-transitory computer-readable storage medium, storing a computer program that when executed by at least one processor causes the at least one processor to:
obtain a target image to be processed, the target image comprising a target detection object;
separate a target detection object image of the target detection object from the target image;
input the target detection object image into a feature object prediction model to obtain a feature object segmentation image of a feature object in the target detection object image;
obtain quantitative feature information of the target detection object according to the target detection object image and the feature object segmentation image; and
classify the target detection object image according to the quantitative feature information to obtain category information of the target detection object in the target image,
wherein the obtain quantitative feature information comprises:
determining a region where the feature object is located and an area value of the region where the feature object is located from the feature object segmentation image;
performing contour extraction on the region where the feature object is located;
calculating a perimeter of a contour of the region where the feature object is located, the perimeter of the contour being represented by a quantity of pixels of the contour of the region where the feature object is located, to calculate an area value of a circle with an equal perimeter; and
obtaining a ratio of the area value of the circle with the equal perimeter to the area value of the region where the feature object is located to obtain the quantitative feature information of the feature object in circularity.

### Claim 14 (depends on 13)

The non-transitory computer-readable storage medium according to claim 13, wherein the separate comprises:
inputting the target image into a target detection object segmentation model to obtain a first target detection object prediction image corresponding to the target image, and obtaining a region where the target detection object is located in the target image according to the first target detection object prediction image; and
segmenting the target image according to the region where the target detection object is located to obtain the target detection object image.

### Claim 15 (depends on 14)

The non-transitory computer-readable storage medium according to claim 14, wherein the inputting the target image into a target detection object segmentation model to obtain a first target detection object prediction image corresponding to the target image, and obtaining a region where the target detection object is located in the target image according to the first target detection object prediction image comprises:
scaling the target image to obtain a scaled image;
inputting the scaled image into the target detection object segmentation model to obtain a second target detection object prediction image of the scaled image;
obtaining region information of a region where a target detection object is located in the scaled image according to the second target detection object prediction image; and
obtaining the region where the target detection object is located in the target image according to the region information.

### Claim 16 (depends on 14)

The non-transitory computer-readable storage medium according to claim 14, wherein the inputting the target image into a target detection object segmentation model to obtain a first target detection object prediction image corresponding to the target image comprises:
cutting the target image into a plurality of target sub-images to be processed according to a cutting rule;
inputting the target sub-images into the target detection object segmentation model to obtain target detection object prediction sub-images corresponding to the target sub-images; and
stitching the target detection object prediction sub-images according to the cutting rule to obtain the first target detection object prediction image corresponding to the target image.