Method for processing yarn spindle data, electronic device and storage medium

An electronic device, comprising:

• at least one processor;
• and
• a memory storing an instruction that, when executed by the at least one processor, causes the electronic device to:
• based on determining that a yarn spindle is transported into a detection area, perform a defect detection on the yarn spindle located in the detection area to obtain a target detection result of the yarn spindle, wherein the target detection result characterizes a defect degree of the yarn spindle;
• and
• based on determining that the target detection result meets a preset defect value, obtain a target level of the yarn spindle based on the target detection result of the yarn spindle and a preset level of the yarn spindle;
• wherein the instruction, when executed by the at least one processor, causes the electronic device to perform the defect detection on the yarn spindle located in the detection area to obtain the target detection result of the yarn spindle by:
• obtaining N images of the yarn spindle, wherein one image of the N images comprises at least a partial area of the yarn spindle, and N is greater than or equal to 3;
• inputting the N images into a target detection model to obtain a first detection result corresponding to each of the N images, wherein the target detection model is used to identify whether there are defects in a local area of the yarn spindle to obtain the first detection result, and the first detection result comprises at least one of:
• a quantity of defects, positions of defects, or types of defects;
• and
• obtaining the target detection result of the yarn spindle based on the first detection result corresponding to each of the N images;
• wherein the target detection model comprises at least a first network layer, a second network layer and a third network layer;
• wherein the first network layer is used to perform feature processing on an input image to obtain a low-level feature map, wherein the low-level feature map characterizes a feature map extracted after filtering out background noise in the input image,
• the second network layer is used to perform feature enhancement on at least key feature information in the low-level feature map to obtain a high-level feature map, and
• the third network layer is used to perform defect recognition based on the high-level feature map to obtain the first detection result;
• wherein the second network layer comprises at least a fifth sub-network layer, a sixth sub-network layer and a seventh sub-network layer,
• wherein the fifth sub-network layer is used to extract features from the low-level feature map, and fuse a plurality of feature maps extracted to obtain M initial fused feature maps, wherein M is an integer greater than or equal to 2,
• the sixth sub-network layer is used to extract key features from each of the M initial fused feature maps, and perform feature enhancement on key feature information extracted from each of the M initial fused feature maps to obtain M target enhanced feature maps, and
• the seventh sub-network layer is used to fuse the obtained M target enhanced feature maps to obtain the high-level feature map;
• wherein an i-th target enhanced feature map among the M target enhanced feature maps is obtained by:
• convolving an i-th initial fused feature map to obtain an i-th weight factor based on a convolution processing result;
• fusing an (i+1)-th initial fused feature map with the i-th initial fused feature map based on the i-th weight factor to obtain an i-th target fused feature map;
• extracting key features from the obtained i-th target fused feature map, performing feature enhancement on a plurality of extracted feature maps respectively to obtain a plurality of i-th initial enhanced feature maps;
• and fusing the plurality of i-th initial enhanced feature maps to obtain the i-th target enhanced feature map, wherein i is an integer greater than or equal to 1 and less than or equal to M−1;
• and
• based on i being equal to M, extracting key features from an M-th initial fused feature map;
• performing feature enhancement on a plurality of extracted feature maps respectively to obtain a plurality of M-th initial enhanced feature maps;
• and fusing the plurality of M-th initial enhanced feature maps to obtain an M-th target enhanced feature map.

A non-transitory computer-readable storage medium storing an instruction that, when executed, causes a computing device to:

• based on determining that a yarn spindle is transported into a detection area, perform a defect detection on the yarn spindle located in the detection area to obtain a target detection result of the yarn spindle, wherein the target detection result characterizes a defect degree of the yarn spindle;
• and
• based on determining that the target detection result meets a preset defect value, obtain a target level of the yarn spindle based on the target detection result of the yarn spindle and a preset level of the yarn spindle;
• wherein the instruction, when executed, causes the computing device to perform the defect detection on the yarn spindle located in the detection area to obtain the target detection result of the yarn spindle by:
• obtaining N images of the yarn spindle, wherein one image of the N images comprises at least a partial area of the yarn spindle, and N is greater than or equal to 3;
• inputting the N images into a target detection model to obtain a first detection result corresponding to each of the N images, wherein the target detection model is used to identify whether there are defects in a local area of the yarn spindle to obtain the first detection result, and the first detection result comprises at least one of:
• a quantity of defects, positions of defects, or types of defects;
• and
• obtaining the target detection result of the yarn spindle based on the first detection result corresponding to each of the N images;
• wherein the target detection model comprises at least a first network layer, a second network layer and a third network layer,
• wherein the first network layer is used to perform feature processing on an input image to obtain a low-level feature map, wherein the low-level feature map characterizes a feature map extracted after filtering out background noise in the input image,
• the second network layer is used to perform feature enhancement on at least key feature information in the low-level feature map to obtain a high-level feature map, and
• the third network layer is used to perform defect recognition based on the high-level feature map to obtain the first detection result;
• wherein the second network layer comprises at least a fifth sub-network layer, a sixth sub-network layer and a seventh sub-network layer,
• wherein the fifth sub-network layer is used to extract features from the low-level feature map, and fuse a plurality of feature maps extracted to obtain M initial fused feature maps, wherein M is an integer greater than or equal to 2,
• the sixth sub-network layer is used to extract key features from each of the M initial fused feature maps, and perform feature enhancement on key feature information extracted from each of the M initial fused feature maps to obtain M target enhanced feature maps, and
• the seventh sub-network layer is used to fuse the obtained M target enhanced feature maps to obtain the high-level feature map;
• wherein an i-th target enhanced feature map among the M target enhanced feature maps is obtained by:
• convolving an i-th initial fused feature map to obtain an i-th weight factor based on a convolution processing result;
• fusing an (i+1)-th initial fused feature map with the i-th initial fused feature map based on the i-th weight factor to obtain an i-th target fused feature map;
• extracting key features from the obtained i-th target fused feature map, performing feature enhancement on a plurality of extracted feature maps respectively to obtain a plurality of i-th initial enhanced feature maps;
• and fusing the plurality of i-th initial enhanced feature maps to obtain the i-th target enhanced feature map, wherein i is an integer greater than or equal to 1 and less than or equal to M−1;
• and
• based on i being equal to M, extracting key features from an M-th initial fused feature map;
• performing feature enhancement on a plurality of extracted feature maps respectively to obtain a plurality of M-th initial enhanced feature maps;
• and fusing the plurality of M-th initial enhanced feature maps to obtain an M-th target enhanced feature map.