Fault classification visualization method based on target detection and fine grit recognition

Abstract

The invention provides a fault classification visualization method based on target detection and fine grit recognition, which adopts a two-stage method, and comprises the following steps: firstly, positioning the position of a fault through target detection, and storing coordinate information as a first stage; after a model is trained by using a fine-grained algorithm, the position where a fault occurs and whether the fault exists can be marked by using coordinate information of the first stage and the training model, namely, visual operation is carried out on a classified result, and the effect the same as that of target identification is achieved. The visual operation of the method is to explicitly mark the position and the category of the fault in the picture. According to the classification method, a weak supervision method is used in the fine-grained recognition stage, the classification method is efficient and rapid, wide in application range and low in computing power consumption, compared with a single target detection method and a single fine-grained recognition algorithm in the present stage, the accuracy of industrial fault classification is improved, explicit positioning can be carried out on faults in pictures, and the classification efficiency is improved. The industrial requirements are met.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a fault classification visualization method based on target detection and fine-grained identification. Background technique [0002] Fine-grained image recognition is an interesting, fundamental and challenging problem in computer vision and has been an active area of ​​research for decades. Fine-grained image analysis mainly deals with multiple sub-categories belonging to the same meta-category, such as birds, dogs, cars, etc. Fine-grained classification needs to be able to find the subtle differences between objects, extract the differences into the features of the objects, and then realize the detection of objects. Classification. The emergence of deep learning technology has accelerated significant breakthroughs in the fine-grained field. [0003] For target detection technology, target detection is a relatively simple task in computer vision, which is ...

AI Technical Summary

Problems solved by technology

[0008] The purpose of the present invention is to provide a fault classification visualization method based on target detection and fine-grained recognition, so as to solve the problem that the existing single target detection algorithm has low fault recognition efficiency, and the position and target position cannot be directly marked in the classification results. Technical issues with classification results

Images