Image target detection method, system and device and storage medium

Abstract

The invention discloses an image target detection method, system and device and a storage medium. According to the method, a Faster-RCNN algorithm is used for sequentially carrying out feature map extraction and other processing steps on a to-be-processed image. The method further comprises the steps of performing multiple times of expansion convolution processing on the feature map, receiving multiple pieces of parallel feature information output by each time of expansion convolution processing, fusing the multiple pieces of parallel feature information to obtain first fusion feature information, fusing the first fusion feature information with the feature map to obtain second fusion feature information, and the like. According to the method, on the basis of the technical advantage that the existing Faster-RCNN algorithm can extract abundant image detail features, the method can overcome defect that the overall recognition precision of the Faster-RCNN is low due to the fact that the resolution of the feature map is too low and detail information in the to-be-processed image is lost too much, and the high target detection accuracy is obtained. The method is widely applied to the technical field of image processing.

Description

Technical field [0001] The present invention relates to the technical field of image processing, in particular to an image target detection method, system, device and storage medium. Background technique [0002] In the field of image recognition, object detection is often performed to detect the objects contained in the image. For example, for an image containing a car, after setting the car as a target, the target in the image needs to be detected, and the image is divided into a car area and a background area. Faster-RCNN is a commonly used algorithm for target detection. Its main steps include extracting feature maps, extracting regions of interest, mapping regions of interest, and input to the fully connected layer for processing and output image classification. The result and the regression result of the image frame can realize the target detection of the image. An important part of Faster-RCNN is the feature extraction network, which can extract feature maps from the ima...

AI Technical Summary

Problems solved by technology

In order to get more features, the convolutional neural network is designed to have a higher depth, which makes the resolution of the feature map too low, and the details in the image to be processed are lost too much, resulting in the overall recognition accuracy of Faster-RCNN lower

Images