Target detection method and system based on low-resolution images

Abstract

The present invention provides a method and system for target detection based on low-resolution images, which relate to the technical field of target detection, wherein the method includes the following steps: acquiring a low-resolution image of the target to be detected; performing feature analysis on the low-resolution image Extract the corresponding feature map; process the feature map to get the corresponding detection frame; process the detection frame to get the coordinates and classification probability score of the detection frame; detect and classify the target to be detected according to the coordinates of the detection frame and classification probability score. The present invention can extract feature maps with deep semantics without enlarging low-resolution images, thereby effectively reducing detection time without loss of detection accuracy.

Description

technical field [0001] The invention relates to the technical field of target detection, in particular to a low-resolution image-based target detection method and a low-resolution image-based target detection system. Background technique [0002] In recent years, low-resolution object recognition technology based on deep learning has continued to receive extensive attention from academia and industry. Taking handwritten digit string recognition as an example, the intuitive way to understand this task is to first find the digits in a picture and then recognize them, which is very similar to target detection. However, these methods based on target detection are detected by enlarging the image to a certain size, and there are some problems: first, after enlarging the low-resolution image, extracting features from it will lead to unnecessary time consumption; second, If the low-resolution image is not enlarged and the image is directly sent to the network, a small feature map w...

AI Technical Summary

Problems solved by technology

However, these methods based on target detection are detected by enlarging the image to a certain size, and there are some problems: first, after enlarging the low-resolution image, extracting features from it will lead to unnecessary time consumption; second, If the image is directly sent to the network without enlarging the low-resolution image, a small feature map will be generated in the process of extracting image features, so that the correct detection frame cannot be generated.

Images