Target detection method and device

Abstract

The embodiment of the invention provides a target detection method and device, and the method comprises the steps: detecting the size of a preset to-be-detected target, and determining the size of a first convolution kernel; carrying out the convolution operation of an original image through the first convolution kernel, and obtaining an initial characteristic pattern of the original image; determining a characteristic extraction model comprising a plurality of preset characteristic extraction layers, wherein the characteristic extraction model is used for outputting the effective characteristic information, according with a preset task, of the image according to the initial characteristic pattern through the convolution operation of the plurality of preset characteristic extraction layers; obtaining the type of the to-be-detected target and the original information of the to-be-detected target in the original image according to the characteristic information through a preset multi-class regression algorithm and a preset position regression algorithm. The method can improve the actual operation efficiency of target detection.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a target detection method and device. Background technique [0002] Object detection is an important technology involving computer vision and image processing. Its main purpose is to detect certain types of object instances in images and videos, such as humans, buildings, vehicles, etc. Object detection technology is currently used in image retrieval, video surveillance, and human-computer interaction. Traditional object detection methods are mostly carried out by artificially constructing features combined with machine learning. This type of method is suitable for situations where the scene is fixed, the target pose does not change much, and the lighting conditions do not change much. [0003] In view of the problems existing in the above-mentioned traditional target detection methods, the method of extracting features by deep learning came into being. The mai...

AI Technical Summary

Problems solved by technology

[0004] However, the method based on the convolutional neural network has an extremely large amount of calculation when calculating the feature image of each layer, far exceeding DSP (Digital Signal Processor, digital signal processor) and ARM (Advanced Reduced Instruction Set Computer Machines, simplified instruction Integrating the processing power of the computer microprocessor) chip, resulting in the actual operation efficiency of target detection is very low, and when reducing the amount of calculation, it is easy to affect the effect of feature extraction

Images