Target image detection method, device and system

Abstract

The invention discloses a target image detection method, a device and a system. The method comprises: inputting an image to be processed into a full convolution neural network model, so that the fullconvolution neural network model identifies an image to be processed to output candidate position information of each candidate target image; and outputting the candidate position information of the candidate target image to be processed; extracting candidate target images corresponding to each candidate position information; enlarging each candidate target image to form a standard image, and recording an enlargement multiple corresponding to each standard image; aiming at the standard image, the standard image being inputted into the cascade convolution neural network model, so that the cascade convolution neural network model recognizing the standard image and outputting the current position information and classification probability of the suspicious target image. According to the current position information and classification probability of the suspicious target image, the magnification corresponding to each standard image and the candidate position information of each candidate target image, the position information of the target image is determined. The technical proposal of the invention can more accurately determine the position information of the target image.

Description

technical field [0001] The present invention relates to the field of computer technology, in particular to a target image detection method, device and system. Background technique [0002] At present, the industry usually uses a fully convolutional neural network to identify the image to be processed, so as to determine the position information of each target image carried by the image to be processed on the image to be processed. For example, determine the position information of the license plate area image, the headlight area image or the car window area image on the image to be processed, and then further track or track one or more vehicles based on the position information of the target image on the image to be processed. Other business. [0003] When the fully convolutional neural network recognizes the image to be processed, it is necessary to extract the feature information of the image to be processed (for example, the edge point, texture, color, etc. position inf...

AI Technical Summary

Problems solved by technology

[0003] When the fully convolutional neural network recognizes the image to be processed, it is necessary to extract the feature information of the image to be processed (for example, the edge point, texture, color, etc. The position information on the image; if the size of a target image is relatively small, the full convolutional neural network cannot accurately extract the feature information corresponding to the target image, resulting in the output of the full convolutional neural network corresponding to the target image The location information is inaccurate, that is, the location information output by the full convolutional neural network corresponds to the current image area in the image to be processed and includes non-target image areas other than the target image

Images