Intelligent identification method for detecting target feature quantity and statistical information

Abstract

The invention discloses an intelligent identification method for detecting the number of target features and statistical information, and the method comprises the steps: comparing the feature values of an input layer and a hidden layer of a convolutional neural network after batch normalization with the same set threshold value, and mapping the feature values according to a comparison result; and mapping the target features of the position to a value close to 1 if the target features are greater than the threshold, otherwise, mapping the target features to a value close to 0, and further counting the number of the target features in the input layer and the hidden layer. According to the method, statistics of the number of the target features in the to-be-detected object can be achieved through the hidden layer of the convolutional neural network, statistics based on the number of the target features can be reversely applied to the target detection process, and the detection accuracy is improved.

Description

technical field [0001] This application relates to an intelligent recognition method for detecting the quantity and statistical information of target features. Background technique [0002] Convolutional Neural Networks (CNN) is a type of feed-forward neural network that includes convolution calculations and has a deep structure. It is one of the representative algorithms for deep learning. The convolutional neural network imitates the biological visual perception mechanism, and can perform supervised learning and unsupervised learning. The convolution kernel parameter sharing in the hidden layer and the sparsity of the inter-layer connection enable the convolutional neural network to use a small calculation. Quantitative learning of lattice features, such as pixels and audio, has a stable effect and has no additional feature engineering requirements on the data. The current convolutional neural network technology is difficult to conveniently detect the number of targets, o...

AI Technical Summary

Problems solved by technology

The current convolutional neural network technology is difficult to conveniently detect the number of targets, or detect the number of certain features or hidden features, and it is difficult to achieve target detection that depends on the number of features

Images