Research On Fast Target Recognition Of Convolution Neural Network Based On Prior Knowledge

In recent years,with the development of artificial intelligence technology,the application of convolutional neural network to target recognition has become the focus of research.The target image data are collected as the input of the convolutional neural network,and the corresponding tag value is set as the ideal output for the input image data.And the most basic idea of convolutional neural network is to fit the nonlinear mapping relation between the given input and the ideal output according to the given input and the ideal output.When the ambient light does not change much,the target image is collected and used as the input of the convolutional neural network to learn and train the network model,and the convolutional neural network could be better complete the target recognition task.However,when the ambient lighting condition changes,the collected dynamic target image changes and part of the image information is lost,making it very difficult to identify the target object in the image.In order to solve this problem,this paper proposes a fast target recognition research based on the prior knowledge of convolutional neural network.Combining with the working characteristics of convolutional neural network,the prior knowledge is put forward,and the prior knowledge is introduced into the convolutional neural network,which effectively solves the problem that the target is difficult to be recognized under the condition of illumination change,and greatly improves the speed of target recognition.Its main research work is as follows:(1)The dynamic target recognition based on the prior knowledge of convolutional neural network is studied.Firstly,the dynamic characteristics of the target image under ambient illumination are analyzed.And then the clips circular objects the biggest external rectangular algorithm is proposed,and based on the convolution neural network,two kinds of prior knowledge are introduced.One is to define prior knowledge based on feature matching method,and the other is to take the test results of network model as prior knowledge through pre-training convolutional neural network model,and then build the convolutional neural network dynamic target recognition based on prior knowledge model.The construction process of the model is as follows: The image of the target object is collected in any light condition,and the image of the target object is used as the input data of the convolutional neural network to train the convolutional neural network model,and the standard light condition is determined according to the test results of the network model.The standard light condition is taken as the prior knowledge.In the standard light condition,the prior knowledge based on the definition of feature matching method is introduced to collect the image of the target object and use it as the input data of the convolutional neural network to train the convolutional neural network model.At the same time,the prior knowledge based on the definition of feature matching method is introduced to process the test image data,which is then put into the trained network model for testing,and the target recognition result is obtained.(2)The experimental environment was set up in the laboratory,and the experimental object was the ball used in the Chinese robot competition.The target ball images collected in the experimental environment are used to verify the target recognition algorithm proposed in this paper under the condition of light variation.The experimental results are given and compared with the recognition effect of the typical convolutional neural network model.The experiment verifies the effectiveness of the algorithm proposed in this paper.