Research On Object Material Recognition Method Based On Visual And Auditory Fusion

With the continuous development of artificial intelligence technology,the ability of target classification and recognition based on visual and auditory signals has been greatly improved in recent years.Object material identification is one of the important basic technologies in the field of radioactive solid waste decommissioning and the basis of waste classification.The decommissioning of nuclear facilities has produced a large number of radioactive solid wastes,which need to be classified and treated safely and reasonably.However,the radioactive solid wastes in the context of decommissioning have many kinds,complex shapes and large deformations,which make it very difficult to classify and recognize materials based on visual images.In view of the above problems,this paper will study material recognition from the perspective of multimodal fusion,propose a method of object material recognition based on visual and auditory fusion,and build a depth neural network model for object material recognition.The research work of this paper includes the following aspects:(1)To explore the method of object material recognition based on visual image.Firstly,the material image database is established,and then the material recognition model based on the depth convolution neural network is constructed,and the parameters of the model are designed and optimized.Experiments on the self-built image database show that the material recognition method based on the deep convolution neural network is feasible and effective.This method is compared with the support vector machine algorithm,and the validity of the proposed model is verified.(2)The object material recognition method based on sound signal is explored.First of all,the establishment of material sound database is completed.The MFCC features and their difference coefficients of different material sound signals are extracted from the timefrequency characteristics of different material sound signals,and the construction of object material identification model and parameter optimization design are completed based on convolution neural network method.The experimental results show that the recognition method based on "F1" feature and convolution neural network is effective,and the validity of the model is verified by comparative experiments.(3)To explore the method of object material recognition based on visual and auditory fusion.According to the complementary features of audio-visual signals,the fusion of audio-visual features based on feature level is completed,and a deep network recognition model is established to further improve the performance of the object material recognition model.The effectiveness and feasibility of the method are verified by off-line recognition experiments.Finally,the design and implementation of object material identification software are completed.The experimental results show that the accuracy of the object material recognition method based on visual and auditory fusion in the self-built material database can reach 99.14%.It has a certain practical application value to realize the intelligent sorting of radioactive solid waste.