Research On Intelligent Detection System Of Empty Container Based On Power Sound Excitation

At the entry and exit gates,container smuggling is a serious problem,so the identification of empty containers is necessary.At present,there are two main noninvasive detection technologies for containers: X-ray detection and ?-ray detection.Both of these are radiation detection methods,and the state of the container body can be clearly observed using the ray reconstruction image technology,which has high accuracy but will cause great harm to the human body.At the same time,these two methods are expensive and consume a lot of energy.In this paper,from the perspective of cost and harmlessness to the human body,an empty box recognition system based on strong sound excitation is proposed.The sound signal of a high sound pressure level emitted by a strong sound generator is applied to the side of the box,and the other side of the box is used.The noncontact micro-acoustic sensor collects the acoustic vibration signal of the cabinet,and judges whether it is an empty tank by means of the acoustic signal.Through theoretical derivation and simulation analysis,it can be seen that the acoustic vibration mode and decay time of the container in the empty and non-empty container states are different,that is,the time domain and frequency domain of the two states are different.When the data is actually collected,the received sound signal is susceptible to environmental noise interference,so the noise reduction method is analyzed and studied.In the case of simple noise interference,both the noise and the acoustic vibration signal of the cabinet have a strong non-Gaussian nature.The collected signal is actually a mixed signal of these two signals.The FastICA algorithm is used to maximize the non-Gaussian nature of the collected signal.Noise reduction processing.The purpose of this article is to distinguish between empty and non-empty containers,which is essentially a binary classification problem.The deep convolutional neural network algorithm has achieved great success in the field of image recognition,etc.In this paper,the one-dimensional audio data after noise reduction is converted into a Mel spectrum chart containing time domain and frequency information,and the convolutional neural network is used to pass the picture The data is learned and trained.In the case of simple interference sources,the system has a high recognition accuracy rate,so this method can identify empty boxes and non-empty boxes.This paper completes the software design and implementation of the empty container identification system of the container,which mainly includes the following four parts:(1)the implementation of the data acquisition system;(2)the algorithm implementation of the Mel spectrum audio classification algorithm based on the deep convolution network;(3)The realization of the dual-channel blind source separation algorithm;(4)The design of the host computer software system based on C# WPF,which mainly includes identification,training data collection and storage management.Finally,this paper completes the design of the empty container software system and conducts experimental verification.