CN111474186A - X-ray imaging and CNN express package contraband detection method - Google Patents

Abstract

The invention discloses an X-ray imaging and CNN express parcel contraband detection method, which relates to the fields of optical engineering and artificial intelligence, and includes the following steps: (1) obtaining the X-ray image information of the parcel; (2) obtaining the X-ray image sample of the parcel (3) Feature extraction of package X-ray images; (4) Construction of random forest contraband package recognition model. The present invention adopts CNN-RF to construct a detection model for express parcel contraband, and proposes a new convolutional neural network model that can extract more and more effective feature information, avoid overfitting, and use two GPUs for training It greatly improves the speed of training, and is very suitable for accurate and rapid detection of contraband in express parcels.

Description

A X-ray Imaging and CNN-based Express Parcel Contraband Detection Method

technical field

The invention relates to the fields of optical engineering and artificial intelligence, and mainly relates to a method for detecting whether express packages contain contraband.

Background technique

With the economic development in recent years, the express delivery market has become increasingly prosperous. The express delivery industry is a labor-intensive industry with long working hours and complicated processes. Due to the lack of professional knowledge of personnel when collecting parcels at express delivery outlets, there may be cases of receiving contraband, which will have an unnecessary impact on economic and social development and public safety.

Artificial intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. It mainly includes robotics, image processing, natural language processing, expert systems, etc. Convolutional Neural Networks (CNN) is a class of Feedforward Neural Networks (Feedforward Neural Networks) that includes convolutional computation and has a deep structure. It is one of the representative algorithms of deep learning. Natural language processing has a wide range of applications.

Optical engineering refers to a class of engineering that applies optical theory to practical applications. Optical engineering designs optical instruments, such as lenses, microscopes, and telescopes, but also includes other devices that exploit the properties of optics. In addition, optical engineering also studies optical sensors and related measurement systems, lasers, optical fiber communications and optical discs (eg CD, DVD), etc. X-rays, also known as Roentgen rays, are invisible to the naked eye, but can cause certain compounds to fluoresce or make photographic negatives sensitive. At the same time, X-rays can also penetrate substances, and have a wide range of applications in security inspection, industrial flaw detection, medical treatment and other fields.

SUMMARY OF THE INVENTION

In view of the above problems, the present invention provides a simple and fast method for identifying contraband without unpacking.

In order to achieve the above-mentioned purpose, the technical solution adopted in the present invention is: a method for detecting whether the express parcel of X-ray imaging and CNN contains contraband, comprising the following steps:

(1) Acquisition of package X-ray image information.

(2) The division of the wrapped X-ray image samples.

(3) Feature extraction of wrapped X-ray images.

(4) Construction of a random forest contraband package recognition model.

Preferably, in the step (1), the image information of the package is obtained by using X-ray imaging technology, and the image data set of the package is obtained.

Preferably, in the step (2), the collected package image data is divided into independent and non-repetitive training sets and test sets according to a certain proportion by random sampling.

Preferably, in the step (3), the convolutional neural network (Convolutional Neural Networks, CNN) used to extract the information features of the X-ray image is a network containing 8 training parameters, and the network structure includes a convolutional layer, a local Response normalization layer, pooling layer, fully connected layer, etc.

Preferably, in the step (4), use CNN to extract X-ray image features to construct a random forest (RF) contraband recognition model on the training set, determine the parameters of the recognition model, and then use the test set to detect the recognition effect , to verify the model performance.

Through the above technical solutions, the beneficial effects of the present invention are that a new convolutional neural network model can be proposed to obtain more and more effective image information, which is conducive to the rapid and accurate detection of contraband in the package.

Description of drawings

FIG. 1 is a flow chart of a method for identifying contraband in express parcels in a case of implementing the present invention.

FIG. 2 is a schematic diagram of the structure of a convolutional neural network used for extracting image features in an embodiment of the invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the specific embodiments and the accompanying drawings.

The present invention works under the Windows 10 environment, uses Keras for analysis, and uses TensorFlow as its backend.

An X-ray imaging and CNN express contraband detection method of the present invention comprises the following steps:

(1) Acquisition of package X-ray image information.

(2) The division of the wrapped X-ray image samples.

(3) Feature extraction of wrapped X-ray images.

(4) Construction of a random forest contraband package recognition model.

In order to make the purposes, technical solutions and advantages of the present invention clearer and clearer, a method for detecting express contraband by X-ray imaging and CNN of the present invention will be further described in detail below in conjunction with specific examples and with reference to the accompanying drawings. An X-ray imaging and CNN express contraband detection method is invented, and the identification steps are shown in Figure 1:

101 The acquisition of the X-ray image of the package is obtained through the X-ray imaging equipment to obtain the X-ray image and information of the package.

102 The division of the X spectral image samples of the package, using the random sampling method to divide the spectral image data of the package containing contraband and not containing the contraband into independent test sets and training sets according to the ratio of 70% of the training set and 30% of the test set .

103 Convolutional neural network wraps X-ray image feature extraction. The convolutional neural network used to extract X-ray image features is a network containing 8 training parameters. The network structure includes convolutional layer, local response normalization layer, pooling Layer, fully connected layer, etc., to determine the category is as follows:

The network contains 8 layers with weights; the first 5 layers are convolutional layers and the remaining 3 layers are fully connected layers. Connect the output features of the last fully-connected layer with a dimension of 1000 (or custom), and then serve as the input of the random forest.

Convolutional layer C1, the processing flow of this layer is: convolution-->ReLU-->pooling-->normalization.

Convolutional layer, the input is 227×227×3, using 96 convolution kernels of 11×11×3.

ReLU, input the FeatureMap output by the convolutional layer into the ReLU function.

Pooling layer, using 3×3 pooling units with stride 2.

Local response normalization layer, using k=2, n=5, α=10-4, β=0.75 for local normalization, the output is 27×27×96, the output is divided into two groups, and the size of each group is 27×27×48.

Convolutional layer C2, the processing flow of this layer is: convolution-->ReLU-->pooling-->normalization

Convolutional layer, the input is 2 sets of 27×27×48. Use 2 groups, each with 128 convolution kernels of size 5 × 5 × 48, with edge padding = 2, and the stride of convolution is 1.

ReLU, input the FeatureMap output by the convolutional layer into the ReLU function.

Pooling layer, the size of the operation is 3×3, and the stride is 2.

Local response normalization layer, using k=2, n=5, α=10-4, β=0.75 for local normalization, the output is 13×13×256, the output is divided into 2 groups, and the size of each group is 13×13×128.

Convolution layer C3, the processing flow of this layer is: convolution-->ReLU.

For the convolution layer, the input is 13×13×256, using 2 groups of 384 convolution kernels of size 3×3×256, with edge padding padding=1, and the convolution stride is 1.

ReLU, input the FeatureMap output by the convolutional layer into the ReLU function.

Convolution layer C4, the processing flow of this layer is: convolution --> ReLU This layer is similar to C3.

Convolutional layer, the input is 13×13×384, divided into two groups, each group is 13×13×192. Use 2 groups, each group has 192 convolution kernels of size 3×3×192, with edge padding padding=1, the stride of convolution is 1, and the output FeatureMap is 13×13times384, which is divided into two groups, each of which is 13×13×192.

ReLU, input the FeatureMap output by the convolutional layer into the ReLU function.

Convolutional layer C5, the processing flow of this layer is: Convolution-->ReLU-->Pooling.

The convolutional layer, the input is 13×13×384, is divided into two groups, each group is 13×13×192. Use 2 groups, each group is 128 convolution kernels of size 3×3×192, with edge padding padding=1, and the convolution stride is 1.

ReLU, input the FeatureMap output by the convolutional layer into the ReLU function.

Pooling layer, the size of the pooling operation is 3×3, and the stride is 2, and the output after pooling is 6×6×256.

Fully connected layer FC6, the process of this layer is: (convolution) fully connected-->ReLU-->Dropout

(Convolutional) Fully connected: The input is 6×6×256, and this layer has 4096 convolution kernels.

ReLU, these 4096 operation results generate 4096 values through the ReLU activation function.

Dropout, suppresses overfitting, randomly disconnecting some neurons or not activating some neurons.

The full connection layer FC7, the process is: full connection-->ReLU-->Dropout

Fully connected, the input is a vector of 4096.

ReLU, these 4096 operation results generate 4096 values through the ReLU activation function.

Dropout, suppresses overfitting, randomly disconnects some neurons or does not activate some neurons.

In the output layer, the 4096 data output by the seventh layer is fully connected with the 1000 neurons in the eighth layer, and after training, 1000 float values are output, which is the extracted feature.

Since AlexNet runs on two GPUs, the pixel data needs to be divided into two groups and stored in the two GPUs respectively. In the convolutional layer C2, the convolutional layer C4, and the convolutional layer C5, the pixel data of the previous layer is stored in this group. The data is connected within the GPU, the convolutional layer C3 is fully connected with the previous two layers, and the full connection is two GPUs fully connected.

104 Random forest contraband package recognition model construction, using CNN to extract X-ray image features to build a random forest contraband recognition model on the training set, determine the parameters of the recognition model, and then use the test set to detect the recognition effect and verify the performance of the model.

Through the above technical solutions, the beneficial effects of the present invention are that a new convolutional neural network model can be proposed to obtain more and more effective image information, which is conducive to the rapid and accurate detection of contraband in the package.

It will be apparent to those skilled in the art that the present invention is not limited to the details of the above-described exemplary embodiments, but that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments are to be regarded in all respects as illustrative and not restrictive, and the scope of the invention is to be defined by the appended claims rather than the foregoing description, which are therefore intended to fall within the scope of the claims. All changes within the meaning and scope of the equivalents of , are included in the present invention. Any reference signs in the claims shall not be construed as limiting the involved claim.

In addition, it should be understood that although this specification is described in terms of embodiments, not each embodiment only includes an independent technical solution, and this description in the specification is only for the sake of clarity, and those skilled in the art should take the specification as a whole , the technical solutions of each embodiment can also be appropriately combined to form other implementations that can be understood by those skilled in the art.

Claims (5)

1. A method for detecting whether an express package of X-ray imaging and CNN contains contraband or not comprises the following steps:

(1) and acquiring package X-ray image information.

(2) And (4) dividing the parcel X-ray image sample.

(3) And (5) extracting the characteristics of the package X-ray image.

(4) And constructing a random forest contraband parcel identification model.

2. The method of claim 1 for detecting whether an express package of X-ray imaging and CNN contains contraband, comprising: in the step (1), the image information of the package is acquired by using an X-ray imaging technology, so as to obtain an image data set of the package.

3. The method of claim 1 for detecting whether an express package of X-ray imaging and CNN contains contraband, comprising: in the step (2), the acquired parcel image data is divided into independent and non-repetitive training sets and test sets according to a certain proportion by adopting a random sampling mode.

4. The method of claim 1 for detecting whether an express package of X-ray imaging and CNN contains contraband, comprising: in the step (3), a Convolutional Neural Network (CNN) for extracting information features of the X-ray image is a network including 8 training parameters, and a network structure includes a convolutional layer, a local response normalization layer, a pooling layer, a full connection layer, and the like.

5. The method of claim 1 for detecting whether an express package of X-ray imaging and CNN contains contraband, comprising: in the step (4), a random forest (Randomforest) contraband identification model is constructed on a training set by using the characteristic of the X-ray image extracted by the CNN, the parameter of the identification model is determined, then the identification effect is detected by using a test set, and the performance of the model is verified.

Images