Intelligent image identification system and method based on terahertz imaging

Abstract

The invention discloses an image intelligent identification system and method based on terahertz imaging, and particularly relates to the field of terahertz image intelligent identification, and the system comprises an image collection module, an image processing module, a first loss function construction module, an image detection module, a mark identification module, and a judgment identification module. According to the invention, the terahertz original image of the contraband in the luggage case is collected, the image is optimized and enhanced, then the first loss function is constructed, whether the contraband is carried is detected according to the image detection module, and if the contraband is carried, the position of the contraband is found and identified according to the first loss function and the terahertz original image. And finally, the terahertz original image and the original image with the mark are displayed to monitoring personnel, so that the monitoring personnel can judge whether the central server has an identification error condition or not.

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology, and more specifically, to an image intelligent identification system and method based on terahertz imaging. Background Art

[0002] Terahertz imaging is a non-contact imaging technology that has emerged in recent years. It boasts strong penetrating properties, is transparent to opaque objects, and is non-invasive to flexible objects. Terahertz waves, with a wavelength between millimeter waves and infrared, are suitable for detecting objects that pose safety risks in fields such as biology, medicine, food, and security.

[0003] Currently, terahertz imaging technology has been applied in fields such as biomedicine, food safety, cultural relic preservation, and security inspection. In this area, it is primarily used to detect people and items carrying prohibited items, such as luggage and clothing. Therefore, intelligent image identification systems based on terahertz imaging can improve terahertz imaging technology, enabling its wider application in security inspection. Furthermore, advances in technologies such as deep learning and machine learning are enabling intelligent detection systems based on terahertz imaging to achieve higher accuracy and reliability. This intelligent image identification technology allows for a balance between safety and efficiency, and holds significant promise for the future development of security inspection technology.

[0004] However, there are still some shortcomings in its actual use. For example, existing terahertz security inspection equipment generally uses an image recognition system to automatically identify terahertz "snapshots". However, since the equipment's resolution, image processing and pattern recognition have certain errors in object characteristics and sizes, this method of automatically identifying terahertz "snapshots" by an image recognition system may, to a certain extent, result in certain missed detections and false detections. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide an image intelligent marking system and method based on terahertz imaging to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] Preferably, the image acquisition module is used by the client to control the terahertz scanning device to acquire a terahertz raw image of the prohibited items in the target suitcase and transmit the acquired image to the image processing module. The process of transmitting the image of the prohibited items in the suitcase to the client of the security inspector for display, so that the security inspector can determine whether the inspected suitcase contains prohibited items based on the image of the prohibited items, further includes: displaying a list of various dangerous goods icons on the display screen of the monitoring terminal for selection by the monitoring personnel; receiving the dangerous goods icon selected by the monitoring personnel from the list, and importing the selected dangerous goods icon into the corresponding mark on the image of the target suitcase.

[0008] Preferably, the image processing module is used to optimize, extract features, and perform image enhancement processing on the terahertz imaging data. The image processing module includes processing methods such as a forward filter, pseudo-colorization, binarization, and morphological processing. The forward filter is used to remove noise and clutter signals from the image data, the pseudo-colorization is used to improve the visual effect of the image, the binarization is used to extract image contour information, and the morphological processing is used to segment and reconstruct the image.

[0009] Preferably, a first loss function module is constructed for training an image recognition network using a plurality of terahertz imaging images as training samples, artificially constructing label images corresponding to each terahertz imaging image as a training sample, and constructing a first loss function, wherein the first loss function is a cross entropy loss function; the algorithm for constructing the first function is specifically as follows:

[0010] Where Li represents the situation loss function of the target suitcase image and its corresponding pre-stored logo image corresponding to the i-th sub-block, α and τ are hyperparameters greater than 0, ω i is the mean grayscale value of the i-th sub-block in the target suitcase image, ω i0 is the mean grayscale value of the corresponding sub-block of the target suitcase image in the identification image, σ is the covariance of the corresponding sub-block of the target suitcase image and its corresponding identification image, is the square of the grayscale value variance of the i-th sub-block in the target suitcase image, is the square of the grayscale value variance of the corresponding sub-block in the identification image for the i-th sub-block in the target suitcase image.

[0011] Preferably, the image detection module is used to identify whether the detected target suitcase carries contraband through the terahertz image.

[0012] Preferably, the marking recognition module is used to mark the location of the item on the pre-stored contraband image according to the marked terahertz original image: in the marking recognition module, if contraband is carried, the location of the item is found and marked in the terahertz original image according to the first loss function and the terahertz original image, and the marked terahertz original image is generated, and the terahertz original image and the marked terahertz original image are sent to the client.

[0013] Preferably, the judgment and recognition module is used to display the terahertz original image and the marked original image to the monitoring personnel, so that the monitoring personnel can judge whether the central server has made an identification error. The monitoring terminal marks the location of the item on the pre-stored suitcase image according to the marked terahertz original image, and displays the terahertz original image and the marked original image to the monitoring personnel, so that the monitoring personnel can judge whether the central server has made an identification error. If there is an identification error, the monitoring terminal changes the mark on the image of the prohibited items in the suitcase according to the instruction input by the monitoring personnel, and sends the changed image of the prohibited items in the suitcase to the client of the security personnel for display, so that the security personnel can judge whether the tested suitcase carries prohibited items based on the changed image of the prohibited items in the suitcase, thereby improving the accuracy of security inspection image recognition and avoiding missed detection and false detection.

[0014] Technical effects and advantages of the present invention:

[0015] Efficient and convenient: Terahertz imaging technology can quickly scan the object being inspected and obtain its characteristic information, without the need for tedious operations such as unpacking the object, thereby improving security inspection efficiency and convenience. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of the overall structure of the present invention.

[0017] Figure 2 It is a schematic diagram of the system flow structure of the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] See also Figure 1As shown, the present invention provides an image intelligent identification system based on terahertz imaging, which includes an image acquisition module, an image processing module, a first function construction module, an image detection module, a mark identification module, and a judgment and recognition module.

[0020] The image acquisition module is connected to the image processing module, the image processing module is connected to the first function construction module, the first function construction module is connected to the image detection module, the image detection module is connected to the mark identification module, and the mark identification module is connected to the judgment and recognition module.

[0021] The image acquisition module is used for the client to control the terahertz scanning device to acquire the original terahertz image of the prohibited items in the target suitcase and transmit the acquired image to the image processing module; before the image of the prohibited items in the suitcase is sent to the client of the security personnel for display so that the security personnel can judge whether the tested suitcase carries prohibited items based on the prohibited items image, the method also includes: displaying a list containing various dangerous goods icons on the display screen of the monitoring end for selection by the monitoring personnel; receiving the dangerous goods icon selected by the monitoring personnel in the list, and importing the dangerous goods icon selected by the monitoring personnel into the corresponding mark on the image of the target suitcase.

[0022] The image processing module is used to optimize, extract features and perform image enhancement processing on the terahertz imaging data;

[0023] In one possible design, the method for optimizing, extracting features, and enhancing the terahertz imaging data is as follows:

[0024] The image processing module includes processing methods such as forward filter, pseudo colorization, binarization and morphological processing. The forward filter is used to remove noise and clutter signals in image data, the pseudo colorization is used to improve the visual effect of the image, the binarization is used to extract image contour information, and the morphological processing is used to segment and reconstruct the image.

[0025] The first loss function construction module is used to use multiple terahertz imaging images as training samples to train the image recognition network, artificially construct label images corresponding to each terahertz imaging image as a training sample, and construct a first loss function, which is a cross entropy loss function.

[0026] In one possible design, the algorithm of the first loss function is specifically:

[0027] Where Li represents the situation loss function of the target suitcase image and its corresponding pre-stored logo image corresponding to the i-th sub-block, α and τ are hyperparameters greater than 0, ω iis the mean grayscale value of the i-th sub-block in the target suitcase image, ω i0 is the mean grayscale value of the corresponding sub-block of the target suitcase image in the identification image, σ is the covariance of the corresponding sub-block of the target suitcase image and its corresponding identification image, is the square of the grayscale value variance of the i-th sub-block in the target suitcase image, is the square of the grayscale value variance of the corresponding sub-block in the identification image for the i-th sub-block in the target suitcase image.

[0028] The image detection module is used to identify whether the detected target suitcase carries prohibited items through the terahertz image.

[0029] The marker recognition module is configured to mark the location of an item on a pre-stored prohibited item image based on the marked terahertz original image. If a prohibited item is present, the module locates the item in the terahertz original image based on the first loss function and the marked terahertz original image, marks the location of the item, generates a marked terahertz original image, and transmits the marked terahertz original image and the marked terahertz original image to the client.

[0030] The client also provides a play button for continuously viewing the terahertz raw image displayed on the screen, a pause button for controlling the monitoring end to stop continuously viewing the terahertz raw image, an up button for viewing the previous terahertz raw image frame, a down button for viewing the next terahertz raw image frame, and a function button for zooming in on a specific part of the image. Client monitors can use these buttons to control the monitoring end to display the terahertz raw image they want to view and zoom in on a selected part of the terahertz raw image, helping them to more accurately determine whether there have been missed or false detections.

[0031] The judgment and recognition module is used to display the terahertz original image and the original image with the mark to the monitoring personnel, so that the monitoring personnel can judge whether the central server has any recognition error.

[0032] The monitoring end marks the location of the item on a pre-stored image of the suitcase based on the marked original terahertz image, and simultaneously displays the original terahertz image and the marked original image to a monitoring person, allowing the monitoring person to determine whether the central server has made a recognition error. If a recognition error occurs, the monitoring end changes the mark on the image of the prohibited items in the suitcase based on an instruction input by the monitoring person, and sends the changed image of the prohibited items in the suitcase to a client of a security inspector for display, allowing the security inspector to determine whether the suitcase under inspection carries prohibited items based on the changed image of the prohibited items in the suitcase, thereby improving the accuracy of security inspection image recognition and avoiding missed detection and false detection.

[0033] See also Figure 2 As shown, in this embodiment, it should be specifically explained that the present invention provides an intelligent identification system based on terahertz imaging, including the following steps:

[0034] S1: The client controls the terahertz scanning device to collect the terahertz original image of the contraband in the target suitcase;

[0035] S2: Optimize, extract features and perform image enhancement on the terahertz imaging data; including methods such as denoising, optimization, blur kernel processing, and morphological analysis

[0036] S3: using multiple terahertz imaging images as training samples to train an image recognition network, artificially constructing label images corresponding to each terahertz imaging image as a training sample, and constructing a first loss function, which is a cross entropy loss function;

[0037] S4: Identifying whether the target suitcase carries prohibited items through the terahertz image;

[0038] S5: Marking the location of the object on the pre-stored contraband image according to the marked terahertz original image;

[0039] S6: Displaying the terahertz original image and the original image with the mark to a monitoring person, so that the monitoring person can determine whether the central server has any recognition error.

[0040] In the step S1, the terahertz scanning device includes a terahertz source and a terahertz detector, the terahertz source is used to emit a terahertz signal, and the terahertz detector is used to receive a terahertz wave reflection signal.

[0041] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An image intelligent identification system and method based on terahertz imaging, characterized by: Image acquisition module: used by the client to control the terahertz scanning device to collect the original terahertz image of the prohibited items in the target suitcase, and transmit the collected image to the image processing module; Image processing module: used for optimizing, extracting features and performing image enhancement processing on the terahertz imaging data; Constructing a first loss function module: used to train an image recognition network using multiple terahertz imaging images as training samples, artificially constructing a label image corresponding to each terahertz imaging image as a training sample, and constructing a first loss function, which is a cross entropy loss function; Image detection module: used to identify whether the target suitcase carries prohibited items through the terahertz image; Marking recognition module: used to mark the location of the object on the pre-stored contraband image according to the marked terahertz original image; Judgment and recognition module: used for displaying the terahertz original image and the original image with the mark to the monitoring personnel, so that the monitoring personnel can judge whether the central server has any recognition error.

2. The image intelligent marking system based on terahertz imaging according to claim 1, characterized in that: Before transmitting the prohibited item image in the suitcase to the security inspector's client for display so that the security inspector can determine whether the inspected suitcase carries prohibited items based on the prohibited item image, the method further includes: displaying a list of various dangerous item icons on the display screen of the monitoring terminal for selection by the monitoring inspector; receiving a dangerous item icon selected by the monitoring inspector from the list, and importing the selected dangerous item icon into a corresponding mark on the target suitcase image.

3. The image intelligent marking system based on terahertz imaging according to claim 1 is characterized in that: The image processing module includes processing methods such as forward filter, pseudo colorization, binarization and morphological processing. The forward filter is used to remove noise and clutter signals in image data, the pseudo colorization is used to improve the visual effect of the image, the binarization is used to extract image contour information, and the morphological processing is used to segment and reconstruct the image.

4. The image intelligent marking system based on terahertz imaging according to claim 1, characterized in that: The algorithm for constructing the first function is as follows: Where Li represents the situation loss function of the target suitcase image and its corresponding pre-stored logo image corresponding to the i-th sub-block, α and τ are hyperparameters greater than 0, ω i is the mean grayscale value of the i-th sub-block in the target suitcase image, ω i0 is the mean grayscale value of the corresponding sub-block of the target suitcase image in the identification image, σ is the covariance of the corresponding sub-block of the target suitcase image and its corresponding identification image, is the square of the grayscale value variance of the i-th sub-block in the target suitcase image, is the square of the grayscale value variance of the corresponding sub-block in the identification image for the i-th sub-block in the target suitcase image.

5. The image intelligent marking system based on terahertz imaging according to claim 1 is characterized in that: In the marking recognition module, if prohibited items are carried, the location of the item is found and marked in the terahertz original image according to the first loss function and the terahertz original image, a marked terahertz original image is generated, and the terahertz original image and the marked terahertz original image are sent to the client.

6. The image intelligent marking system based on terahertz imaging according to claim 1, characterized in that: The monitoring terminal marks the location of the item on the pre-stored image of the suitcase according to the marked original terahertz image, and displays the original terahertz image and the marked original image to the monitoring personnel, so that the monitoring personnel can determine whether the central server has made any recognition errors; If there is a recognition error, the monitoring terminal changes the mark on the image of the prohibited items in the suitcase according to the instruction input by the monitoring personnel, and sends the changed image of the prohibited items in the suitcase to the client of the security personnel for display. The security personnel can determine whether the tested suitcase carries prohibited items based on the changed image of the prohibited items in the suitcase, thereby improving the accuracy of security inspection image recognition and avoiding missed detection and wrong detection.

7. An image intelligent marking method based on terahertz imaging, using the image intelligent marking system based on terahertz imaging according to any one of claims 1 to 6, characterized in that: The following steps are involved: S1: The client controls the terahertz scanning device to collect the terahertz original image of the contraband in the target suitcase; S2: Optimizing, extracting features, and performing image enhancement processing on the terahertz imaging data; Including methods such as denoising, optimization, blur kernel processing, and morphological analysis; S3: using multiple terahertz imaging images as training samples to train an image recognition network, artificially constructing label images corresponding to each terahertz imaging image as a training sample, and constructing a first loss function, which is a cross entropy loss function; S4: Identifying whether the target suitcase carries prohibited items through the terahertz image; S5: Marking the location of the object on the pre-stored contraband image according to the marked terahertz original image; S6: Displaying the terahertz original image and the original image with the mark to a monitoring person, so that the monitoring person can determine whether the central server has any recognition error.

8. The method for intelligent image marking based on terahertz imaging according to claim 1, characterized in that: In step S1 , the terahertz scanning device includes a terahertz source and a terahertz detector, wherein the terahertz source is used to transmit a terahertz signal, and the terahertz detector is used to receive a reflected terahertz wave signal.

Images