CN117132842A - Intelligent forbidden article detection method based on terahertz characteristics - Google Patents
Intelligent forbidden article detection method based on terahertz characteristics Download PDFInfo
- Publication number
- CN117132842A CN117132842A CN202311399219.9A CN202311399219A CN117132842A CN 117132842 A CN117132842 A CN 117132842A CN 202311399219 A CN202311399219 A CN 202311399219A CN 117132842 A CN117132842 A CN 117132842A
- Authority
- CN
- China
- Prior art keywords
- connected domain
- noise
- vibration noise
- image
- suspected
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 62
- 238000000034 method Methods 0.000 claims abstract description 33
- 238000010586 diagram Methods 0.000 claims abstract description 32
- 238000003062 neural network model Methods 0.000 claims abstract description 28
- 238000012216 screening Methods 0.000 claims description 61
- 230000011218 segmentation Effects 0.000 claims description 42
- 238000012549 training Methods 0.000 claims description 15
- 238000012545 processing Methods 0.000 claims description 11
- 238000004458 analytical method Methods 0.000 claims description 9
- 238000012937 correction Methods 0.000 claims description 7
- 230000008030 elimination Effects 0.000 claims description 7
- 238000003379 elimination reaction Methods 0.000 claims description 7
- 238000003491 array Methods 0.000 claims description 6
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000003708 edge detection Methods 0.000 claims description 3
- 238000009432 framing Methods 0.000 claims description 3
- 238000002372 labelling Methods 0.000 claims description 3
- 230000007613 environmental effect Effects 0.000 abstract description 2
- 238000007781 pre-processing Methods 0.000 abstract description 2
- 238000010183 spectrum analysis Methods 0.000 abstract description 2
- 238000007689 inspection Methods 0.000 description 4
- 101150031017 BIP2 gene Proteins 0.000 description 3
- 101000749871 Mus musculus Contactin-associated protein 1 Proteins 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 238000010438 heat treatment Methods 0.000 description 3
- LFQSCWFLJHTTHZ-UHFFFAOYSA-N Ethanol Chemical compound CCO LFQSCWFLJHTTHZ-UHFFFAOYSA-N 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000010521 absorption reaction Methods 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000035939 shock Effects 0.000 description 1
- 238000011410 subtraction method Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V8/00—Prospecting or detecting by optical means
- G01V8/10—Detecting, e.g. by using light barriers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/28—Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Geophysics (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The application belongs to the technical field of spectrum analysis, and provides an intelligent detection method for forbidden articles based on terahertz characteristics, which comprises the steps of obtaining a certain frame of image in a shot terahertz image, preprocessing the image, and primarily eliminating noise; calculating the preprocessed image, accurately distinguishing noise caused by vibration, and denoising the image according to a judging result; and identifying whether the target object is the forbidden article or not by adopting a neural network model for the denoised image, and finishing intelligent detection of the forbidden article. After the special denoising method provided by the application is adopted, vibration ghost noise generated by vibration is eliminated, the image quality of the terahertz characteristic diagram is greatly improved, the forbidden article detection method based on the terahertz characteristic has higher environmental robustness, and the final intelligent detection result can be more accurate.
Description
Technical Field
The application relates to the technical field of spectrum analysis, in particular to an intelligent forbidden article detection method based on terahertz characteristics.
Background
Terahertz waves refer to light with a specific frequency at the wavelength of light, and the light with the specific frequency has the characteristics of high penetrability, low energy, water absorption and the like. Based on the characteristics of the terahertz wave, the terahertz wave can penetrate through a shielding object in the current contraband detection, and the contraband objects such as liquid, explosive, metal products and the like in the detected objects are found, so that the terahertz wave is widely applied to the contraband detection.
In the current forbidden article detection system based on terahertz features, a great deal of background noise appears around an article to be detected possibly because of the vibration which is difficult to avoid when the detected article is transmitted by a conveyor belt, and the noise shape is a ghost generated when the object to be detected vibrates. The traditional denoising method is not ideal, and the image quality is reduced by retaining the noise, so that the final contraband detection result is affected.
Therefore, there is a need for a method to effectively remove the above background noise to improve the speed and quality of the contraband detection results.
Disclosure of Invention
In order to solve the technical problems, the application provides an intelligent forbidden article detection method based on terahertz characteristics, which is used for improving the speed and quality of forbidden article detection results.
The application provides an intelligent detection method for forbidden articles based on terahertz characteristics, which comprises the following steps:
collecting and acquiring a background fixed noise correction image and an original detection gray level image, carrying out background denoising on the original detection gray level image to obtain a primary denoising image, and continuously carrying out small hole elimination to obtain a binary image;
detecting the connected domain of the binary image, and then carrying out search analysis on the external contour of the connected domain obtained by detection to obtain a suspected vibration noise distance, a suspected vibration noise connected domain thickness and a search coordinate pair of the binary image;
based on the suspected vibration noise distance and the thickness of the suspected vibration noise connected domain, acquiring a suspected noise connected domain segmentation index of the binary image;
according to the suspected noise connected domain segmentation index, combining the search coordinate pair, performing first noise screening on the binary image, and segmenting a first noise screening connected domain graph from the binary image to obtain a corrected vibration noise distance and a corrected vibration noise connected domain thickness;
analyzing the corrected vibration noise distance and the corrected vibration noise connected domain thickness to obtain a unique characteristic value of a first noise screening connected domain graph, and grouping the first noise screening connected domain graph;
analyzing the unique characteristic value to obtain a noise judgment value;
according to the noise judgment value, based on the first noise screening connected domain diagram, performing second noise screening and processing on the binary image to obtain a vibration noise mask diagram; performing AND operation on the vibration noise mask image and the primary denoising image to obtain a secondary denoising image;
and identifying whether the target object is the forbidden article or not by adopting a neural network model for the secondary denoising image, and finishing intelligent detection of the forbidden article.
In some embodiments of the present application, pinhole elimination is performed to obtain a binary image, including:
obtaining a threshold value of the primary denoising image by adopting an Ojin threshold value method, and performing binary segmentation on the primary denoising image according to the threshold value to obtain a binary segmentation image;
carrying out Canny edge detection on the binary segmentation image to obtain a stripe information graph of the binary segmentation image;
detecting whether all the end-to-end connected stripes completely comprise the end-to-end connected stripes in the area surrounded by the stripes based on the stripe information graph, if so, marking a point inside the area surrounded by the internal stripes in the binary segmentation image, and changing the gray level of the data point where the internal stripes are positioned into the background gray level of the stripe information graph;
and in the marked binary segmentation image, adopting a flooding filling algorithm to eliminate all small holes by taking 255 as a filling value for each marked data point to obtain a binary image.
In some embodiments of the present application, detecting the connected domain of the binary image, and then performing search analysis on the external contour of the connected domain obtained by the detection to obtain a suspected vibration noise distance, a suspected vibration noise connected domain thickness and a search coordinate pair of the binary image, where the search comprises:
detecting the connected domain of the binary image to obtain a connected domain mark m;
selecting a data point from the outer contour of the connected domain, starting from the data point, and carrying out data point retrieval along the tangential line of the data point in the vertical outward direction until the edge of the picture is retrieved or the data point of the outer contour of another connected domain is touched;
if the picture edge is touched when the search is stopped, no operation is performed; if the external contour data point of another connected domain is touched during the stopping of the search, recording how many data points are searched in total in the searching process, recording as the distance d of the connected domain and recording the coordinates (x) 1 ,y 1 ) The method comprises the steps of carrying out a first treatment on the surface of the Then, searching is continued in the touched connected domain according to the searching direction until the outline of the searched connected domain is touched, how many data points are searched in the connected domain are recorded as the connected domain thickness s and the coordinates (x 2 ,y 2 ) The method comprises the steps of carrying out a first treatment on the surface of the Obtain the search coordinate pair am= (x) 1 ,y 1 ,x 2 ,y 2 ) The method comprises the steps of carrying out a first treatment on the surface of the The initial connected domain is marked with the number m a The end connected domain is marked as m b Obtaining the vibrationDistance d of dynamic noise ab Thickness s of vibration noise connected domain ab Search for coordinate pair AM ab ；
Traversing all data points on all connected outer contours in a clockwise direction, and obtaining n vibration noise distances d for the same ending connected domain ab According to d ab In the initial connected domain m a Numbering the images from the first to the last clockwise to obtain a suspected vibration noise distance d nab Pair s of the same theory ab Numbering to obtain the thickness s of the suspected vibration noise connected domain nab For AM ab Numbering to obtain search coordinate pair AM nab 。
In some embodiments of the present application, the suspected noise connected domain segmentation index is:
;
in nsm n A suspected noise connected domain segmentation index d representing a binary image nab Representing a suspected vibration noise distance, d, numbered n n-1ab Representing a suspected vibration noise distance, s, numbered n-1 nab The thickness s of the suspected vibration noise connected domain with the reference number n is represented n-1ab The thickness mu of the suspected vibration noise connected domain with the reference number of n-1 is shown dab Representing the suspected vibration noise distance d nab Mean, mu sab S representing thickness of suspected vibration noise connected domain nab And (5) an average value.
In some embodiments of the present application, according to the index of segmentation of the suspected noise connected domain, in combination with the search coordinate pair, performing a first noise screening on the binary image, and segmenting a first noise screening connected domain graph from the binary image, to obtain a corrected vibration noise distance and a corrected vibration noise connected domain thickness, including:
dividing the suspected noise connected domain into nsm indexes n Less than 0.1 and the adjacent elements are divided into a group, each group of the obtained beginning or ending labels n, and the coordinate pairs AM are searched correspondingly according to the labels n nab Two coordinates of (a) and corresponding search coordinate pairs AM nab The connecting line of two coordinates in the structure is used as a dividing line to connect the end connected domain m b Dividing;
each connected domain after segmentation has a group of suspected noise connected domain segmentation indexes nsm corresponding to the connected domain n Value of the division index nsm of the suspected noise connected domain n The connected domain with the value smaller than 0.1 is reserved, the gray value of each data point of other connected domains is set to be 0, and a first noise screening connected domain diagram is obtained;
renumbering connected domains in the first noise screening connected domain diagram, and the reference sign is marked as l t The initial connected domain is marked as l c The end connected domain is marked as l e Obtaining a corrected vibration noise distance d nce And correcting the thickness s of the vibration noise connected domain nce 。
In some embodiments of the present application, analyzing the corrected vibration noise distance and the corrected vibration noise connected domain thickness, obtaining a unique feature value of a first noise screening connected domain graph, and grouping the first noise screening connected domain graph, including:
distance d of the corrected vibration noise nce And the corrected vibration noise connected domain thickness s nce The subscript c of the pair is divided into a plurality of groups according to the subscript e, and the corrected vibration noise distance d in each group is obtained nce Mean mu dce The thickness s of the corrected vibration noise connected domain nce Mean mu sce ；
For average mu dce A minimum set of data, μ therein sce As a suspected vibration noise connected domain l e Thickness characteristic value W of (a);
for average mu dce Two minimum groups of data are obtained by connecting corresponding two initial connected domains c As a suspected vibration noise connected domain l e The two connected domains are numbered l in the order from left to right and from top to bottom according to the centroid position q And l p Obtaining a unique characteristic value W of the first noise screening connected domain diagram qep ；
The first noise screening connected domain diagram is processed according to W qep The connected domains included in the position information q, e, and p are divided into a group.
In some embodiments of the application, the noise determination value is: taking the sum of the variance of the u different thickness characteristic value W arrays and the variance of the u-1 different thickness characteristic value W arrays, wherein u is more than or equal to 2.
In some embodiments of the present application, according to the noise determination value, based on the first noise screening connected domain map, performing a second noise screening and processing on the binary image to obtain a vibration noise mask map, including:
noise judgment value snr in first noise screening connected domain diagram u Setting the gray value of each data point of the corresponding connected domain to be 0 or more than 0.1 to obtain a second noise screening connected domain diagram;
and setting the data values of all connected domains in the second noise screening connected domain graph to be 0, and setting the background data point to be 1 to obtain the vibration noise mask graph VBfix.
In some embodiments of the present application, for the secondary denoising image, a neural network model is used to identify whether a target object is an forbidden article, so as to complete intelligent detection of the forbidden article, including:
the forbidden article detection system based on terahertz features adopts a YOLO2 algorithm to obtain a neural network model;
training the neural network model to obtain a trained neural network model;
and inputting the secondary denoising image into the trained neural network model to finish intelligent detection of forbidden articles based on terahertz characteristics.
In some embodiments of the present application, training a neural network model to obtain a trained neural network model includes:
constructing a training set consisting of 5 dangerous goods and 1 conventional goods, respectively placing the training set into paper packages, and collecting 5000 pictures by continuously changing the size of the placed packages and the placing positions and states of the goods, wherein 600 dangerous goods are selected from the five dangerous goods, and 2000 conventional goods are selected from the 1 conventional goods;
and framing and labeling each picture, adopting a cross entropy function as a loss function, and adopting a Adam (Adaptive Moment Estimation) optimizer to train the neural network to obtain a trained neural network model.
As can be seen from the above embodiments, the intelligent detection method for the forbidden articles based on terahertz features provided by the embodiment of the application has the following beneficial effects:
aiming at the ghost-like noise generated by the vibration of an object in a terahertz image, the embodiment of the application measures the vibration noise distance and the vibration noise communication domain thickness through a specific algorithm according to the characteristics that the image shape of the ghost-like noise is parallel and similar to the edge contour of the object, and further calculates the communication domain segmentation index and the suspected noise object judgment value. The connected domain segmentation index can divide and overlap vibration ghost noise and object images which are connected, so that the risk that the vibration ghost noise and the object images are overlapped together and eliminated is avoided; the suspected noise article judgment value can distinguish article images which are easy to be confused with noise images, accurately judge vibration ghost noise images and finish accurate denoising of terahertz images. After the special denoising method provided by the application is adopted, vibration ghost noise generated by vibration is eliminated, the image quality of the terahertz characteristic diagram is greatly improved, the forbidden article detection method based on the terahertz characteristic has higher environmental robustness, and the final intelligent detection result can be more accurate.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application as claimed.
Drawings
In order to more clearly illustrate the technical solution of the present application, the drawings that are needed in the embodiments will be briefly described below, and it will be obvious to those skilled in the art that other drawings can be obtained from these drawings without inventive effort.
Fig. 1 is a basic flow diagram of an intelligent forbidden article detection method based on terahertz characteristics according to an embodiment of the present application.
Detailed Description
In order to make the technical solution of the present application better understood by those skilled in the art, the technical solution of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
The application aims to detect articles in security inspection by the physical characteristics of terahertz characteristics, and respectively detect forbidden articles in the terahertz characteristics, so that the security inspection speed and quality are increased, and the security is ensured.
The specific scene aimed by the application is as follows: according to the application, the terahertz images of the objects to be detected are analyzed aiming at scenes of express delivery, airports, stations and the like, which need to carry out rapid security inspection on the parcel baggage, the terahertz images are denoised aiming at special noise in the scenes, the quality of the terahertz images is improved, the identification of forbidden objects is completed based on a YOLO2 algorithm, the security inspection speed and quality are accelerated, and the security is ensured.
The intelligent detection method for the forbidden articles based on the terahertz characteristics provided by the embodiment is described in detail below with reference to the accompanying drawings.
Fig. 1 is a basic flow chart of an intelligent forbidden article detection method based on terahertz characteristics, which is provided by the embodiment of the application, and as shown in fig. 1, the method specifically comprises the following steps:
s100: and acquiring a background fixed noise correction image and an original detection gray level image, carrying out background denoising on the original detection gray level image to obtain a primary denoising image, and continuously carrying out pinhole elimination to obtain a binary image.
In the example scene of the application, the adopted terahertz characteristic detection camera is as follows: teraFAST-256 from TeraCase. The original image is a gray image, the background is white, and noise or the detected object is black.
Because the application adopts the conveyor belt to transport the articles aiming at the scene, the friction between the driving wheel and the conveyor belt can cause the local heating of the conveyor belt, and the conveyor belt can possibly generate unexpected terahertz waves due to the heating to influence the final image quality, and firstly, the noise is eliminated by adopting an image addition and subtraction method.
Because the structure of the object to be detected is complex, terahertz light can completely penetrate through the internal partial area of the object to be detected, so that small holes appear in the final image of the object to be detected, the denoising effect of special noise is affected, and therefore the small holes are required to be identified first and eliminated.
Therefore, the embodiment of the application firstly acquires the background fixed noise correction image and the original detection gray level image, carries out background denoising on the original detection gray level image to obtain a primary denoising image, and continuously carries out pinhole elimination to obtain a binary image.
Further, in some embodiments of the present application, the specific method of background denoising may be:
1. when the conveyor belt has no article to be detected at the beginning of operation, a gray image at the moment is obtained and named as a background fixed noise correction image BGfix;
2. when an object to be detected starts on a conveyor belt, a terahertz characteristic detection image is obtained, the terahertz characteristic detection image of any frame is named as an original detection gray level image OGP, and the original detection gray level image OGP is subtracted by a background fixed noise correction image BGfix to obtain a primary denoising image GP1.
Through the steps, the background fixed noise is subtracted, a primary denoising image GP1 is obtained, the intensity of terahertz waves generated by local heating of the conveyor belt is small, the image of the object to be detected is not affected, and the background noise generated by vibration is clearer.
Further, in some embodiments of the present application, pinhole elimination is performed to obtain a binary image, and the specific method includes the following steps:
1. obtaining a threshold value of the primary denoising image GP1 by adopting an Ojin threshold method, and performing binary segmentation on the primary denoising image GP1 according to the threshold value to obtain a binary segmentation image BIP;
2. carrying out Canny edge detection on the binary segmentation image BIP to obtain a stripe information map SMP of the binary segmentation image BIP;
3. detecting whether all the end-to-end connected stripes are completely contained in the area surrounded by the end-to-end connected stripes based on the stripe information map SMP, if so, marking a point inside the area surrounded by the internal stripes in the binary segmentation image BIP, and changing the gray level of the data point where the internal stripes are positioned into the background gray level of the stripe information map SMP in the stripe information map SMP;
4. in the marked binary segmentation image BIP, as the gray value of the data point of the article connected domain is 255, all small holes are eliminated by adopting a flooding filling algorithm with 255 as a filling value for each marked data point, and the binary image BIP2 with small holes removed is obtained.
Through the steps, the binary image BIP2 is obtained on the premise of not influencing the external contour information of the image, the hole image in the article is eliminated, and the image only contains the position and shape information with the detected article and vibration noise.
The embodiment of the application refers to noise around objects caused by vibration of a conveyor belt as vibration ghost noise, and constructs a vibration noise distance d based on the shape and position characteristics of the vibration ghost noise nab And the thickness s of the vibration noise connected domain nab Reflecting the shape and position characteristics of each connected domain; constructing a suspected connected domain segmentation index nsm according to the two indexes, and completing segmentation of vibration noise and the article connected domain; and constructing a suspected noise article judgment value snr, completing the distinction between the noise connected domain and the suspected noise article connected domain, and accurately positioning the noise connected domain. And using the resolved second noise screening connected domain diagram for denoising to obtain a second denoising image GP2, thereby completing the removal of the problematic noise.
The method specifically comprises the following steps:
s200: and detecting the connected domain of the binary image, and then carrying out search analysis on the external contour of the connected domain obtained by detection to obtain a suspected vibration noise distance, a suspected vibration noise connected domain thickness and a search coordinate pair of the binary image.
In some embodiments of the present application, a connected domain is detected on a binary image, and then a search analysis is performed on an external contour of the connected domain obtained by the detection, so as to obtain a suspected vibration noise distance, a suspected vibration noise connected domain thickness and a search coordinate pair of the binary image, including the following steps:
1. detecting connected domains of the binary image BIP2 to obtain connected domains, numbering each connected domain, marking the number as m, and performing the following operation on each connected domain;
2. selecting a data point from the outer contour of the connected domain, starting from the data point, and carrying out data point retrieval along the tangential line of the data point in the vertical outward direction until the edge of the picture is retrieved or the data point of the outer contour of another connected domain is touched;
3. if the picture edge is touched when the search is stopped, no operation is performed; if the external contour data point of another connected domain is touched during the stopping of the search, recording how many data points are searched in total in the searching process, recording as the distance d of the connected domain and recording the coordinates (x) 1 ,y 1 ) The method comprises the steps of carrying out a first treatment on the surface of the Then, searching is continued in the touched connected domain according to the searching direction until the outline of the searched connected domain is touched, how many data points are searched in the connected domain are recorded as the connected domain thickness s and the coordinates (x 2 ,y 2 ) The method comprises the steps of carrying out a first treatment on the surface of the Obtain the search coordinate pair am= (x) 1 ,y 1 ,x 2 ,y 2 ) The method comprises the steps of carrying out a first treatment on the surface of the The initial connected domain is marked with the number m a The end connected domain is marked as m b Obtaining the vibration noise distance d ab Thickness s of vibration noise connected domain ab Search for coordinate pair AM ab ；
4. Traversing all data points on all connected outer contours in a clockwise direction, and obtaining n vibration noise distances d for the same ending connected domain ab According to d ab In the initial connected domain m a Numbering the images from the first to the last clockwise to obtain a suspected vibration noise distance d nab Pair s of the same theory ab Numbering to obtain the thickness s of the suspected vibration noise connected domain nab For AM ab Numbering to obtain search coordinate pairsAM nab 。
Thus far, the suspected vibration noise distance d is obtained through the above processing nab Comprising the representation m a 、m b Contour similarity information between two connected domains; obtaining the thickness s of the suspected vibration noise connected domain nab Comprises m b Thickness information of different positions on the connected domain; retrieving the coordinate pair AM nab The position information of the two indexes is recorded.
S300: and obtaining a suspected noise connected domain segmentation index of the binary image based on the suspected vibration noise distance and the thickness of the suspected vibration noise connected domain.
Obtaining the suspected vibration noise distance d in step S300 nab For the vibration noise connected domain, because the shape of the vibration noise connected domain is ghost, the edge contour of the vibration noise connected domain is highly matched with the edge contour of an object. If m is b The representative connected domain is vibration noise, m a The communicating domain represented is an article or is associated with m b Vibration noise generated in the same batch, d nab Should be a constant fixed value. Also for the thickness s of the communication domain of suspected vibration noise nab Since the shape of the vibration noise is a residual shadow of vibration residual, the thickness of the vibration noise should be quite uniform, s nab And should also be a constant fixed value.
Based on the above analysis, if between n-1 and n, d nab Or s nab A drastic change occurs, representing vibration noise overlapping the image of other objects at n, at which time m needs to be applied at n b The communicating region is divided into two communicating regions.
The method for constructing the suspected noise connected domain segmentation index based on the method comprises the following steps:
;
in nsm n A suspected noise connected domain segmentation index d representing a binary image nab Representing a suspected vibration noise distance, d, numbered n n-1ab Representing a suspected vibration noise distance, s, numbered n-1 nab A suspected earthquake denoted by the reference number nDynamic noise connected domain thickness s n-1ab The thickness mu of the suspected vibration noise connected domain with the reference number of n-1 is shown dab Representing the suspected vibration noise distance d nab Mean, mu sab S representing thickness of suspected vibration noise connected domain nab The mean value, n.gtoreq.2, is to ensure that the subscript is correct, and all data is calculated without pointing to the data which does not exist.
If d nab Or s nab Stable nsm n A value close to 0; if d nab Or s nab Severe change, nsm n The value is larger, far from 0; empirical value nsm n If the number is greater than 0.1, it is determined that the connected domain needs to be divided.
S400: and according to the suspected noise connected domain segmentation index, combining the search coordinate pair, carrying out first noise screening on the binary image, and segmenting a first noise screening connected domain graph from the binary image to obtain a corrected vibration noise distance and a corrected vibration noise connected domain thickness.
According to the suspected noise connected domain segmentation index, combining with the search coordinate pair, carrying out first noise screening on the binary image, segmenting a first noise screening connected domain graph from the binary image, and obtaining a corrected vibration noise distance and a corrected vibration noise connected domain thickness, wherein the method comprises the following steps:
1. suspected noise connected domain segmentation index nsm n Is a one-dimensional vector matrix, suspected noise connected domain segmentation index nsm n The connected domains corresponding to elements smaller than 0.1 are regarded as a whole. Therefore, the suspected noise connected domain division index nsm n Less than 0.1 and the adjacent elements are divided into a group, each group of the obtained beginning or ending labels n, and the coordinate pairs AM are searched correspondingly according to the labels n nab Two coordinates of (a) and corresponding search coordinate pairs AM nab The connecting line of two coordinates in the pair is used as a dividing line, the data point value of the dividing line is set to be 0, and the end connected domain m b Dividing;
2. traversing all d nab The division of the suspected vibration noise connected domain and the article connected domain is completed, and each connected domain after division has a group of corresponding suspected noise connected domain division indexes nsm n Value to be suspected of noiseConnected domain division index nsm n And reserving the connected domain with the value smaller than 0.1, and setting the gray value of each data point of other connected domains to be 0 to obtain a first noise screening connected domain diagram NCP1.
3. Renumbering the connected domain in the first noise screening connected domain map NCP1, and the reference sign is marked as l t The initial connected domain is marked as l c The end connected domain is marked as l e Obtaining a corrected vibration noise distance d nce And correcting the thickness s of the vibration noise connected domain nce 。
So far, through the above processing, the suspected vibration noise communicating domain is separated from the article communicating domain, and a new corrected vibration noise distance d is obtained nce And correcting the thickness s of the vibration noise connected domain nce 。
S500: analyzing the corrected vibration noise distance and the corrected vibration noise connected domain thickness, obtaining the unique characteristic value of the first noise screening connected domain graph, and grouping the first noise screening connected domain graph.
Correction of vibration noise distance d nce And correcting the thickness s of the vibration noise connected domain nce In the connected domain l e Is a connected domain after preliminary deletion, wherein each suspected vibration noise connected domain I e Nsm of connected domain n Values of less than 0.1 are all very likely to be vibration noise connected domains. However, the special condition that the shape of the detected object is special and the self communicating domain is similar to the shape of the vibration noise communicating domain generated by the detected object can also occur, so that erroneous judgment is caused.
However, since vibration noise is caused by the fact that the object vibrates to generate the ghost on the photo, the thickness of the ghost generated by the object is usually far lower than that of the object itself, and the further away from the object, the thinner the ghost.
Based on the analysis, the corrected vibration noise distance and the corrected vibration noise connected domain thickness are analyzed, the unique characteristic value of the first noise screening connected domain diagram is obtained, and the first noise screening connected domain diagram is grouped. The method comprises the following steps:
1. first, the vibration noise distance d is corrected nce And correcting the thickness s of the vibration noise connected domain nce And (5) preprocessing. Due to each suspected vibration noise connected domain e Will correspond to multiple groups d nce Sum s nce Data, the application corrects the vibration noise distance d nce And correcting the thickness s of the vibration noise connected domain nce The same subscript c is divided into a plurality of groups according to the difference of subscripts e, and the corrected vibration noise distance d in each group is obtained nce Mean mu dce Correcting the thickness s of the vibration noise connected domain nce Mean mu sce ；
2. For average mu dce A minimum set of data, μ therein sce As a thickness characteristic value W of the first noise screening connected domain diagram;
3. for average mu dce Two minimum groups of data are obtained by connecting corresponding two initial connected domains c As a suspected vibration noise connected domain l e The two connected domains are numbered l in the order from left to right and from top to bottom according to the centroid position q And l p Obtaining a unique characteristic value W of the first noise screening connected domain diagram qep ；
4. Unique characteristic value W qep W represents the thickness characteristic value of the connected domain, and q, e, and p represent adjacent connected domains of the connected domain. Each suspected vibration noise connected domain e All have a corresponding q, e, p, where q represents its previous connected domain and p represents its subsequent connected domain, these associated connected domains being a set of consecutive shock artifacts. The first noise screening connected domain diagram is processed according to W qep The connected domains included in the position information q, e, and p are divided into a group.
So far, the unique characteristic value W of the first noise screening connected domain diagram is obtained through the above processing qep And grouping the first noise screening connected domain graphs.
S600: and analyzing the unique characteristic value to obtain a noise judgment value.
For any group of unique characteristic values W obtained in step S500 qep Arranging according to the size of the thickness characteristic values W from small to large, wherein u different thickness characteristic values W are taken from small to large to obtain an array W u Which is provided withWhere u is the number of eigenvalues taken. If a ribbon is to be confused with vibration noise, its thickness characteristic should be the largest and much larger than the thickness characteristic of vibration noise in the same group.
Based on the above analysis, a noise judgment value is constructed as follows: taking the sum of the variance of the u different thickness characteristic value W arrays and the variance of the u-1 different thickness characteristic value W arrays, wherein u is more than or equal to 2, namely:
;
wherein, snr u Noise judgment value sigma representing noise and suspected noise article u Representing the variance, sigma, of an array of u non-identical thickness eigenvalues, W u-1 The variance of the array of u-1 non-identical thickness eigenvalues W is represented.
If the u-th thickness feature value W represents an object confused with vibration noise, then σ u And sigma (sigma) u-1 Should be very large, representing a very dramatic change in the thickness characteristic W. If all of the u thickness feature values W are vibration noise, σ u And sigma (sigma) u-1 Should be very small, representing very slow changes in the thickness characteristic W. Sigma (sigma) u +σ u-1 The method is used for preventing different variances of thickness characteristic values caused by different transmission vibration intensity in terahertz characteristic detection images without frames, and is suitable for snr u Has influence on sigma u And sigma (sigma) u-1 And (3) carrying out normalization processing on the difference value of the two images.
S700: according to the noise judgment value, based on the first noise screening connected domain diagram, performing second noise screening and processing on the binary image to obtain a vibration noise mask diagram; and performing AND operation on the vibration noise mask image and the primary denoising image to obtain a secondary denoising image.
According to the noise judgment value, based on the first noise screening connected domain diagram, carrying out second noise screening and processing on the binary image to obtain a vibration noise mask diagram, and comprising the following steps:
1.snr u the empirical threshold for judgment is 0.1, when snr u And when the value is more than or equal to 0.1, judging that the corresponding connected domain including u after the characteristic value is sequenced in u is the object connected domain mixed with noise. Noise judgment value snr in first noise screening connected domain diagram NCP1 u Setting the gray value of each data point of the corresponding connected domain to be 0 or more than 0.1 to obtain a second noise screening connected domain map NCP2;
2. because the data value of the background data is 0, setting the data value of all connected domains in the NCP2 of the second noise screening connected domain to be 0 and setting the background data point to be 1, and obtaining a VBfix of the vibration noise mask;
3. and performing AND operation on the vibration noise mask image VBfix and the primary denoising image GP1 to obtain a secondary denoising image GP2.
Thus far, through the above processing, the image GP2 after the second denoising is obtained.
S800: and (3) identifying whether the target object is the forbidden article or not by adopting a neural network model for the secondary denoising image, and finishing intelligent detection of the forbidden article.
And identifying whether the target object is a forbidden article or not according to the secondary denoising image by adopting a neural network model, and finishing intelligent detection of the forbidden article, wherein the method comprises the following steps of:
1. the forbidden article detection system based on terahertz features adopts a YOLO2 algorithm to obtain a neural network model;
2. training the neural network model to obtain a trained neural network model;
3. and inputting the secondary denoising image into a trained neural network model to finish intelligent detection of forbidden articles based on terahertz characteristics.
Training the neural network model to obtain a trained neural network model, wherein the training comprises the following steps:
2.1 constructing a training set consisting of 5 types of dangerous goods and 1 type of conventional goods, wherein the 5 types of dangerous goods are respectively: the method comprises the steps of respectively putting training sets into paper packages of gun models, powdery objects, ethanol, cutters and combustible five dangerous objects, such as clothes, books and the like, and collecting 5000 pictures by continuously changing the sizes of the placed packages and the placing positions and states of the objects, wherein 600 dangerous objects are respectively arranged in the five dangerous objects, and 2000 conventional objects are arranged in the 1 class;
2.2 framing and labeling each picture, adopting a cross entropy function as a loss function, adopting a Adam (Adaptive Moment Estimation) optimizer, and training the neural network to obtain a trained neural network model. The training process is a well-known technique in the art and will not be described in detail.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments.
It is noted that unless specified and limited otherwise, relational terms such as "first" and "second", and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. The terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a circuit structure, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such article or apparatus. Without further limitation, the statement "comprises one … …" does not exclude that an additional identical element is present in an article or device that comprises the element. In addition, the term "and/or" as used herein includes any and all combinations of one or more of the associated listed items, and the symbol/label is used herein for convenience of description only.
Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. This application is intended to cover any variations, uses, or adaptations of the application following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice within the art to which the application pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the application being indicated by the following claims.
It is to be understood that the application is not limited to the precise arrangements and instrumentalities shown in the drawings, which have been described above, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the application is limited only by the appended claims.
Claims (10)
1. The intelligent detection method for the forbidden articles based on the terahertz characteristics is characterized by comprising the following steps of:
collecting and acquiring a background fixed noise correction image and an original detection gray level image, carrying out background denoising on the original detection gray level image to obtain a primary denoising image, and continuously carrying out small hole elimination to obtain a binary image;
detecting the connected domain of the binary image, and then carrying out search analysis on the external contour of the connected domain obtained by detection to obtain a suspected vibration noise distance, a suspected vibration noise connected domain thickness and a search coordinate pair of the binary image;
based on the suspected vibration noise distance and the thickness of the suspected vibration noise connected domain, acquiring a suspected noise connected domain segmentation index of the binary image;
according to the suspected noise connected domain segmentation index, combining the search coordinate pair, performing first noise screening on the binary image, and segmenting a first noise screening connected domain graph from the binary image to obtain a corrected vibration noise distance and a corrected vibration noise connected domain thickness;
analyzing the corrected vibration noise distance and the corrected vibration noise connected domain thickness to obtain a unique characteristic value of a first noise screening connected domain graph, and grouping the first noise screening connected domain graph;
analyzing the unique characteristic value to obtain a noise judgment value;
according to the noise judgment value, based on the first noise screening connected domain diagram, performing second noise screening and processing on the binary image to obtain a vibration noise mask diagram; performing AND operation on the vibration noise mask image and the primary denoising image to obtain a secondary denoising image;
and identifying whether the target object is the forbidden article or not by adopting a neural network model for the secondary denoising image, and finishing intelligent detection of the forbidden article.
2. The intelligent detection method for contraband based on terahertz features of claim 1, wherein performing pinhole elimination to obtain a binary image comprises:
obtaining a threshold value of the primary denoising image by adopting an Ojin threshold value method, and performing binary segmentation on the primary denoising image according to the threshold value to obtain a binary segmentation image;
carrying out Canny edge detection on the binary segmentation image to obtain a stripe information graph of the binary segmentation image;
detecting whether all the end-to-end connected stripes completely comprise the end-to-end connected stripes in the area surrounded by the stripes based on the stripe information graph, if so, marking a point inside the area surrounded by the internal stripes in the binary segmentation image, and changing the gray level of the data point where the internal stripes are positioned into the background gray level of the stripe information graph;
and in the marked binary segmentation image, adopting a flooding filling algorithm to eliminate all small holes by taking 255 as a filling value for each marked data point to obtain a binary image.
3. The intelligent detection method for contraband based on terahertz features according to claim 1, wherein the detecting the connected domain of the binary image, and then performing search analysis on the external contour of the connected domain obtained by the detection, to obtain a suspected vibration noise distance, a suspected vibration noise connected domain thickness and a search coordinate pair of the binary image, includes:
detecting the connected domain of the binary image to obtain a connected domain mark m;
selecting a data point from the outer contour of the connected domain, starting from the data point, and carrying out data point retrieval along the tangential line of the data point in the vertical outward direction until the edge of the picture is retrieved or the data point of the outer contour of another connected domain is touched;
if the picture edge is touched when the search is stopped, no operation is performed; if the external contour data point of another connected domain is touched during the stopping of the search, recording how many data points are searched in total in the searching process, recording as the distance d of the connected domain and recording the coordinates (x) 1 ,y 1 ) The method comprises the steps of carrying out a first treatment on the surface of the Then, searching is continued in the touched connected domain according to the searching direction until the outline of the searched connected domain is touched, how many data points are searched in the connected domain are recorded as the connected domain thickness s and the coordinates (x 2 ,y 2 ) The method comprises the steps of carrying out a first treatment on the surface of the Obtain the search coordinate pair am= (x) 1 ,y 1 ,x 2 ,y 2 ) The method comprises the steps of carrying out a first treatment on the surface of the The initial connected domain is marked with the number m a The end connected domain is marked as m b Obtaining the vibration noise distance d ab Thickness s of vibration noise connected domain ab Search for coordinate pair AM ab ；
Traversing all data points on all connected outer contours in a clockwise direction, and obtaining n vibration noise distances d for the same ending connected domain ab According to d ab In the initial connected domain m a Numbering the images from the first to the last clockwise to obtain a suspected vibration noise distance d nab Pair s of the same theory ab Numbering to obtain the thickness s of the suspected vibration noise connected domain nab For AM ab Numbering to obtain search coordinate pair AM nab 。
4. The intelligent detection method for contraband based on terahertz features of claim 1, wherein the suspected noise connected domain segmentation index is:
;
in nsm n Suspected noise connectivity representing binary imagesDomain division index d nab Representing a suspected vibration noise distance, d, numbered n n-1ab Representing a suspected vibration noise distance, s, numbered n-1 nab The thickness s of the suspected vibration noise connected domain with the reference number n is represented n-1ab The thickness mu of the suspected vibration noise connected domain with the reference number of n-1 is shown dab Representing the suspected vibration noise distance d nab Mean, mu sab S representing thickness of suspected vibration noise connected domain nab And (5) an average value.
5. The intelligent detection method for contraband based on terahertz features as in claim 1, wherein, according to the suspected noise connected domain segmentation index, in combination with the search coordinate pair, performing a first noise screening on the binary image, segmenting a first noise screening connected domain graph from the binary image, and obtaining a corrected vibration noise distance and a corrected vibration noise connected domain thickness, comprising:
dividing the suspected noise connected domain into nsm indexes n Less than 0.1 and the adjacent elements are divided into a group, each group of the obtained beginning or ending labels n, and the coordinate pairs AM are searched correspondingly according to the labels n nab Two coordinates of (a) and corresponding search coordinate pairs AM nab The connecting line of two coordinates in the structure is used as a dividing line to connect the end connected domain m b Dividing;
each connected domain after segmentation has a group of suspected noise connected domain segmentation indexes nsm corresponding to the connected domain n Value of the division index nsm of the suspected noise connected domain n The connected domain with the value smaller than 0.1 is reserved, the gray value of each data point of other connected domains is set to be 0, and a first noise screening connected domain diagram is obtained;
renumbering connected domains in the first noise screening connected domain diagram, and the reference sign is marked as l t The initial connected domain is marked as l c The end connected domain is marked as l e Obtaining a corrected vibration noise distance d nce And correcting the thickness s of the vibration noise connected domain nce 。
6. The intelligent detection method for contraband based on terahertz features of claim 1, wherein analyzing the corrected vibration noise distance and the corrected vibration noise connected domain thickness to obtain unique feature values of a first-time noise screening connected domain graph and grouping the first-time noise screening connected domain graph includes:
distance d of the corrected vibration noise nce And the corrected vibration noise connected domain thickness s nce The subscript c of the pair is divided into a plurality of groups according to the subscript e, and the corrected vibration noise distance d in each group is obtained nce Mean mu dce The thickness s of the corrected vibration noise connected domain nce Mean mu sce ；
For average mu dce A minimum set of data, μ therein sce As a suspected vibration noise connected domain l e Thickness characteristic value W of (a);
for average mu dce Two minimum groups of data are obtained by connecting corresponding two initial connected domains c As a suspected vibration noise connected domain l e The two connected domains are numbered l in the order from left to right and from top to bottom according to the centroid position q And l p Obtaining a unique characteristic value W of the first noise screening connected domain diagram qep ；
The first noise screening connected domain diagram is processed according to W qep The connected domains included in the position information q, e, and p are divided into a group.
7. The intelligent detection method for contraband based on terahertz characteristics according to claim 1, wherein the noise judgment value is: taking the sum of the variance of the u different thickness characteristic value W arrays and the variance of the u-1 different thickness characteristic value W arrays, wherein u is more than or equal to 2.
8. The intelligent detection method for contraband based on terahertz features of claim 1, wherein the performing, based on the first noise screening connected domain graph according to the noise determination value, a second noise screening and processing on the binary image to obtain a vibration noise mask graph includes:
noise judgment value snr in first noise screening connected domain diagram u Setting the gray value of each data point of the corresponding connected domain to be 0 or more than 0.1 to obtain a second noise screening connected domain diagram;
and setting the data values of all connected domains in the second noise screening connected domain graph to be 0, and setting the background data point to be 1 to obtain the vibration noise mask graph VBfix.
9. The intelligent detection method for contraband based on terahertz characteristics according to claim 1, wherein for the secondary denoising image, a neural network model is adopted to identify whether a target object is a contraband, and the intelligent detection for the contraband is completed, comprising:
the forbidden article detection system based on terahertz features adopts a YOLO2 algorithm to obtain a neural network model;
training the neural network model to obtain a trained neural network model;
and inputting the secondary denoising image into the trained neural network model to finish intelligent detection of forbidden articles based on terahertz characteristics.
10. The method for intelligently detecting contraband based on terahertz features according to claim 9, wherein training the neural network model to obtain a trained neural network model comprises:
constructing a training set consisting of 5 dangerous goods and 1 conventional goods, respectively placing the training set into paper packages, and collecting 5000 pictures by continuously changing the size of the placed packages and the placing positions and states of the goods, wherein 600 dangerous goods are selected from the five dangerous goods, and 2000 conventional goods are selected from the 1 conventional goods;
and framing and labeling each picture, adopting a cross entropy function as a loss function, and adopting an Adam optimizer to train the neural network to obtain a trained neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311399219.9A CN117132842B (en) | 2023-10-26 | 2023-10-26 | Intelligent forbidden article detection method based on terahertz characteristics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311399219.9A CN117132842B (en) | 2023-10-26 | 2023-10-26 | Intelligent forbidden article detection method based on terahertz characteristics |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117132842A true CN117132842A (en) | 2023-11-28 |
| CN117132842B CN117132842B (en) | 2024-01-23 |
Family
ID=88858586
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311399219.9A Active CN117132842B (en) | 2023-10-26 | 2023-10-26 | Intelligent forbidden article detection method based on terahertz characteristics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117132842B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117930375A (en) * | 2024-03-22 | 2024-04-26 | 国擎(山东)信息科技有限公司 | Multi-dimensional detection technology fused channel type terahertz human body security inspection system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108764358A (en) * | 2018-05-31 | 2018-11-06 | 广东工业大学 | A kind of Terahertz image-recognizing method, device, equipment and readable storage medium storing program for executing |
| CN110544227A (en) * | 2018-05-29 | 2019-12-06 | 中国科学院电子学研究所 | Passive terahertz human body security inspection image target detection method |
| CN115063315A (en) * | 2022-06-22 | 2022-09-16 | 西安电子科技大学 | Terahertz image preprocessing method based on sliding division and small-area noise filling |
| CN115082339A (en) * | 2022-06-22 | 2022-09-20 | 西安电子科技大学 | Terahertz image denoising method based on connected domain adaptive partitioning |
| CN115601697A (en) * | 2022-09-30 | 2023-01-13 | 山东省科学院自动化研究所（Cn） | Pretreatment method for terahertz security inspection image |
-
2023
- 2023-10-26 CN CN202311399219.9A patent/CN117132842B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110544227A (en) * | 2018-05-29 | 2019-12-06 | 中国科学院电子学研究所 | Passive terahertz human body security inspection image target detection method |
| CN108764358A (en) * | 2018-05-31 | 2018-11-06 | 广东工业大学 | A kind of Terahertz image-recognizing method, device, equipment and readable storage medium storing program for executing |
| CN115063315A (en) * | 2022-06-22 | 2022-09-16 | 西安电子科技大学 | Terahertz image preprocessing method based on sliding division and small-area noise filling |
| CN115082339A (en) * | 2022-06-22 | 2022-09-20 | 西安电子科技大学 | Terahertz image denoising method based on connected domain adaptive partitioning |
| CN115601697A (en) * | 2022-09-30 | 2023-01-13 | 山东省科学院自动化研究所（Cn） | Pretreatment method for terahertz security inspection image |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117930375A (en) * | 2024-03-22 | 2024-04-26 | 国擎(山东)信息科技有限公司 | Multi-dimensional detection technology fused channel type terahertz human body security inspection system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117132842B (en) | 2024-01-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110363047B (en) | Face recognition method and device, electronic equipment and storage medium | |
| CN117132842B (en) | Intelligent forbidden article detection method based on terahertz characteristics | |
| CN108710910B (en) | Target identification method and system based on convolutional neural network | |
| CN102254319B (en) | Method for carrying out change detection on multi-level segmented remote sensing image | |
| US8340420B2 (en) | Method for recognizing objects in images | |
| US11392800B2 (en) | Computer vision systems and methods for blind localization of image forgery | |
| CN111915704A (en) | Apple hierarchical identification method based on deep learning | |
| CN107103317A (en) | Fuzzy license plate image recognition algorithm based on image co-registration and blind deconvolution | |
| CN115082419A (en) | Blow-molded luggage production defect detection method | |
| CN108830832A (en) | A kind of plastic barrel surface defects detection algorithm based on machine vision | |
| KR20180070472A (en) | Image matching method and device | |
| CN108109163A (en) | A kind of moving target detecting method for video of taking photo by plane | |
| CN107194393B (en) | Method and device for detecting temporary license plate | |
| CN111126393A (en) | Vehicle appearance refitting judgment method and device, computer equipment and storage medium | |
| CN103413149B (en) | Method for detecting and identifying static target in complicated background | |
| CN109117746A (en) | Hand detection method and machine readable storage medium | |
| CN117252882A (en) | Cylinder head quality detection method and system | |
| CN105138992A (en) | Coastline detection method based on regional active outline model | |
| CN111444773B (en) | Image-based multi-target segmentation identification method and system | |
| CN113780110A (en) | Method and device for detecting weak and small targets in image sequence in real time | |
| CN110245600A (en) | Adaptively originate quick stroke width unmanned plane Approach for road detection | |
| CN114092682A (en) | Small hardware fitting defect detection algorithm based on machine learning | |
| CN118153607A (en) | One-dimensional code positioning method based on machine vision | |
| CN110889817A (en) | Image fusion quality evaluation method and device | |
| CN114037650A (en) | Ground target visible light damage image processing method for change detection and target detection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |