CN115601697A - A preprocessing method for terahertz security images - Google Patents

Abstract

The invention discloses a pretreatment method and system for a terahertz security inspection image, electronic equipment and a computer readable storage medium, and belongs to the technical field of terahertz waveband digital image analysis. The method comprises the steps of obtaining a terahertz security inspection image not containing a sample and a terahertz security inspection image containing the sample; performing histogram analysis on the terahertz security inspection image without the sample to acquire a gray value with the maximum number of corresponding pixels; carrying out corrosion operation on the terahertz security inspection image containing the sample, and carrying out binarization gray scale transformation on the processed image by taking the gray scale value with the maximum number of corresponding pixels as a threshold value; performing connected domain analysis on the image subjected to the binary gray processing to obtain a boundary of a sample, and performing burr trimming on the boundary; and acquiring a corrected image according to the image after the burr trimming. The interference of packages such as paper boxes, envelopes and the like is eliminated, and researchers can conveniently identify suspicious dangerous goods in the paper boxes and the envelopes visually or through an image recognition algorithm.

Description

A preprocessing method for terahertz security images

technical field

The present application relates to the technical field of terahertz band number image analysis, in particular to a pre-processing method for terahertz security images.

Background technique

The statements in this section merely mention the background art related to this application, and do not necessarily constitute the prior art.

Paper shell packaging materials have high permeability to terahertz waves, metals have strong reflection characteristics on terahertz waves, polar liquids such as water have strong absorption characteristics on terahertz waves, and irregular interfaces and curved surfaces have a relatively strong scattering effect on terahertz waves. Therefore, the terahertz continuous wave imaging system has a strong detection ability for metals covered by paper shells and polar liquids in bottles, and is sensitive to the boundaries of various objects. It is initially applied to the inspection of dangerous goods in packages and envelopes.

The development direction of terahertz security inspection is intelligence. Based on terahertz image features and advanced artificial intelligence algorithms, it is not only necessary to realize the positioning of internal dangerous goods, but also to complete the identification of dangerous goods. However, at this stage, the image recognition algorithm in the visible light band is not effective in the terahertz band, because the terahertz wavelength is longer and the diffraction resolution limit is larger.

In addition, the power of the terahertz source is low, the number of pixels in the imaging device is small, and the spatial scale of the pixels is large, all of which make it difficult to identify terahertz continuous wave images.

The terahertz security images obtained in the security inspection based on terahertz continuous wave imaging, such as cartons and envelopes, have unique characteristics: potential targets are located inside cartons and envelopes, although the packaging materials themselves have weak absorption of terahertz waves, cartons, envelopes The edge has a strong effect on the scattering of terahertz waves. Therefore, the images of the edge of the carton and the potential targets inside are prone to adhesion, which is extremely unfavorable for researchers to identify suspicious and dangerous items in cartons and envelopes visually or later through image recognition algorithms; When using an image classifier, testers are required to keep a certain distance between the target and the edges of cartons and envelopes during the image training process.

Contents of the invention

The inventor found that an ideal method for terahertz security inspection image recognition is: first detect the complete boundaries of cartons and envelopes, locate the cartons and envelopes, then process them through algorithms, and finally identify the cartons through visual or image recognition algorithms , an image of the contents of the envelope. However, few researchers have proposed targeted terahertz security image processing methods based on this idea, which limits the accuracy of dangerous goods image recognition.

In order to solve the deficiencies of the prior art, this application provides a pre-processing method, system, electronic equipment and computer-readable storage medium for terahertz security images, which are used to eliminate the interference of packaging such as cartons and envelopes, which is convenient for researchers Identify suspicious and dangerous items in cartons and envelopes visually or through image recognition algorithms.

In the first aspect, the present application provides a preprocessing method for terahertz security images;

A preprocessing method for terahertz security images, including:

Obtaining a terahertz security check image that does not contain a sample and a terahertz security check image that contains a sample; wherein the sample is a carton or envelope that may contain dangerous goods;

Perform histogram analysis on the terahertz security inspection image that does not contain samples, and obtain the gray value with the largest number of corresponding pixels;

Corrosion operation is performed on the terahertz security inspection image containing the sample, and the gray value corresponding to the largest number of pixels is used as the threshold value, and the processed image is subjected to binary gray scale transformation;

Perform connected domain analysis on the image after binary grayscale processing, obtain the boundary of the sample, and perform burr trimming on the boundary;

According to the burr-trimmed image, a corrected image is obtained.

In the second aspect, the present application provides a pre-processing system for terahertz security images;

A pre-processing system for terahertz security images, including:

The image acquisition module is configured to: acquire a terahertz security check image that does not contain a sample and a terahertz security check image that contains a sample; wherein the sample is a carton or envelope that may contain dangerous goods;

The background image analysis module is configured to: perform histogram analysis on the terahertz security inspection image that does not contain samples, and obtain the gray value corresponding to the largest number of pixels;

The noise reduction module is configured to: perform a corrosion operation on the terahertz security inspection image containing the sample, use the gray value corresponding to the largest number of pixels as the threshold value, and perform binary gray-scale transformation on the processed image;

The burr pruning module is configured to: perform connected domain analysis on the binary grayscale processed image, obtain the boundary of the sample, and perform burr pruning on the boundary;

The corrected image acquisition module is configured to: acquire the corrected image according to the image trimmed by the burr.

In a third aspect, the present application provides an electronic device;

An electronic device includes a memory, a processor, and computer instructions stored in the memory and run on the processor. When the computer instructions are run by the processor, the steps of the above-mentioned preprocessing method for terahertz security images are completed.

In a fourth aspect, the present application provides a computer-readable storage medium;

A computer-readable storage medium is used for storing computer instructions, and when the computer instructions are executed by a processor, the steps of the above-mentioned preprocessing method for terahertz security images are completed.

Compared with prior art, the beneficial effect of the present application is:

1. Based on the characteristics of terahertz security inspection images containing samples and terahertz security inspection images not containing samples, this application processes the terahertz security inspection images containing samples to eliminate the interference of packaging boundaries such as cartons or envelopes, which is convenient for researchers to see Identify suspicious dangerous goods in the package visually or through image recognition algorithms, improving the accuracy of dangerous goods image recognition;

2. This application marks the boundary of the connected domain during the burr pruning process, which is convenient for combining with other edge enhancement algorithms; no additional gradient calculation is required: for binary images, the gradient on the left side of the search direction is always equal to -255 ; When the search fails in the original search direction, the gradient of the pixel toward the direction of travel is -255. In addition, the gradient of the gray value in any direction elsewhere in the image is 0, which significantly simplifies the calculation of the gradient.

Description of drawings

The accompanying drawings constituting a part of the present application are used to provide further understanding of the present application, and the schematic embodiments and descriptions of the present application are used to explain the present application, and do not constitute improper limitations to the present application.

FIG. 1 is a schematic flow chart of a preprocessing method for terahertz security images provided in the embodiment of the present application;

Fig. 2 is a schematic diagram of the corrosion operation on the terahertz security image containing the sample provided by the embodiment of the present application; wherein, (a) is the image before corrosion, (b) is the structural element, and (c) is the image after corrosion;

Fig. 3 is a schematic diagram of connected domains of a terahertz image including samples provided by the embodiment of the present application;

4 is a schematic diagram of comparison before and after burr pruning provided by the embodiment of the present application; wherein, (a) is a schematic diagram of an initial outer boundary queue of a connected domain, and (b) is a schematic diagram of an outer boundary queue after burr pruning;

FIG. 5 is a schematic flow diagram of the generation of the initial outer boundary queue of the connected domain provided by the embodiment of the present application;

FIG. 6 is a schematic flow diagram of queue adjustment outside the connected domain provided by the embodiment of the present application;

Fig. 7 is a schematic diagram of boundary markers in a connected domain provided by the embodiment of the present application; wherein (a) is a schematic diagram of a connected domain, (b) is a statistical table for calculating key values, and (c) is a schematic flow chart for judging whether it is a real empty connected domain.

detailed description

It should be pointed out that the following detailed description is exemplary and is intended to provide further explanation to the present application. Unless defined otherwise, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

It should be noted that the terminology used here is only for describing specific implementations, and is not intended to limit the exemplary implementations according to the present application. As used herein, unless the context clearly dictates otherwise, the singular is intended to include the plural, and it should also be understood that the terms "comprising" and "having" and any variations thereof are intended to cover a non-exclusive Comprising, for example, a process, method, system, product, or device comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include steps or units not explicitly listed or for these processes, methods, Other steps or units inherent in a product or equipment.

In the case of no conflict, the embodiments and the features in the embodiments of the present invention can be combined with each other.

Embodiment one

In the existing technology, the terahertz images obtained during security inspection based on terahertz continuous wave imaging have their unique characteristics. The images of the edges of envelopes or carton packaging and the images of potential targets inside are prone to adhesion. It is extremely disadvantageous to identify dangerous goods; therefore, this application provides a pre-processing method for terahertz security images, which eliminates the interference of packaging such as cartons and envelopes, so that researchers can identify dangerous goods and improve the accuracy of identification.

A preprocessing method for terahertz security images, including:

Obtain terahertz security inspection images that do not contain samples and terahertz security inspection images that contain samples; where the samples are cartons or envelopes that may contain dangerous goods;

Perform histogram analysis on the terahertz security inspection image that does not contain samples, and obtain the gray value with the largest number of corresponding pixels;

Corrosion operation is performed on the terahertz security inspection image containing the sample, and the gray value corresponding to the largest number of pixels is used as the threshold value, and the processed image is subjected to binary gray scale transformation;

Perform connected domain analysis on the image after binary grayscale processing, obtain the boundary of the sample, and perform burr trimming on the boundary;

According to the burr-trimmed image, a corrected image is obtained.

Further, the connected domain analysis is performed on the binary grayscale processed image, the boundary of the sample is obtained, and the burr trimming of the boundary includes:

Perform connected domain analysis on the image after binary grayscale processing, obtain the outer boundary of the sample, and perform burr trimming on the outer boundary;

Detect whether there is an inner boundary in the connected domain;

If there is an inner boundary in the connected domain, burr pruning is performed on the inner boundary.

Preferably, the connected domain analysis is carried out to the image after binary grayscale processing, and the specific steps of obtaining the outer boundary of the sample are:

Based on the image after binary grayscale processing, the connected domain with a grayscale value of 255 is obtained, the abnormal region is eliminated, and the outer boundary of the sample is obtained.

Preferably, the specific steps for performing burr trimming on the outer boundary are:

Mark the initial outer boundary queue of the connected domain;

Based on the initial outer boundary queue, according to the queue index, find the pixel coordinates corresponding to the queue index; if the pixel coordinates corresponding to different queue indexes are the same, delete the index value in the initial outer boundary queue and delete the pixel coordinates between the two queue indexes ;

Traverse and search the adjusted outer boundary queue until the entire queue is processed.

Preferably, the specific steps of marking the initial outer boundary queue of the connected domain are:

Find the pixel with the smallest x-coordinate or y-coordinate in the connected domain, and set the pixel coordinate as the first of the initial outer boundary queue;

Start the search direction to the right, and search in the order of left, front, right, and back relative to the search direction; if the searched pixel is in the connected domain, add the pixel coordinates to the initial outer boundary queue, and update the search at the same time direction;

If the added pixel coordinates are the same as the starting point coordinates, the search is terminated. Further, the specific steps for performing binary grayscale transformation on the processed image are as follows:

Adjust the gray value of all pixels greater than or equal to the threshold to 0, and adjust the gray value of pixels whose gray value is lower than the threshold to 255.

Further, according to the burr-trimmed image, the specific steps for obtaining the corrected image are:

Restoring the pixel gray value corresponding to each coordinate in the connected domain after burr pruning to the corresponding pixel gray value in the terahertz security image containing the sample;

Adjust the other pixels in the glitch-trimmed image to 255 to obtain the corrected image.

Next, a preprocessing method for terahertz security images disclosed in this embodiment will be described in detail with reference to FIGS. 1-7 .

This embodiment provides a preprocessing method for terahertz security images, the preprocessing method for terahertz security images includes:

S1. Obtain a terahertz security inspection image that does not contain a sample and a terahertz security inspection image that contains a sample; wherein, the terahertz security inspection image that does not contain a sample and the terahertz security inspection image that contains a sample are collected by a continuous wave terahertz security inspection system, both of which are terahertz security inspection images. Hertz band grayscale image; the parameters of the external equipment remain unchanged during image acquisition: the first acquisition obtains a terahertz security image that does not contain samples (that is, a blank background image that does not contain samples, which is recorded as a gray value matrix P0), and the second The terahertz security inspection picture containing the sample is collected for the first time (denoted as the gray value matrix P1); the sample can be a carton or envelope containing one or more dangerous goods, and the dangerous goods are dangerous or suspected to be chemically analyzed items.

In this example, the sample is a carton containing one hazardous product.

S2. Perform a histogram analysis on P0, obtain the gray value corresponding to the largest number of pixels and record the gray value as g, which is the mode of the background gray value.

S3. Perform a morphological corrosion operation on P1, and the structure operator is as follows:

As shown in Figure 2, in order to increase readability, the light gray indicates that the gray value is 255, and the pixels are divided by white thin lines. Figure 2(a) shows the image before corrosion, and Figure 2(b) shows the structural elements. Figure 2(c) shows the image after etching.

In this embodiment, assuming that the center of the structural element coincides with a light gray color block in Figure 2(a), if all the color blocks of the structural element are covered by the light gray area in Figure 2(a), the overlapping color The block is preserved; the pixels in the original image are traversed using the structural element as a template, and the image shown in Figure 2(c) is obtained after processing.

Note that the image processed by the corrosion operation is P2, and g is the threshold value, and the processed image is subjected to binary grayscale transformation, and the grayscale values of all pixels greater than or equal to g are adjusted to 0 ("logic 0"), and the The gray value of the pixel whose gray value is lower than g is adjusted to 255 ("logic 1"), and the image P3 is obtained.

S4. Perform connected domain analysis on the image after binary grayscale processing, obtain the boundary of the sample, and perform burr trimming on the boundary;

After the image is processed and binarized by the morphological method, the edge of the image may appear uneven, which is mainly caused by the unsatisfactory morphological template and the complex scattering of terahertz waves at the edge of the object. The purpose of trimming is to eliminate the "burrs" left by the edges. The burrs defined here have a common feature that the minimum width is equal to 1 pixel.

Specific steps include:

S401. Based on the binary grayscale processed image, obtain a connected domain with a grayscale value of 255, remove abnormal regions, and obtain the outer boundary of the sample; specifically, perform connected domain analysis on P3, and find a connected domain with logic 1, From the result, remove the area whose size is less than or equal to 4, and the area whose size is less than or equal to 2 pixels in the x direction or y direction, and mark the remaining connected domain with a gray value of 255 as a valid target, and the corresponding pixel sets are numbered A ₁ , A ₂ , A ₃ . . . An, define other regions with gray value 0 as the background, and record the set of all pixel coordinates as B.

In this embodiment, as shown in FIG. 3 , a connected domain is a set of light gray pixel coordinates and the minimum distance between any element in the set and other elements is equal to 1 pixel. The upper left, upper right, lower left, and lower right connected domains are respectively marked as A ₁ , A ₂ , A ₃ , and A ₄ . Since the size of A ₃ is equal to 4, it is eliminated, and the pixel in the y direction of A ₄ is equal to 1.

S402. Mark the initial outer boundary queue of the connected domain; specifically, find the pixel with the smallest x-coordinate of a certain connected domain (if there are more than one pixel with the smallest x-coordinate, then find the pixel with the smallest y-coordinate among these points), and record its coordinates (that is, Starting point coordinates) and added to the first position of the initial outer boundary queue; the initial search direction is to the right, and the relative search direction is searched in the order of priority (descending order) of left, front, right, and back, and the pixel coordinate step size of each search is 1. If the searched pixel is in the connected domain, add the coordinate value of the pixel to the queue, and update the search direction at the same time; check whether it is the same as the starting point coordinate every time a new coordinate is added, if the starting point coordinate appears again in the queue, then The search is terminated and the queue length does not change any more.

In this embodiment, as shown in Fig. 4(a) and Fig. 5, taking A1 as _an example, the outer boundary of A1 is _first marked, and the first position of each pixel in the queue is marked inside the pixel, and the queue length for 42.

S403. Based on the initial outer boundary queue, according to the queue index, search for the pixel coordinates corresponding to the queue index; if the pixel coordinates corresponding to different queue indexes are the same, delete the index value in the initial outer boundary queue and delete the pixel coordinates between the two queue indexes Pixel coordinates; specifically, starting from the xth coordinate of the queue to search for coordinate values in the interval from x to n0, n is the total length of the current queue; if a coordinate with an index value of x is found at a position with an index value of x+Δx , then delete the part whose index value is from x+1 to x+Δx from the queue, reconstitute a new queue, and record its length as n1; the x+1th to x+Δx-1 coordinates of the original queue are included in the set B _n . Continue searching from the next coordinate value of the current search coordinate value in n1 until the coordinate value has been searched in the entire queue. The new queue obtained by the kth deletion is nk, nk≤n; the initial value of x is 2, and the index value of the search coordinates is added to the current queue after each search to the end of the queue, and the above search process is repeated until the queue The coordinate value at the end of , is used as the search object and the search is completed. Assume that the currently processed connected domain is A _n , and B _n includes all pixels to be deleted after all searches are completed.

In this embodiment, as shown in Fig. 4(b) and Fig. 6, the coordinates of the second pixel are searched backward from the queue index 2, and it is found that the coordinates do not appear in the queue; then search backward from the index 3 The coordinates of 3 pixels were not found either. By analogy, search backward from index 9, and find that the coordinates corresponding to index 15 are the same as the coordinates corresponding to index 9, then delete the 10th, 11th, 12th, 13th, 14th, 15th part of the current queue index value, and put the index The coordinates corresponding to values 10, 11, 12, 13, and 14 (corresponding to three pixels) are added to the set B ₁ , and then the adjusted new queue is searched from the end of the deleted interval until the entire queue is processed. According to Figure 4, the length of the initial queue has been adjusted 5 times in total, that is, the length of the queue after the last adjustment is n5=28, and the final result is shown in Figure 4(b), with a total of 28 pixels on the outer boundary. B ₁ in the figure contains 7 different coordinates, and their first appearance positions in the first queue are 10, 11, 12, 25, 30, 33, and 37 respectively.

S404. Detect whether there is an inner boundary in the above connected domain; specifically, remember that the smallest x-coordinate of a certain connected domain is x _min , the largest x-coordinate is x _max , the smallest y-coordinate is y _min , and the largest y-coordinate is y _max . Start searching from y=y _min , assuming that the set of x-coordinates of points from y=y _min in the connected domain is R ₁ , arrange the elements in R ₁ in ascending order, and find out the integer discontinuity of R ₁ according to the boundary of the x-coordinates of this column and form several sets H1, H2, H3... containing these x-coordinate values, which are called sets in the state of y=1, and the minimum difference between elements in each set is 1. Add 1 to the search coordinate (that is, execute y=y _min +1), repeat the above process until the search is completed y=y _max , and the search is completed. Initialize i=1, and based on the x coordinates of each element in the set, find the intersection of the sets named at the beginning of H in the state of y=i and y=i+1. If the same x element exists in the two sets, the two sets will be merged. After merging, the set is named with the smaller number, that is, Hc=Ha∪Hb, c=min{a,b}; if the x coordinates of the elements in the two sets do not have the same value, the set is added to y=i in the state of y=i +1 status. Search for y=y _min +2, y=y _min +3...until y=y _max search is completed, and the merging of the first-level set is completed with the value of y increasing by 1 (maybe merging more than once), and finally the connected domain of the suspected hole is obtained. A connected domain corresponds to a set.

If a connected domain suspected to be a hole is detected, then step S405 is performed, and if there is no connected domain, then step S5 is performed.

In this embodiment, the connected domain is shown in FIG. 4( b ), there is no integer discontinuity in the x-coordinate, and no connected domain suspected to be a hole is detected, then step S5 is executed.

Exemplarily, in some embodiments, the connected domain is as shown in FIG. 7( a ), a target connected domain (light gray), the upper row of numbers indicates the y coordinate, and the left column of numbers indicates the x coordinate. When scanning, the range is determined according to the distribution of the target connected domain y. Obviously, 2≤y≤11 requires 10 scans. First scan a column of y=2, record the x coordinates involved in the column of the target connected domain, then y=y+1, and so on to get the second row of the table in Figure 7(b); according to the column x Find out the discontinuous part of the Ri integer and form several sets, the minimum difference between the elements in each set is 1, for example, y=2 Find {4,5,6,7,8,9,10,11,12 } and {4,6,7,8,9,12} to get {5}, {10,11}, and so on, each column is searched to get the third row of the table in Figure 7(b); then according to The distribution of these disconnected parts builds sets H1, H2...H12, and the sets generated when y=i is scanned are shown in the fourth row of the table in Figure 7(b). The numbers in curly brackets indicate the order in which the sets are generated (ie The label of the H set); the pixel marked with the white number i in Figure 7(a) is the element in the set Hi. According to whether there is an intersection of the x coordinates corresponding to the pixels in the set, the set in the state of y=i and the set generated when y=i+1 are conditionally merged in pairs to obtain the set in the state of y=i+1: such as y If the x-coordinates of each element in a set in the =i state and a certain set generated by y=i+1 have no common value, then the set continues to exist in the state of y=i+1. By analogy, with the increase of y=2 to y=12, the 9-level combination of sets is completed, and the number of H sets after each level remains unchanged or increases on the basis of the previous level, under the state of y=12 The number of H sets that exist is the largest, and finally the fifth row of the table in Figure 5(b) is obtained. Each curly bracket in the fifth line represents a hole or suspected hole connected domain, and the value in the curly brackets is the label of the H set contained in the connected domain (the white number in Figure 7(a), the label is the state before the start of set merging and renaming ); count the number of curly brackets in the fifth row of Figure 7(b), and obtain the total number of holes or suspected hole connected domains corresponding to the target connected domain, as shown in the last row of the table in Figure 7(b). The label of the H set in the last row of the table is no longer continuous. This is because the merge and renaming of the sets occurred with the increase of the scan value. At the same time, although the same set name appears in multiple positions in the line, the right set The number of elements is equal to or greater than the elements in the collection on the left. Then, step S405 is executed.

S405. Check the suspected connected domain of the hole to obtain the connected domain that actually has an inner boundary; specifically, take the union of the suspected connected domain to be analyzed and the corresponding target connected domain, and respectively verify that the suspected connected domain is upward, downward, and leftward , whether the area after shifting one pixel to the right is completely included in the union. If yes, it indicates that the suspected void connected domain is a hole inside the target connected domain A _i being analyzed, which corresponds to the structure of the object itself; if not, it indicates that the connected domain is a groove or concave structure on the outer boundary.

As shown in Figure 7(c), the target connected domain to be analyzed in Figure 7(c) is denoted as A _i , and the corresponding suspected hole connected domain is H _j , H _j is translated up, down, left, and right to obtain H _j-1 , H _j-2 , H _j-3 , H _j-4 . Find the union of A _i and H _i , and then use H _j-1 , H _j-2 , H _j-3 , H _j-4 to find the difference with the union of A _i and H _i respectively, if all the results are empty Set, indicating that the suspected hole connected domain after four translations is included in the union of the original suspected hole connected domain and the target connected domain to be analyzed, that is, H _i is the hole inside A _i ; otherwise, it is a concave structure on the edge.

Exemplarily, in Fig. 7 (a), all the pixels labeled 4, 6, 7, and 8 form a suspected hole-connected domain, and after they are translated to the left, right, up, and down, the pixel coverage area is itself and Mark the area corresponding to the white circular pixels, obviously they are completely contained in the union of the gray connected domain and itself, so it belongs to a hole in the target connected domain; all the pixels labeled 2 and 3 in Figure 7(a) consist A suspected hole-connected domain is established. After they are translated to the left, right, up, and down, the pixel coverage area is the area corresponding to itself and the marked white triangle pixel. Obviously, when it translates to the left, two pixels marking the triangle (coordinates (10, 1), (11, 1)) move out of the union of itself and the gray connected domain, so it is not in the gray connected domain A real void, but a concave structure that appears at the border.

After the detection is completed, step S406 is executed.

S406. Perform burr pruning on the detected hole connected domains. The principle of pruning is the same as step S403, and will not be repeated here; after pruning is completed, a set C _n is generated; and step S6 is executed.

S5. The final boundary of the nth target is D _n =A _n -B _n , restore the gray scale of the pixel corresponding to each coordinate in D ₁ , D ₂ , D ₃ ... D _n to the state in P1, and adjust the other pixels to 255. Obtain the corrected image P4 that removes the influence of the carton package.

S6. The final boundary of the nth object is D _n =A _n -B _n +C _n , restore the pixel grayscale corresponding to each coordinate in D ₁ , D ₂ , D ₃ ... D _n to the status in P1, and other The pixel is adjusted to 255 to obtain the corrected image P4 that removes the influence of the carton packaging.

Embodiment two

This embodiment discloses a pre-processing system for terahertz security images, including:

The image acquisition module is configured to: acquire a terahertz security check image not containing a sample and a terahertz security check image containing a sample;

The background image analysis module is configured to: perform histogram analysis on the terahertz security inspection image that does not contain samples, and obtain the gray value corresponding to the largest number of pixels;

The noise reduction module is configured to: perform a corrosion operation on the terahertz security inspection image containing the sample, use the gray value corresponding to the largest number of pixels as the threshold value, and perform binary gray-scale transformation on the processed image;

The burr pruning module is configured to: perform connected domain analysis on the binary grayscale processed image, obtain the boundary of the sample, and perform burr pruning on the boundary;

The corrected image acquisition module is configured to: acquire the corrected image according to the image trimmed by the burr.

It should be noted here that the above-mentioned image acquisition module, background image analysis module, noise reduction module, burr trimming module and corrected image acquisition module correspond to the steps in Embodiment 1, and the examples and applications realized by the above-mentioned modules and corresponding steps The scenarios are the same, but are not limited to the content disclosed in Embodiment 1 above. It should be noted that, as a part of the system, the above-mentioned modules can be executed in a computer system such as a set of computer-executable instructions.

Embodiment three

Embodiment 3 of the present invention provides an electronic device, including a memory, a processor, and computer instructions stored in the memory and run on the processor. When the computer instructions are run by the processor, the above-mentioned preprocessing method for terahertz security images is completed. A step of.

Embodiment Four

Embodiment 4 of the present invention provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps of the above-mentioned preprocessing method for terahertz security images are completed.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each process and/or block in the flowchart and/or block diagram, and a combination of processes and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions may be provided to a general purpose computer, special purpose computer, embedded processor, or processor of other programmable data processing equipment to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing equipment produce a An apparatus for realizing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions The device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions can also be loaded onto a computer or other programmable data processing device, and a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device Steps are provided for implementing the functions specified in the flow chart or flow charts and/or block diagram block or blocks.

The description of each embodiment in the foregoing embodiments has its own emphases, and for parts not described in detail in a certain embodiment, reference may be made to relevant descriptions of other embodiments.

The above descriptions are only preferred embodiments of the present application, and are not intended to limit the present application. For those skilled in the art, various modifications and changes may be made to the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of this application shall be included within the protection scope of this application.

Claims (10)

1. A pretreatment method for a terahertz security inspection image is characterized by comprising the following steps:

acquiring a terahertz security check image not containing a sample and a terahertz security check image containing the sample; wherein the sample is a carton or envelope that may contain hazardous materials;

performing histogram analysis on the terahertz security inspection image without the sample to acquire the gray value with the maximum number of corresponding pixels;

carrying out corrosion operation on the terahertz security inspection image containing the sample, and carrying out binarization gray scale transformation on the processed image by taking the gray scale value with the maximum number of corresponding pixels as a threshold value;

performing connected domain analysis on the image subjected to the binary gray processing to obtain a boundary of a sample, and performing burr trimming on the boundary;

and acquiring a corrected image according to the image after the burr trimming.

2. The preprocessing method for the terahertz security inspection image as claimed in claim 1, wherein the performing connected domain analysis on the binary gray processed image to obtain the boundary of the sample, and performing burr trimming on the boundary comprises:

performing connected domain analysis on the image subjected to the binary gray processing to obtain the outer boundary of the sample, and performing burr trimming on the outer boundary;

detecting whether an inner boundary exists in a connected domain;

and if the inner boundary exists in the connected domain, performing burr trimming on the inner boundary.

3. The pretreatment method for the terahertz security inspection image as claimed in claim 2, wherein the step of performing connected domain analysis on the binary gray-scale processed image to obtain the outer boundary of the sample comprises the following specific steps:

and acquiring a connected domain with the gray value of 255 based on the image after the binary gray processing, eliminating abnormal regions and acquiring the outer boundary of the sample.

4. The pretreatment method for the terahertz security inspection image as claimed in claim 2, wherein the specific steps of performing burr trimming on the outer boundary are as follows:

marking an initial outer boundary queue of a connected domain;

based on the initial outer boundary queue, searching a pixel coordinate corresponding to the queue index according to the queue index; if the pixel coordinates corresponding to different queue indexes are the same, deleting the pixel coordinates of the index value between the two deleted queue indexes in the initial outer boundary queue;

and traversing and searching the adjusted outer boundary queue until the whole queue is processed.

5. The pretreatment method for the terahertz security inspection image as claimed in claim 4, wherein the specific steps of marking the initial outer boundary queue of the connected domain are as follows:

finding out the pixel with the minimum x coordinate or y coordinate in the connected domain, and taking the pixel coordinate as the head of the initial outer boundary queue;

taking the right as an initial searching direction, and carrying out searching in the order of left, front, right and back relative to the searching direction; if the pixel of the searched pixel is in the connected domain, adding the pixel coordinate into the initial outer boundary queue, and updating the searching direction;

if the added pixel coordinate is the same as the start point coordinate, the search terminates.

6. The pretreatment method for the terahertz security inspection image as claimed in claim 1, wherein the specific steps of performing binarization grayscale transformation on the processed image by using the grayscale value with the maximum number of corresponding pixels as a threshold value are as follows:

the gray scale values of all pixels with the gray scale values larger than or equal to the threshold are adjusted to be 0, and the gray scale values of the pixels with the gray scale values lower than the threshold are adjusted to be 255.

7. The preprocessing method for the terahertz security inspection image as claimed in claim 1, wherein the specific steps of obtaining the corrected image according to the image after the burr trimming are as follows:

restoring the pixel gray value corresponding to each coordinate in the connectivity domain after the burrs are trimmed to be the pixel gray value corresponding to the terahertz security inspection image containing the sample;

the other pixels in the image after the burr trimming are adjusted to 255, and a corrected image is obtained.

8. A pretreatment system for a terahertz security inspection image is characterized by comprising:

an image acquisition module configured to: acquiring a terahertz security check image not containing a sample and a terahertz security check image containing the sample; wherein the sample is a carton or envelope that may contain hazardous materials;

a background picture analysis module configured to: performing histogram analysis on the terahertz security inspection image without the sample to acquire a gray value with the maximum number of corresponding pixels;

a noise reduction module configured to: carrying out corrosion operation on the terahertz security inspection image containing the sample, and carrying out binarization gray scale transformation on the processed image by taking the gray scale value with the maximum number of corresponding pixels as a threshold value;

a spike trimming module configured to: performing connected domain analysis on the image subjected to the binary gray processing to obtain the boundary of the sample, and performing burr trimming on the boundary;

a modified image acquisition module configured to: and acquiring a corrected image according to the image after the burr trimming.

9. An electronic device comprising a memory and a processor and computer instructions stored on the memory and executed on the processor, the computer instructions when executed by the processor performing the steps of any of claims 1-7.

10. A computer-readable storage medium storing computer instructions which, when executed by a processor, perform the steps of any one of claims 1 to 7.

Images