A method for segmenting body parts from a millimeter wave human body image

Abstract

The application discloses a kind of millimeter wave human body image human body part segmentation methods, method includes: to original millimeter wave image is threshold segmentation, obtain front and back human body binary image;Front and back human body binary image is respectively in front and back direction superposition and obtain front and back human body rough area binary image;Based on distance transformation, human body rough area binary image is morphologically processed, and human body area binary image is obtained;Take the complement image of human body area in human body rough area binary image, analyze the connected region feature of complement image and judge whether the posture of examinee is standard;Based on complement image, connected region analysis is carried out, and then the positioning and segmentation of each part of human body are carried out, and privacy part is filtered according to segmentation result. Using the method of the application, each part of human body in millimeter wave human body image can be accurately positioned and segmented, and privacy filtering of specified human body part can be completed based on this.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and specifically to a method for segmenting human body parts in millimeter-wave human images. Background Technology

[0002] With societal progress and heightened public safety awareness, security checks on personnel entering and exiting important locations have become a crucial issue in the field of public safety. Traditional metal detectors have limitations in personnel security checks, failing to effectively detect non-metallic hazardous materials and lacking non-contact screening capabilities. Millimeter-wave security technology, with its strong penetrating power, non-contact detection, and sensitivity to non-metallic substances, effectively compensates for the shortcomings of traditional metal detectors, making it a current research hotspot and development trend in the field of personnel security checks.

[0003] Because human bodies vary in shape and the types, sizes and materials of concealed objects differ, current target detection for millimeter-wave human images is generally based on the idea of ​​image segmentation to extract features and edge information. However, based on the need for abnormal posture recognition and privacy protection, it is also necessary to be able to quickly and accurately locate and segment various parts of the human body.

[0004] Chinese patent application CN109410151A discloses an active millimeter-wave imaging privacy protection algorithm. The main process is as follows: Noisy millimeter-wave images are filtered and smoothed; longitudinal energy superposition of the millimeter-wave images yields a transverse echo energy distribution curve of the human body; the transverse echo energy curve of the superimposed longitudinal energy is truncated, retaining the main body portion to remove the influence of arm echoes on subsequent detection of private parts of the human body; the obtained echo energy curve of the main body portion is analyzed to obtain the transverse center coordinates of the human body; the transverse center coordinates of the human body are analyzed using the transverse distribution curve of the leg echo energy, and averaged with the aforementioned center coordinates; the obtained images of the main body portion are transversely superimposed to obtain a longitudinal echo energy distribution curve of the human body; the obtained longitudinal energy distribution curve is analyzed to obtain the longitudinal coordinate positions of the head, neck, chest, and groin; the groin and chest portions are blurred to protect personal privacy.

[0005] This technical solution is based on the horizontal and vertical echo energy distribution curves of millimeter-wave images for segmentation. The curve-based segmentation method is not only sensitive to parameters, but also requires high image quality. It may not achieve ideal segmentation results for millimeter-wave images with a lot of noise and foreign objects. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention aims to provide a method for segmenting human body parts from millimeter-wave human images.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A method for segmenting human body parts in millimeter-wave human images, comprising the following steps:

[0009] S1. Obtain frontal and rear millimeter-wave human body images by scanning with millimeter-wave equipment;

[0010] S2. Threshold segmentation is performed on the front millimeter-wave human body image and the back millimeter-wave human body image to obtain the front human body binary image and the back human body binary image, respectively.

[0011] S3. Overlay the frontal human body binary image and the back human body binary image to obtain a rough binary image of the frontal human body region and a rough binary image of the back human body region.

[0012] S4. Based on distance transformation, perform morphological processing on the frontal and back human body rough region binary images to obtain the frontal and back human body region binary images.

[0013] S5. Take the complement images of the human body regions from the front and back binary images of the human body regions respectively, and analyze the connected component features of the complement images to determine whether the posture of the inspected person is standard. If it is standard, proceed to step S6; otherwise, return to step S1 to reacquire the front millimeter-wave human body image and the back millimeter-wave human body image.

[0014] S6. Perform connected component analysis on the complement images of the frontal and back human body regions, respectively, and locate and segment each human body part in the frontal and back millimeter-wave human body images accordingly.

[0015] Furthermore, the above method also includes the following steps:

[0016] S7. Based on the segmentation results of step S6, perform privacy filtering on the human body parts set in the front millimeter-wave human body image and the back millimeter-wave human body image.

[0017] Further, the specific process of step S3 is as follows: the front human body binary image and the back human body binary image are symmetrical along the height direction. The same position of the front human body binary image and the back human body binary image is ORed pixel by pixel to obtain the front human body rough area binary image and the back human body rough area binary image respectively.

[0018] Furthermore, the specific process of step S4 is as follows:

[0019] S4.1 Calculate the distance transformation matrix of the binary image of the coarse human body region;

[0020] S4.2. Dilation processing based on the distance transformation matrix of the coarse human body region binary image: Traverse the distance transformation matrix of the coarse human body region binary image and record all pixel positions less than or equal to the threshold d; in the coarse human body region binary image, set the pixel values ​​of the recorded pixel positions less than or equal to the threshold d to 255, and set the pixel values ​​of other positions to 0, to obtain a binary image based on distance dilation d times.

[0021] S4.3. Extract the edges from the binary image obtained in step S4.2 based on distance dilation d times to obtain the edge image;

[0022] S4.4 Calculate the distance transformation matrix of the edge image obtained in step S4.3;

[0023] S4.5. Erosion processing based on the distance transformation matrix of the edge image: Traverse the distance transformation matrix of the edge image and record all pixel positions less than the threshold e; In the binary image based on distance dilation d times, set the pixel values ​​of the recorded pixel positions less than the threshold e to 0, and leave the pixel values ​​of other positions unchanged, to obtain the binary image based on distance erosion e times.

[0024] S4.6 Delete the small connected regions outside the human body in the binary map obtained in step S4.5 based on distance erosion e times, to obtain the binary map of the human body region.

[0025] Furthermore, the specific process of step S5 is as follows:

[0026] S5.1 Find the bounding rectangle of the human body region in the binary image of the human body region, and set the pixel value of the bounding rectangle to 255;

[0027] S5.2 Invert the binary image of the human body region within the bounding rectangle to obtain the complement image of the human body region;

[0028] S5.3 Connected component labeling: Label each connected component of the complement image obtained in step S5.2;

[0029] S5.4 Calculate the area and position of each connected region in the complement image;

[0030] S5.5. Compare the area and position of each connected region in the complement image with the area and position of each connected region in the preset standard pose to determine whether the pose is standard.

[0031] Furthermore, in step S6, the complement image of the human body region is specifically subjected to connected component analysis to obtain the segmentation positions of the human head, left arm, right arm, left hand, right hand, body, thigh, and calf, and the millimeter-wave human body image is segmented into 10 regions accordingly.

[0032] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0033] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0034] The beneficial effects of this invention are as follows: using the method of this invention, various parts of the human body in millimeter-wave human body images can be accurately located and segmented, and based on this, privacy filtering of specified human body parts can be completed. Attached Figure Description

[0035] Figure 1 This is a flowchart of the method in Embodiment 1 of the present invention;

[0036] Figure 2 The images are the frontal and rear millimeter-wave human body images obtained in Embodiment 2 of the present invention.

[0037] Figure 3 The two images are the frontal and back binary images of a human body obtained by threshold segmentation of the frontal and back millimeter-wave human body images in Embodiment 2 of the present invention.

[0038] Figure 4 These are the frontal and back rough human body region binary images obtained in Embodiment 2 of the present invention.

[0039] Figure 5 The binary images of the front and back human body regions, which are morphologically processed and contain only the human body, obtained in Embodiment 2 of the present invention.

[0040] Figure 6 This is the complement image obtained from the binary image of the frontal human body region and the binary image of the back human body region in Embodiment 2 of the present invention;

[0041] Figure 7 This is a schematic diagram of the dividing line obtained from the front millimeter-wave human body image and the back millimeter-wave human body image in Embodiment 2 of the present invention;

[0042] Figure 8 This is the image obtained by segmenting the frontal millimeter-wave human body image and the rear millimeter-wave human body image in Embodiment 2 of the present invention;

[0043] Figure 9 This is the image after head privacy filtering in Embodiment 2 of the present invention;

[0044] Figure 10 The images are the frontal and rear millimeter-wave human body images obtained in Embodiment 3 of the present invention.

[0045] Figure 11 The two images are the frontal and back binary images of a human body obtained by threshold segmentation of the frontal and back millimeter-wave human body images in Embodiment 3 of the present invention.

[0046] Figure 12 These are the frontal and rear rough human body region binary images obtained in Embodiment 3 of the present invention.

[0047] Figure 13 The binary images of the front and back human body regions, which are morphologically processed and contain only the human body, obtained in Embodiment 3 of the present invention.

[0048] Figure 14 This is the complement image obtained from the binary image of the frontal human body region and the binary image of the back human body region in Embodiment 3 of the present invention;

[0049] Figure 15 This is a schematic diagram of the segmentation lines obtained from the front millimeter-wave human body image and the back millimeter-wave human body image in Embodiment 3 of the present invention;

[0050] Figure 16 This is the image obtained by segmenting the frontal millimeter-wave human body image and the rear millimeter-wave human body image in Embodiment 3 of the present invention;

[0051] Figure 17 This is the image after head privacy filtering in Embodiment 3 of the present invention. Detailed Implementation

[0052] The present invention will be further described below with reference to the accompanying drawings. It should be noted that this embodiment is based on the present technical solution and provides detailed implementation methods and specific operation processes, but the protection scope of the present invention is not limited to this embodiment.

[0053] Example 1

[0054] This embodiment provides a method for human body part segmentation in millimeter-wave human images, which solves the problems of human body part localization and segmentation in millimeter-wave human images, and achieves accurate segmentation of the original millimeter-wave human images.

[0055] like Figure 1 As shown, the method for segmenting human body parts in millimeter-wave human images mainly includes the following steps:

[0056] S1. Obtain frontal and rear millimeter-wave human body images by scanning with millimeter-wave equipment;

[0057] S2. Threshold segmentation is performed on the front millimeter-wave human body image and the back millimeter-wave human body image to obtain the front human body binary image and the back human body binary image, respectively.

[0058] S3. Overlay the frontal human body binary image and the back human body binary image to obtain a rough binary image of the frontal human body region and a rough binary image of the back human body region.

[0059] S4. Based on distance transformation, perform morphological processing on the frontal and back human body rough region binary images to obtain the frontal and back human body region binary images.

[0060] S5. Take the complement images of the human body regions from the front and back binary images of the human body regions respectively, and analyze the connected component features of the complement images to determine whether the posture of the inspected person is standard. If it is standard, proceed to step S6; otherwise, return to step S1 to reacquire the front millimeter-wave human body image and the back millimeter-wave human body image.

[0061] S6. Perform connected component analysis on the complement images of the frontal and back human body regions respectively, and locate and segment each human body part in the frontal and back millimeter-wave human body images accordingly.

[0062] S7. Based on the segmentation results of step S6, perform privacy filtering on the human body parts set in the front millimeter-wave human body image and the back millimeter-wave human body image.

[0063] Specifically, in this embodiment, the specific process of step S3 is as follows: the front human body binary image and the back human body binary image are symmetrical along the height direction, and the same position of the front human body binary image and the back human body binary image are ORed pixel by pixel to obtain the front human body rough area binary image and the back human body rough area binary image respectively.

[0064] Specifically, in step S4, the morphological processing of the rough binary image of the human body from the front and the rough binary image of the human body from the back is performed as follows:

[0065] S4.1 Calculate the distance transformation matrix of the binary image of the coarse human body region;

[0066] S4.2. Dilation processing based on the distance transformation matrix of the coarse human body region binary image: Traverse the distance transformation matrix of the coarse human body region binary image and record all pixel positions less than or equal to the threshold d; in the coarse human body region binary image, set the pixel values ​​of the recorded pixel positions less than or equal to the threshold d to 255, and set the pixel values ​​of other positions to 0, to obtain a binary image based on distance dilation d times.

[0067] S4.3. Extract the edges from the binary image obtained in step S4.2 based on distance dilation d times to obtain the edge image;

[0068] S4.4 Calculate the distance transformation matrix of the edge image obtained in step S4.3;

[0069] S4.5. Erosion processing based on the distance transformation matrix of the edge image: Traverse the distance transformation matrix of the edge image and record all pixel positions less than the threshold e; In the binary image based on distance dilation d times, set the pixel values ​​of the recorded pixel positions less than the threshold e to 0, and leave the pixel values ​​of other positions unchanged, to obtain the binary image based on distance erosion e times.

[0070] S4.6 Delete the connected regions outside the human body in the binary map obtained in step S4.5 based on distance erosion e times, to obtain the binary map of the human body region.

[0071] Specifically, in this embodiment, in step S5, the binary images of the front and back human body regions are analyzed to determine whether the posture of the person being examined is standard, according to the following process:

[0072] S5.1 Find the bounding rectangle of the human body region in the binary image of the human body region, and set the pixel value of the bounding rectangle to 255;

[0073] S5.2. Invert (complement) the binary image of the human body region within the bounding rectangle to obtain the complement image of the human body region;

[0074] S5.3 Connected component labeling: Label each connected component of the complement image obtained in step S5.2;

[0075] S5.4 Calculate the area and position of each connected region in the complement image;

[0076] S5.5. Compare the area and position of each connected region in the complement image with the area and position of each connected region in the preset standard pose to determine whether the pose is standard.

[0077] Specifically, in this embodiment, in step S6, the complement image of the human body region is subjected to connected component analysis to obtain the segmentation positions of the human head, left arm, right arm, left hand, right hand, body, thigh, and calf, and the millimeter-wave human body image is segmented into 10 regions accordingly.

[0078] Example 2

[0079] This embodiment provides an application example of the method described in Embodiment 1, including the following steps:

[0080] S1. Obtain a frontal millimeter-wave human body image using millimeter-wave human body security inspection equipment. Figure 2 (a) and the back millimeter-wave human body image ( Figure 2 (b)

[0081] S2. Perform fixed threshold segmentation on the frontal and back millimeter-wave human body images to obtain the frontal and back human body binary images as shown below. Figure 3 of (a), Figure 3 As shown in (b). In this embodiment, the segmentation threshold is approximately 25% of the image saturation value. For example, if the image saturation value is 255, this value can be 60.

[0082] S3. Overlay the frontal and back binary images of the human body, ensuring the front and back images are symmetrical along the height direction. Perform an OR operation on each pixel at the same location to obtain coarse binary images of the frontal and back human body regions, as shown below. Figure 4 of (a), Figure 4 As shown in (b).

[0083] S4. Perform morphological processing based on distance transformation on the rough binary image of the human body region from the front and back, respectively, to obtain the rough binary image of the human body region from the front and back. The specific process is as follows:

[0084] S4.1 Calculate the distance transformation matrix of the binary image of the coarse human body region, that is, calculate the distance between each pixel in the image and its nearest non-zero pixel.

[0085] S4.2. Based on the distance transformation matrix calculated in step S4.1, dilate the coarse human body region binary image: traverse the distance transformation matrix and record all pixel positions less than or equal to the threshold 5. Set the pixel values ​​of all pixel positions less than or equal to the threshold 5 in the coarse human body region binary image to 255, and set the other pixel values ​​to 0, to obtain a binary image based on distance dilation 5 times;

[0086] S4.3. Extract the edges from the binary image based on distance dilation of 5 times to obtain the edge image;

[0087] S4.4 Calculate the distance transformation matrix of the edge image;

[0088] S4.5. Erosion processing based on the distance transformation matrix of the edge image: Traverse the distance transformation matrix of the edge image and record the positions of all pixels less than the threshold 2; In the binary image after 5 dilations based on distance, set the pixel values ​​of all pixel positions less than the threshold 2 to 0, and leave the pixel values ​​of the remaining positions unchanged, to obtain a binary image after 2 erosions based on distance.

[0089] S4.6. Delete small connected regions (excluding the human body) from the binary image based on distance erosion twice, obtaining the binary image of the human body region. The resulting frontal and back view binary images of the human body region are shown below. Figure 5 of (a), Figure 5 As shown in (b).

[0090] S5. Perform complement image extraction on the human body regions in the front human body region binary image and the back human body region binary image obtained in step S4 respectively, conduct connected region analysis on the complement images, and determine whether the posture of the current inspected person is standard. The specific process is as follows:

[0091] S5.1. Find the bounding rectangle border of the human body region, and set the pixel value at the position of the bounding rectangle border to 255;

[0092] S5.2. Invert (take the complement of) the image obtained in step S5.1 within the bounding rectangle border region to obtain the complement image of the human body region. The complement image of the front human body region and the complement image of the back human body region are respectively as shown in Figure 6 (a) and (b) therein;

[0093] S5.3. Connected region marking: Mark the 5 connected regions of the complement image obtained in step S5.2 as 1 to 5 respectively, and the marking sequence is as Figure 6 shown. In the following description, connected region 1 refers to the connected region marked as 1, and so on;

[0094] S5.4. Calculate the area and position of each of the 5 connected regions respectively;

[0095] S5.5. Determine whether the posture is standard. In this embodiment, the determination rules are as follows (taking the complement image of the front human body region as an example):

[0096] If S1*1.3 < S3, the right arm is positioned too high;

[0097] If S1*1.7 > S3, the right arm is positioned too low;

[0098] If S2*1.3 < S4, the left arm is positioned too high;

[0099] If S2*1.7 > S4, the left arm is positioned too low;

[0100] If M5 < 0.5*width - 10, the standing position is deviated to the right;

[0101] If M5 > 0.5*width + 10, the standing position is deviated to the left;

[0102] If the posture is non-standard;

[0103] Wherein, S1 to S4 respectively represent the areas of connected region 1, connected region 2, connected region 3 and connected region 4, M5 represents the midpoint position of connected region 5 in the image width direction, and width represents the width of the image.

[0104] S6. Perform connected component analysis on the complemented images of the frontal and back human body regions obtained in step S5 to obtain the segmentation positions of the head, left arm, right arm, left hand, right hand, body, thigh, and calf. Based on this, the frontal and back millimeter-wave human body images are segmented into 10 regions respectively. The specific process is as follows (taking the complemented image of the frontal human body region as an example):

[0105] S6.1 Find the shortest horizontal distance between connected component 1 and connected component 2 in the complement image, and combine it with the head size to obtain the dividing line 1;

[0106] S6.2 Find the highest vertex of connected component 3 and connected component 4, and draw a dividing line upwards to obtain dividing line 2 and dividing line 3;

[0107] S6.3 Find the shortest distance in the image height direction between connected region 1 and connected region 3 and between connected region 2 and connected region 4, and obtain the dividing line 4 and dividing line 5.

[0108] S6.4 Calculate the peak points of the curve above the connected region 5, and take the two largest peak points, one to the left and one to the right, which are the two thigh dividing lines 6 and 7.

[0109] S6.5 Find the shortest distance between connected region 3 and connected region 5 and between connected region 5 and connected region 4 in the image width direction where there is overlap. Combined with the proportion of thigh and calf, obtain the dividing lines 8 and 9 of thigh and calf.

[0110] S6.6. Divide the millimeter-wave human body image into 10 parts according to the above dividing lines. The dividing lines for the front and back millimeter-wave human body images are as follows: Figure 7 As shown in (a) and (b), the segmented frontal and rear millimeter-wave human body images are respectively as follows: Figure 8 As shown in (a) and (b).

[0111] S7. Based on the human body region location obtained in step S6, perform image blurring on the head of the frontal millimeter-wave human body image to obtain a head privacy-filtered image, such as... Figure 9 As shown.

[0112] Example 3

[0113] This embodiment provides another application example of the method in Embodiment 1, including the following steps:

[0114] S1. Obtain a frontal millimeter-wave human body image using millimeter-wave human body security inspection equipment. Figure 10 (a) and the back millimeter-wave human body image ( Figure 10 (b)

[0115] S2. Perform fixed threshold segmentation on the frontal and back millimeter-wave human body images to obtain the frontal and back human body binary images as shown below. Figure 11 of (a), Figure 11 As shown in (b). In this embodiment, the segmentation threshold is approximately 25% of the image saturation value. For example, if the image saturation value is 255, this value can be 60.

[0116] S3. Overlay the frontal and back binary images of the human body, ensuring the front and back images are symmetrical along the height direction. Perform an OR operation on each pixel at the same location to obtain coarse binary images of the frontal and back human body regions, as shown below. Figure 12 of (a), Figure 12 As shown in (b).

[0117] S4. Perform morphological processing based on distance transformation on the rough binary image of the human body region from the front and back, respectively, to obtain the rough binary image of the human body region from the front and back. The specific process is as follows:

[0118] S4.1 Calculate the distance transformation matrix of the binary image of the coarse human body region, that is, calculate the distance between each pixel in the image and its nearest non-zero pixel.

[0119] S4.2. Based on the distance transformation matrix calculated in step S4.1, dilate the coarse human body region binary image: traverse the distance transformation matrix and record all pixel positions less than or equal to the threshold 5. Set the pixel values ​​of all pixel positions less than or equal to the threshold 5 in the coarse human body region binary image to 255, and set the other pixel values ​​to 0, to obtain a binary image based on distance dilation 5 times;

[0120] S4.3. Extract the edges from the binary image based on distance dilation of 5 times to obtain the edge image;

[0121] S4.4 Calculate the distance transformation matrix of the edge image;

[0122] S4.5. Erosion processing based on the distance transformation matrix of the edge image: Traverse the distance transformation matrix of the edge image and record the positions of all pixels less than the threshold 2; In the binary image after 5 dilations based on distance, set the pixel values ​​of all pixel positions less than the threshold 2 to 0, and leave the pixel values ​​of the remaining positions unchanged, to obtain a binary image after 2 erosions based on distance.

[0123] S4.6. Delete small connected regions (excluding the human body) from the binary image based on distance-based erosion twice, obtaining the binary image of the human body region. The frontal and back view binary images of the human body region are shown below. Figure 13 of (a), Figure 13 As shown in (b).

[0124] S5. Respectively taking the complement image of the human body region in the front human body region binary map and the back human body region binary map obtained in step S4, performing connected region analysis on the complement images, and determining whether the posture of the current inspected person is standard. The specific process is as follows:

[0125] S5.1. Find the bounding rectangle border of the human body region, and set the pixel value at the position of the bounding rectangle border to 255;

[0126] S5.2. Invert (complement) the image obtained in step S5.1 within the bounding rectangle border region to obtain the complement image of the human body region. The complement image of the front human body region and the complement image of the back human body region are respectively as Figure 14 shown in (a) and (b) in the figure;

[0127] S5.3. Connected region labeling: Label the 5 connected regions of the complement image obtained in step S5.2 as 1 to 5 respectively, and the labeling sequence is as Figure 14 shown below. In the following text, connected region 1 refers to the connected region labeled 1, and so on;

[0128] S5.4. Respectively calculate the area and position of the 5 connected regions;

[0129] S5.5. Determine whether the posture is standard. In this embodiment, the determination rules are as follows (taking the complement image of the front human body region as an example):

[0130] If S1*1.3 < S3, the right arm is too high;

[0131] If S1*1.7 > S3, the right arm is too low;

[0132] If S2*1.3 < S4, the left arm is too high;

[0133] If S2*1.7 > S4, the left arm is too low;

[0134] If M5 < 0.5*width - 10, the standing position is too far to the right;

[0135] If M5 > 0.5*width + 10, the standing position is too far to the left;

[0136] If the posture is non-standard;

[0137] wherein S1 to S4 respectively represent the areas of connected region 1, connected region 2, connected region 3, and connected region 4, M5 represents the midpoint position of connected region 5 in the width direction of the image, and width represents the width of the image.

[0138] S6. Perform connected component analysis on the complemented images of the frontal and back human body regions obtained in step S5 to obtain the segmentation positions of the head, left arm, right arm, left hand, right hand, body, thigh, and calf. Based on this, the frontal and back millimeter-wave human body images are segmented into 10 regions respectively. The specific process is as follows (taking the complemented image of the frontal human body region as an example):

[0139] S6.1 Find the shortest horizontal distance between connected component 1 and connected component 2 in the complement image, and combine it with the head size to obtain the dividing line 1;

[0140] S6.2 Find the highest vertex of connected component 3 and connected component 4, and draw a dividing line upwards to obtain dividing line 2 and dividing line 3;

[0141] S6.3 Find the shortest distance in the image height direction between connected region 1 and connected region 3 and between connected region 2 and connected region 4, and obtain the dividing line 4 and dividing line 5.

[0142] S6.4 Calculate the peak points of the curve above the connected region 5, and take the two largest peak points, one to the left and one to the right, which are the two thigh dividing lines 6 and 7.

[0143] S6.5 Find the shortest distance between connected region 3 and connected region 5 and between connected region 5 and connected region 4 in the image width direction where there is overlap. Combined with the proportion of thigh and calf, obtain the dividing lines 8 and 9 of thigh and calf.

[0144] S6.6. Divide the millimeter-wave human body image into 10 parts according to the above dividing lines. The dividing lines for the front and back millimeter-wave human body images are as follows: Figure 15 As shown in (a) and (b), the segmented frontal and rear millimeter-wave human body images are respectively as follows: Figure 16 As shown in (a) and (b).

[0145] S7. Based on the human body region location obtained in step S6, perform image blurring on the head of the frontal millimeter-wave human body image to obtain a head privacy-filtered image, such as... Figure 17 As shown.

[0146] For those skilled in the art, various corresponding changes and modifications can be made based on the above technical solutions and concepts, and all such changes and modifications should be included within the protection scope of the claims of this invention.

Claims

1. A method for segmenting human body parts in millimeter-wave human images, characterized in that, Includes the following steps: S1. Obtain frontal and rear millimeter-wave human body images by scanning with millimeter-wave equipment; S2. Threshold segmentation is performed on the front millimeter-wave human body image and the back millimeter-wave human body image to obtain the front human body binary image and the back human body binary image, respectively. S3. Overlay the frontal human body binary image and the back human body binary image to obtain a rough binary image of the frontal human body region and a rough binary image of the back human body region. S4. Based on distance transformation, perform morphological processing on the coarse binary image of the human body region from the front and back, respectively, to obtain the binary image of the human body region from the front and back; the specific process is as follows: S4.1 Calculate the distance transformation matrix of the binary image of the coarse human body region; S4.

2. Dilation processing based on the distance transformation matrix of the coarse human body region binary image: Traverse the distance transformation matrix of the coarse human body region binary image and record the positions of all pixels less than or equal to the threshold d. In the binary map of the coarse human body region, the pixel values ​​of the recorded pixel positions that are less than or equal to the threshold d are set to 255, while the pixel values ​​of other positions are set to 0, resulting in a binary map based on distance dilation d times. S4.

3. Extract the edges from the binary image obtained in step S4.2 based on distance dilation d times to obtain the edge image; S4.4 Calculate the distance transformation matrix of the edge image obtained in step S4.3; S4.

5. Erosion processing based on the distance transformation matrix of the edge image: Traverse the distance transformation matrix of the edge image and record all pixel positions less than the threshold e; In the binary image based on distance dilation d times, set the pixel values ​​of the recorded pixel positions less than the threshold e to 0, and leave the pixel values ​​of other positions unchanged, to obtain the binary image based on distance erosion e times. S4.6 Delete the small connected regions outside the human body in the binary map obtained in step S4.5 based on distance erosion e times, to obtain the binary map of the human body region. S5. Take the complement images of the human body regions from the front and back binary images of the human body regions respectively, and analyze the connected component features of the complement images to determine whether the posture of the inspected person is standard. If it is standard, proceed to step S6; otherwise, return to step S1 to reacquire the front millimeter-wave human body image and the back millimeter-wave human body image. S6. Perform connected component analysis on the complement images of the frontal and back human body regions, respectively, and locate and segment each human body part in the frontal and back millimeter-wave human body images accordingly.

2. The method according to claim 1, characterized in that, It also includes the following steps: S7. Based on the segmentation results of step S6, perform privacy filtering on the human body parts set in the front millimeter-wave human body image and the back millimeter-wave human body image.

3. The method according to claim 1, characterized in that, The specific process of step S3 is as follows: the front human body binary image and the back human body binary image are symmetrical along the height direction. The same position of the front human body binary image and the back human body binary image is ORed pixel by pixel to obtain the front human body rough area binary image and the back human body rough area binary image respectively.

4. The method according to claim 1, characterized in that, The specific process of step S5 is as follows: S5.1 Find the bounding rectangle of the human body region in the binary image of the human body region, and set the pixel value of the bounding rectangle to 255; S5.2 Invert the binary image of the human body region within the bounding rectangle to obtain the complement image of the human body region; S5.3 Connected component labeling: Label each connected component of the complement image obtained in step S5.2; S5.4 Calculate the area and position of each connected region in the complement image; S5.

5. Compare the area and position of each connected region in the complement image with the area and position of each connected region in the preset standard pose to determine whether the pose is standard.

5. The method according to claim 1, characterized in that, In step S6, the complement image of the human body region is specifically subjected to connected component analysis to obtain the segmentation positions of the human head, left arm, right arm, left hand, right hand, body, thigh, and calf, and the millimeter-wave human body image is segmented into 10 regions accordingly.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method described in any one of claims 1-5.

Images