CN115311684A - Method for integrating millimeter wave image multi-angle detection results - Google Patents
Method for integrating millimeter wave image multi-angle detection results Download PDFInfo
- Publication number
- CN115311684A CN115311684A CN202210938978.7A CN202210938978A CN115311684A CN 115311684 A CN115311684 A CN 115311684A CN 202210938978 A CN202210938978 A CN 202210938978A CN 115311684 A CN115311684 A CN 115311684A
- Authority
- CN
- China
- Prior art keywords
- rec
- value
- confidence
- map
- gray value
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
-
- 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
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- 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
- G06V10/34—Smoothing or thinning of the pattern; Morphological operations; Skeletonisation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an integration method of millimeter wave image multi-angle detection results, which comprises the steps of firstly calculating the maximum confidence of single detection; then generating a confidence map, and storing all detection results of a group of samples in the same image together in a mode of overlapping regional gray values, so as to integrate the detection results in batches; then calculating the maximum circumscribed rectangle of all connected domains in the confidence map; and sequentially calculating the highest/next-highest gray value in each maximum circumscribed rectangle, and performing integration treatment to obtain a final integration result. The invention effectively integrates the multi-angle detection result of the millimeter wave image of the human body by referring to the probability weight thought of the NMS algorithm and combining with the confidence interval theory of statistics.
Description
Technical Field
The invention belongs to the field of human body millimeter wave image target detection, and particularly relates to a method for integrating millimeter wave image multi-angle detection results.
Background
Millimeter wave image target detection is the key to realize the detection of contraband carried on the body surface, can be widely applied to security inspection work of airports, stations and other places, and is an effective substitute for the existing human security inspection means. The millimeter wave imaging is a precondition for realizing millimeter wave image target detection, and the surrounding imaging is a mainstream imaging mode at present, in which an upright human body is used as a rotation axis, an imaging element is rotated at a certain interval angle to generate a plurality of human body images (as shown in fig. 1, hereinafter referred to as a group of samples), and a group of detection results can be correspondingly obtained for a group of samples by using a millimeter wave image target detection technology.
For a contraband object, the detection result forms and positions of the contraband object at different angles are different; for the security inspection task based on the millimeter wave image target detection technology, the requirements for the inspection result are as follows: the number of the detection results is the same as the number of the contraband actually carried by the human body, and the position of the detection results is close to the actual carrying position of the contraband (relative to the front view of the human body, such as the 2 nd row and 2 nd column images in fig. 1). Therefore, it is necessary to integrate the detection results of the multiple angles.
Non-Maximum Suppression (NMS, neubeck a, gool l. Efficient Non-Maximum Suppression [ C ]// International Conference Pattern recognition. Ieee Computer Society, 2006.) is commonly used to integrate multiple detected results belonging to the same target in a target detection task, which requires that a deep learning technique be used to calculate the probability that each detected result contains a contraband target, and integrate the detected results with that probability as a weight. Because the probability weight coefficient is added, the integration effect of the NMS algorithm on the detection result is better, but the traditional target detection algorithm which does not use the deep learning technology cannot use the NMS algorithm; and the NMS algorithm only relates to the integration work of the multi-detection result of a single sample, but not relates to the integration work of the multi-angle sample.
Aiming at the problem, the invention effectively integrates the multi-angle detection result of the human body millimeter wave image by referring to the probability weight thought of the NMS algorithm and combining with the confidence interval theory of statistics.
Disclosure of Invention
The invention effectively integrates the multi-angle detection results of the human body millimeter wave images by referring to the probability weight thought of the NMS algorithm and combining with the confidence interval theory of statistics aiming at the problem that the multi-angle detection results (figure 2 is marked by a rectangular frame, and the total number of 35 in the example) of the human body millimeter wave images are difficult to be properly integrated by using the existing method.
A method for integrating multi-angle detection results of millimeter wave images comprises the following steps:
and storing all detection results of a group of samples in the same image together in a mode of overlapping regional gray values, thereby performing batch integration on the detection results.
Step 3, calculating the maximum circumscribed rectangles of all connected domains in the confidence map;
and 4, sequentially calculating the highest/next-highest gray value in each maximum circumscribed rectangle in the step 3, and performing integration processing to obtain a final integration result.
Further, the specific method in step 1 is as follows:
note the maximum confidence of a single detection as onseprabability, expressed in gray scale values. The maximum gray value which can be stored by the confidence map is maxProbability, and if n human body images are shared in a group of samples, the maximum confidence of single detection is represented as follows:
further, the specific method of step 2 is as follows:
and creating an image with the same size as the sample image and the initial gray value of 0 as a confidence Map, and recording the image as Map. Then Map is expressed as:
wherein i represents a sample number, i ∈ [1,n ]](ii) a j represents the number of the detection result in one sample, and j belongs to [0, + ∞); omega ij A jth detection result region representing the ith sample; omega ij | Value＝onceProbability The gray value of each detection result area is defined as the onceProavailability.
Thus, equation (2) can be interpreted as: the Map generation process is a process of superposing the onceprability as a gray value at the corresponding position in the Map in all the detection result areas in the group of samples.
Further, the specific method of step 3 is as follows;
an interconnected region with non-zero values is called a connected domain, and the connected domain in the Map is formed by superposing the formula (2) and represents a region containing contraband. A Map may have 0 to a number of connected domains, i.e., the number of contraband contained in the set of samples.
One connected domain can be circumscribed by one rectangle only, and the rectangle is called the largest circumscribed rectangle of the connected domain and is marked as Rec k The lower subscript k denotes the number in Map of the largest circumscribed rectangle of the current connected component. Record Rec k The processing of step 4 is performed for each Rec.
Further, the specific method in step 4 is as follows:
Rec k and representing the confidence degree in the two-dimensional confidence interval by using the image gray value, wherein the two-dimensional confidence interval is related to the contraband target. The higher the confidence, the closer the true form and location of the contraband target is to the form and location of the confidence interval. The purpose of step 4 is therefore to calculate Rec k And estimating the real form and position of the contraband target according to the form and position of the confidence interval.
Firstly, traversing the pixel points in the current Rec to obtain the highest gray Value, and marking as Value Peak . And note that the gray Value in the current Rec is equal to Value Peak Has a region of Ω Peak . Note that the next highest gray Value of the current Rec is Value SubPeak As shown in the formula (2), value SubPeak And Value Peak The following relationships exist:
Value SubPeak ＝Value Peak -onceProbability (3)
note that the gray Value in current Rec is equal to Value SubPeak Has a region of Ω SubPeak 。
Secondly, to prevent the size of the integration result from being too small relative to the real contraband target, a degradation threshold k is introduced dg . Record omega Peak Has an area of S Peak I.e. the gray Value in the current Rec is equal to Value Peak The number of the pixel points can be obtained into omega in the same way SubPeak Area S of SubPeak Then k is dg Expressed as:
when k is dg When the content is less than or equal to 0.5, judging that S exists Peak ＜＜S SubPeak In this case, let Ω Peak Degenerated to omega SubPeak I.e. omega Peak The grey values of all pixels are reduced by oneProbasic.
Finally, the gray Value in the current Rec is lower than Value Peak The gray Value of the pixel point is set to zero, and only the gray Value in Rec is kept to be equal to Value Peak And calculating the number of connected domains in Rec and recording as p. The mathematical morphological dilation operation enables to communicate adjacent connected domains, noting a rectangular structural element with dimensions p × p and origin at the center as B, using the morphological dilation p times of the connected domains within B pairs Rec to guarantee the integrity of the integration result. And similarly, calculating the maximum circumscribed rectangle of the connected domain in the current Rec by the method in the step 3, wherein the rectangular area is the integration result of the multi-angle detection results of the contraband target in the current Rec. For Rec in Map k And (4) processing according to the step 4 in sequence to obtain an integrated result of the multi-angle detection results of a group of samples.
The invention has the following beneficial effects:
aiming at the problem that the multi-angle detection results (shown by red rectangular boxes in figure 2, the total number of 35 in the example) of the human body millimeter wave images are difficult to be properly integrated by using the existing method, the multi-angle detection results of the human body millimeter wave images are effectively integrated by referring to the probability weight thought of an NMS (network management system) algorithm and combining with the confidence interval theory of statistics.
Drawings
FIG. 1 is an imaging schematic diagram of a millimeter wave human body imaging device imaging once per 18 degrees of rotation in a 180 degree rotation region;
FIG. 2 is a graph showing the results of a set of samples (FIG. 1);
FIG. 3 is a schematic diagram illustrating a process of generating confidence that contraband is generated at the ankle of the left side of the human body of FIG. 1;
FIG. 4 is a schematic illustration of a confidence process for the example of contraband at the left ankle of the human body of FIG. 1.
Detailed Description
The method of the present invention is further illustrated with reference to the following figures and examples.
A method for integrating multi-angle detection results of millimeter wave images specifically comprises the following steps:
note the maximum confidence of a single detection as onseprabability, expressed in gray scale values. Noting that the maximum gray value (determined by the image file format) that the confidence map can store is maxProbability, noting that there are n human body images in a group of samples (i.e. the imaging device images n times in the range of 180 °), the maximum confidence of a single detection is represented as:
for example, if the confidence map format is an 8-bit unsigned single channel image, with a total of 9 images in a set of samples, then there are:
and creating an image with the initial gray value of 0 and the same size as the sample image as a confidence Map, and marking the image as Map. Then Map is expressed as:
wherein i represents a sample number, i ∈ [1,n ]](ii) a j represents the number of the detection result in one sample, and j belongs to [0, + ∞); omega ij A jth detection result region representing the ith sample; omega ij | Value＝onceProbability The gray value of each detection result area is defined as the onseproavailability.
Thus, equation (3) can be interpreted as: and the Map generation process is a process of superposing corresponding positions of all detected result areas in a group of samples by using the onacephobicity as a gray value in the Map.
Fig. 3 is a schematic diagram of a partial Map generation process corresponding to a partial region at the ankle of the left side of the human body in fig. 1, wherein the single detection results of the 1 st, 2 nd and 9 th samples are marked by solid boxes, and the accumulated confidence region at the position corresponding to the Map is marked by a dashed box.
Step 3, calculating the maximum circumscribed rectangles of all connected domains in the confidence map;
an interconnected region with non-zero values is called a connected domain, and the connected domain in the Map is formed by superposing the formula (3) and represents a region containing contraband. A Map may have 0 to a number of connected domains, i.e., the number of contraband contained in the set of samples.
As shown by the dashed box in FIG. 4 (a), a connected component can be circumscribed by only one rectangle, which is called the largest circumscribed rectangle of the connected component, and is denoted as Rec k The lower subscript k denotes the number in Map of the largest circumscribed rectangle of the current connected domain. Record Rec k The processing of step 4 is performed for each Rec.
And 4, sequentially calculating the highest/next-highest gray value in each maximum circumscribed rectangle in the step 3, and performing integration processing.
Rec k And representing the confidence coefficient in the two-dimensional confidence interval by using an image gray value as the two-dimensional confidence interval related to the contraband target. The higher the confidence, the contraband targetThe closer the true morphology and location are to the morphology and location of the confidence interval. The purpose of step 4 is therefore to calculate Rec k And estimating the real form and position of the contraband target according to the form and position of the confidence interval.
Firstly, traversing the pixel points in the current Rec to obtain the highest gray Value, and marking as Value Peak . And note that the gray Value in the current Rec is equal to Value Peak Has a region of Ω Peak ，Ω Peak In fig. 4 (a), the left oblique lines are marked. Note that the next highest gray Value of the current Rec is Value SubPeak As shown in the formula (3), value SubPeak And Value Peak The following relationships exist:
Value SubPeak ＝Value Peak -onceProbability (4)
note that the gray Value in the current Rec is equal to Value SubPeak Has a region of Ω SubPeak ，Ω SubPeak In FIG. 4 (a), the grid lines are shown.
Secondly, to prevent the size of the integration result from being too small relative to the real contraband target, a degradation threshold k is introduced dg . Record omega Peak Has an area of S Peak I.e. the gray Value in the current Rec is equal to Value Peak The number of the pixels can be obtained to be omega in the same way SubPeak Area S of SubPeak Then k is dg Expressed as:
when k is dg When the concentration is less than or equal to 0.5, judging that S exists Peak ＜＜S SubPeak In this case, let Ω Peak Degenerated to omega SubPeak I.e. omega Peak The gray values of all pixels are reduced by oneprobability.
Finally, the gray Value in the current Rec is lower than Value Peak The gray Value of the pixel point is set to zero, and only the gray Value in Rec is kept to be equal to Value Peak And calculating the number of connected domains in Rec and marking as p. Mathematical morphology (Serra J. Image Analysis and chemical Morpho)log-Volume i.academic, 1982.) inflation operation enables to connect adjacent connected domains, taking the rectangular structural element with size p × p and origin at the center as B, using B to morphologically inflate connected domains within Rec p times to ensure integrity of the integrated result. And (c) calculating the maximum circumscribed rectangle of the connected domain in the current Rec by the method in the step (3) (a solid frame in FIG. 4 (c)), wherein the rectangular region is the integration result of the multi-angle detection results of the contraband target in the current Rec, and the solid frame in FIG. 4 (d) is the integration result. For Rec in Map k And (4) processing according to the step 4 in sequence to obtain an integrated result of the multi-angle detection results of a group of samples.
Claims (5)
1. A method for integrating multi-angle detection results of millimeter wave images is characterized by comprising the following steps:
step 1, calculating the maximum confidence of single detection;
step 2, generating a confidence map;
all detection results of a group of samples are stored in the same image together in a mode of overlapping regional gray values, so that the detection results are integrated in batches;
step 3, calculating the maximum circumscribed rectangles of all connected domains in the confidence map;
and 4, sequentially calculating the highest/next-highest gray value in each maximum circumscribed rectangle in the step 3, and performing integration treatment to obtain a final integration result.
2. The method for integrating the multi-angle detection results of millimeter wave images as claimed in claim 1, wherein the specific method in step 1 is as follows:
recording the maximum confidence of single detection as the onceProba availability, and expressing the maximum confidence as a gray value; the maximum gray value which can be stored by the confidence map is maxProbability, and if n human body images are recorded in a group of samples, the maximum confidence of single detection is represented as follows:
3. the method for integrating the multi-angle detection results of millimeter wave images as claimed in claim 2, wherein the specific method in step 2 is as follows:
creating an image which has the same size as the sample image and has an initial gray value of 0 as a confidence Map, and marking the image as a Map; then Map is expressed as:
wherein i represents a sample number, i ∈ [1,n ]](ii) a j represents the number of the detection result in one sample, and j belongs to [0, + ∞ ]; omega ij A jth detection result region representing the ith sample; omega ij | Value＝onceProbability Defining the gray value of each detection result area as the onceProba availability;
therefore, equation (2) can be interpreted as: and the Map generation process is a process of superposing corresponding positions of all detected result areas in a group of samples by using the onacephobicity as a gray value in the Map.
4. The method for integrating multi-angle detection results of millimeter wave images as claimed in claim 3, wherein the specific method in step 3 is as follows;
a region which has non-zero values and is communicated with each other is called a communicated domain, the communicated domains in the Map are superposed by a formula (2) and represent a region containing contraband; one Map may have 0 to a plurality of connected domains, and the number of the connected domains is the number of the contraband contained in the group of samples;
one connected domain can be circumscribed by only one rectangle, and the rectangle is called the maximum circumscribed rectangle of the connected domain and is marked as Rec k The lower subscript k represents the number of the maximum circumscribed rectangle of the current connected domain in the Map; record Rec k The processing of step 4 is performed for each Rec.
5. The method for integrating the multi-angle detection results of millimeter wave images as claimed in claim 4, wherein the specific method in step 4 is as follows:
Rec k representing the confidence degree in the two-dimensional confidence interval by using an image gray value, wherein the two-dimensional confidence interval is related to the contraband target; the higher the confidence coefficient is, the closer the real form and position of the contraband target are to the form and position of the confidence coefficient interval; the purpose of step 4 is therefore to calculate Rec k Estimating the real form and position of the contraband target according to the form and position of the confidence interval;
firstly, traversing the pixel points in the current Rec to obtain the highest gray Value, and marking as Value Peak (ii) a And note that the gray Value in the current Rec is equal to Value Peak Has a region of Ω Peak (ii) a Note that the next highest gray Value of the current Rec is Value SubPeak As shown in the formula (2), value SubPeak And Value Peak The following relationships exist:
Value SubPeak ＝Value Peak -onceProbability (3)
note that the gray Value in the current Rec is equal to Value SubPeak Has a region of Ω SubPeak ；
Secondly, to prevent the size of the integration result from being too small relative to the real contraband target, a degradation threshold k is introduced dg (ii) a Record omega Peak Has an area of S Peak I.e. the gray Value in the current Rec is equal to Value Peak The number of the pixels can be obtained to be omega in the same way SubPeak Area S of SubPeak Then k is dg Expressed as:
when k is dg When the content is less than or equal to 0.5, judging that S exists Peak ＜＜S SubPeak In this case, let Ω Peak Degenerated to omega SubPeak I.e. omega Peak The gray values of all pixels are reduced by oneProavailability;
finally, the gray Value in the current Rec is lower than Value Peak The gray Value of the pixel point is set to zero, and only the gray Value in Rec is kept to be equal to Value Peak Calculating the number of connected domains in Rec and recording as p; the mathematical morphological dilation operation can communicate adjacent connected domains, the size is p multiplied by p, a rectangular structural element with an origin at the center is B, and the connected domains in the Rec are morphologically dilated p times by using the B pair to ensure the integrity of an integration result; similarly, calculating the maximum circumscribed rectangle of the connected domain in the current Rec by the method in the step 3, wherein the rectangular area is the integration result of the multi-angle detection results of the contraband target in the current Rec; for Rec in Map k And (4) processing according to the step 4 in sequence to obtain an integrated result of the multi-angle detection results of a group of samples.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210938978.7A CN115311684B (en) | 2022-08-05 | 2022-08-05 | Method for integrating multi-angle detection results of millimeter wave images |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210938978.7A CN115311684B (en) | 2022-08-05 | 2022-08-05 | Method for integrating multi-angle detection results of millimeter wave images |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115311684A true CN115311684A (en) | 2022-11-08 |
| CN115311684B CN115311684B (en) | 2023-03-28 |
Family
ID=83861531
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210938978.7A Active CN115311684B (en) | 2022-08-05 | 2022-08-05 | Method for integrating multi-angle detection results of millimeter wave images |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115311684B (en) |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106291738A (en) * | 2016-10-11 | 2017-01-04 | 深圳万发创新进出口贸易有限公司 | A kind of safe examination system on bus |
| CN106371148A (en) * | 2016-09-27 | 2017-02-01 | 华讯方舟科技有限公司 | Millimeter wave image-based human body foreign substance detection method and system |
| CN106529602A (en) * | 2016-11-21 | 2017-03-22 | 中国科学院上海微系统与信息技术研究所 | Automatic millimeter wave image target identification method and device |
| CN109325490A (en) * | 2018-09-30 | 2019-02-12 | 西安电子科技大学 | Terahertz image target identification method based on deep learning and RPCA |
| CN109799544A (en) * | 2018-12-28 | 2019-05-24 | 深圳市华讯方舟太赫兹科技有限公司 | Intelligent detecting method, device and storage device applied to millimeter wave safety check instrument |
| CN110059640A (en) * | 2019-04-22 | 2019-07-26 | 长光卫星技术有限公司 | The in-orbit recognition methods of sea ship based on Optical remote satellite near-infrared spectral coverage |
| CN110426745A (en) * | 2019-01-30 | 2019-11-08 | 西安电子科技大学 | The millimeter-wave image foreign matter detecting method decomposed based on block mixed Gaussian low-rank matrix |
| CN111046877A (en) * | 2019-12-20 | 2020-04-21 | 北京无线电计量测试研究所 | Millimeter wave image suspicious article detection method and system |
| CN111260607A (en) * | 2019-12-23 | 2020-06-09 | 北京无线电计量测试研究所 | Automatic suspicious article detection method, terminal device, computer device and medium |
| CN112016387A (en) * | 2019-07-08 | 2020-12-01 | 杭州芯影科技有限公司 | Contraband identification method and device suitable for millimeter wave security check instrument |
| CN113487642A (en) * | 2021-07-08 | 2021-10-08 | 杭州电子科技大学 | Method for detecting in-vitro target by using millimeter waves for significance vision |
| US11231498B1 (en) * | 2020-07-21 | 2022-01-25 | International Business Machines Corporation | Concealed object detection |
| CN114581950A (en) * | 2022-03-08 | 2022-06-03 | 博微太赫兹信息科技有限公司 | Millimeter wave image target detection method and system |
-
2022
- 2022-08-05 CN CN202210938978.7A patent/CN115311684B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106371148A (en) * | 2016-09-27 | 2017-02-01 | 华讯方舟科技有限公司 | Millimeter wave image-based human body foreign substance detection method and system |
| CN106291738A (en) * | 2016-10-11 | 2017-01-04 | 深圳万发创新进出口贸易有限公司 | A kind of safe examination system on bus |
| CN106529602A (en) * | 2016-11-21 | 2017-03-22 | 中国科学院上海微系统与信息技术研究所 | Automatic millimeter wave image target identification method and device |
| CN109325490A (en) * | 2018-09-30 | 2019-02-12 | 西安电子科技大学 | Terahertz image target identification method based on deep learning and RPCA |
| CN109799544A (en) * | 2018-12-28 | 2019-05-24 | 深圳市华讯方舟太赫兹科技有限公司 | Intelligent detecting method, device and storage device applied to millimeter wave safety check instrument |
| CN110426745A (en) * | 2019-01-30 | 2019-11-08 | 西安电子科技大学 | The millimeter-wave image foreign matter detecting method decomposed based on block mixed Gaussian low-rank matrix |
| CN110059640A (en) * | 2019-04-22 | 2019-07-26 | 长光卫星技术有限公司 | The in-orbit recognition methods of sea ship based on Optical remote satellite near-infrared spectral coverage |
| CN112016387A (en) * | 2019-07-08 | 2020-12-01 | 杭州芯影科技有限公司 | Contraband identification method and device suitable for millimeter wave security check instrument |
| CN111046877A (en) * | 2019-12-20 | 2020-04-21 | 北京无线电计量测试研究所 | Millimeter wave image suspicious article detection method and system |
| CN111260607A (en) * | 2019-12-23 | 2020-06-09 | 北京无线电计量测试研究所 | Automatic suspicious article detection method, terminal device, computer device and medium |
| US11231498B1 (en) * | 2020-07-21 | 2022-01-25 | International Business Machines Corporation | Concealed object detection |
| CN113487642A (en) * | 2021-07-08 | 2021-10-08 | 杭州电子科技大学 | Method for detecting in-vitro target by using millimeter waves for significance vision |
| CN114581950A (en) * | 2022-03-08 | 2022-06-03 | 博微太赫兹信息科技有限公司 | Millimeter wave image target detection method and system |
Non-Patent Citations (3)
| Title |
|---|
| YUAN M, ET AL.: "A Suspicious Multi-Object Detection and Recognition Method for Millimeter Wave SAR Security Inspection Images Based on Multi-Path Extraction Network", 《REMOTE SENSING》 * |
| 姚家雄,等: "利用卷积神经网络进行毫米波图像违禁物体定位", 《红外与毫米波学报》 * |
| 王崇剑,等: "一种用于主动式毫米波图像的低复杂度隐匿物品检测方法", 《红外与毫米波学报》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115311684B (en) | 2023-03-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ahmed | Comparative study among Sobel, Prewitt and Canny edge detection operators used in image processing | |
| Fan et al. | Dim infrared image enhancement based on convolutional neural network | |
| Younus et al. | Effective and fast deepfake detection method based on haar wavelet transform | |
| CN112508013B (en) | Lock catch loss fault detection method, system and device | |
| Lee et al. | Edge detection analysis | |
| CN112215780B (en) | Image evidence obtaining and resistance attack defending method based on class feature restoration fusion | |
| Xu et al. | COCO-Net: A dual-supervised network with unified ROI-loss for low-resolution ship detection from optical satellite image sequences | |
| CN111259792A (en) | Face living body detection method based on DWT-LBP-DCT characteristics | |
| Babu et al. | An efficient image dahazing using Googlenet based convolution neural networks | |
| Saealal et al. | Three-Dimensional Convolutional Approaches for the Verification of Deepfake Videos: The Effect of Image Depth Size on Authentication Performance | |
| Wang et al. | A multiscale gradient fusion method for edge detection in color images utilizing the cbm3d filter | |
| CN106778822B (en) | Image straight line detection method based on funnel transformation | |
| CN115311684B (en) | Method for integrating multi-angle detection results of millimeter wave images | |
| CN108961322B (en) | Mismatching elimination method suitable for landing sequence images | |
| Chi et al. | Single image dehazing using a novel histogram tranformation network | |
| CN111652906B (en) | Adaptive tracking method, device and equipment for satellite video ground dynamic target rotation | |
| Narasimharao et al. | Advanced Techniques for Color Image Blind Deconvolution to Restore Blurred Images | |
| CN112085683A (en) | Depth map reliability detection method in significance detection | |
| Menaka et al. | Wavelet transform-based land cover classification of satellite images | |
| Balamurugan et al. | Classification of Land Cover in Satellite Image using supervised and unsupervised Techniques | |
| Nair et al. | Single Image Dehazing Using Multi-Scale DCP-BCP Fusion | |
| Lou et al. | Research and analysis of infrared image enhancement algorithm based on fractional differentiation | |
| Yao et al. | Shadow removal from images using an improved single-scale retinex color restoration algorithm | |
| Arvind et al. | Bit-plane specific selective histogram equalization for image enhancement and representation | |
| Luong et al. | Illumination Compensation of Facial Image Using Combination Algorithm for Face Recognition |
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 |