CN111445522B - Passive night vision intelligent lightning detection system and intelligent lightning detection method - Google Patents

Passive night vision intelligent lightning detection system and intelligent lightning detection method Download PDF

Info

Publication number
CN111445522B
CN111445522B CN202010168178.2A CN202010168178A CN111445522B CN 111445522 B CN111445522 B CN 111445522B CN 202010168178 A CN202010168178 A CN 202010168178A CN 111445522 B CN111445522 B CN 111445522B
Authority
CN
China
Prior art keywords
cast
night vision
mine
land
network model
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.)
Active
Application number
CN202010168178.2A
Other languages
Chinese (zh)
Other versions
CN111445522A (en
Inventor
王驰
于明坤
李思远
李富迪
孙建美
朱俊
栾信群
方东
李金辉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
63983 Troops of PLA
University of Shanghai for Science and Technology
Original Assignee
63983 Troops of PLA
University of Shanghai for Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 63983 Troops of PLA, University of Shanghai for Science and Technology filed Critical 63983 Troops of PLA
Priority to CN202010168178.2A priority Critical patent/CN111445522B/en
Publication of CN111445522A publication Critical patent/CN111445522A/en
Application granted granted Critical
Publication of CN111445522B publication Critical patent/CN111445522B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
    • G06T7/75—Determining position or orientation of objects or cameras using feature-based methods involving models
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/20—Scenes; Scene-specific elements in augmented reality scenes

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Multimedia (AREA)
  • Image Processing (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a passive night vision intelligent lightning detection system and an intelligent lightning detection method, wherein a fixed focus lens and a low-light camera are connected through an F bayonet, the low-light camera is arranged on a support frame, an FPGA development board is connected with the low-light camera through a data transmission line, the FPGA development board is connected with a display screen through a data line, a lead storage battery is sequentially connected with the low-light camera, the FPGA development board and the display screen through data lines, and all parts are arranged on an instrument loading vehicle; constructing a passive night vision intelligent mine detection network model by using a YOLOv2 algorithm, constructing a data set of a landmine, training, evaluating and optimizing the passive night vision intelligent mine detection network, and embedding the trained passive night vision intelligent mine detection network model into an FPGA development board; a passive night vision intelligent exploratory ranging model is built by using the imaging principle of geometrical optics and is embedded into an FPGA development board; the invention realizes intelligent and rapid detection of the cast land mines at night.

Description

Passive night vision intelligent lightning detection system and intelligent lightning detection method
Technical Field
The invention relates to a land mine detection system and a land mine detection method, in particular to a night vision intelligent mine detection system and a mine detection method, which are applied to the technical field of cast land mine detection.
Background
Modern warfare is very focused on rapid maneuvers and requires all-weather observation and monitoring mechanisms. Rapid detection of the cast land mines at night is one of the important factors to ensure rapid maneuver of the war at night. At present, two main night detection technologies for land mine casting are available: imaging technology detection and non-imaging technology detection. Non-imaging technology such as pulse radar, obtains the radiation signal of the landmine by an active and passive working mode, and detects the target signal by utilizing a signal processing technology and a pattern recognition technology, but the technology is greatly influenced by the surrounding environment of the landmine and complicated in communication transmission, so that the detection difficulty of the landmine scattered at night is high; imaging technologies such as infrared, radar, satellite remote sensing and the like are easily affected by noise of surrounding environments of the landmine, and detection accuracy is low.
The passive night vision intelligent mine detection system and the mine detection method show good detection prospect, but a commercial or special passive night vision intelligent mine detection system is not available at present, and a typical passive night vision intelligent mine detection system comprises an imaging part, a data transmission part and a ground mine throwing night detection part. At present, commercial products exist in a night low-light imaging part, but the combination of the night intelligent detection method for the land mine throwing and the night intelligent detection method for the land mine throwing is absent; the night detection part for throwing the land mine is not used commercially or is not used for special products. The invention aims at solving the key problems of poor imaging quality, low detection accuracy, low detection speed and the like during the detection of the ground mines thrown at night.
Disclosure of Invention
In order to solve the problems in the prior art, the invention aims to overcome the defects in the prior art and provide a passive night vision intelligent mine detection system and an intelligent mine detection method, which are characterized in that a night vision camera is used for collecting night vision images of a cast mine, a YOLO v2 algorithm is used for constructing a passive night vision intelligent mine detection network model, a geometrical optical imaging principle is used for measuring the cast mine, and quick and intelligent detection of the cast mine at night is realized.
In order to achieve the aim of the invention, the invention adopts the following technical scheme:
a passive night vision intelligent lightning detection system comprises a low-light camera, a fixed focus lens, a Camellia data transmission line, an FPGA development board, a display screen, a support frame, a lead storage battery and an instrument loading vehicle; the fixed focus lens is connected with the low-light-level camera through an F bayonet, the low-light-level camera is arranged on the support frame, the FPGA development board is connected with the low-light-level camera through the Cameralink data transmission line, the FPGA development board is connected with the display screen through a data line, the lead storage battery supplies power for the low-light-level camera, the FPGA development board and the display screen, and the support frame, the FPGA development board, the display screen and the lead storage battery are arranged on the instrument loading vehicle; the micro-camera acquires night vision images of the landmine through the fixed focus lens, transmits image data to the FPGA development board through the Cameralink data transmission line, and the FPGA development board identifies the detected night vision images of suspected targets of the landmine, judges whether the night vision images are the landmine, measures the landmine through data processing, obtains distance data of the landmine and the micro-camera, and displays detection results and distance measurement results on the display screen.
As the preferable technical scheme of the invention, the passive night vision intelligent mine detection system utilizes a passive night vision intelligent mine detection network model to identify night vision images of the cast mines and judge whether the cast mines exist, and the construction and use method of the passive night vision intelligent mine detection network model comprises the following specific implementation steps:
1) Using the micro-light camera to shoot and collect at least 3000 night vision images of the landmine, and manufacturing a data set of the landmine according to a PASCAL VOC2007 format;
2) Constructing a passive night vision intelligent mine detection network model by utilizing a YOLO v2 algorithm, adopting DarkNet-19 as a basic network structure, training the passive night vision intelligent mine detection network model by utilizing the data set of the cast land mine generated in the step 1), evaluating the passive night vision intelligent mine detection network model by adopting a loss function, and storing a weight model when the loss value of the loss function is reduced to 0.05, wherein the loss function formula is as follows:
Figure BDA0002408207130000021
where loss is the loss function value, lambda coord For locating the weighting coefficient of the prediction error, lambda noobj For classifying the weight coefficient of the prediction error, S 2 Representing the division of the night vision image of the cast mine into S x S boxes, B representing the number of bounding boxes in each box,
Figure BDA0002408207130000022
judging whether an ith grid and a jth boundary box in the night vision image are responsible for throwing the land mine or not; (x) i ,y i ) Center coordinates of an object boundary frame predicted by a passive night vision intelligent mine detection network model are indicated,/-or%>
Figure BDA0002408207130000023
Center coordinates of an actual bounding box for the cast land mine, (w i ,h i ) For the length and width of the bounding box of the ith cell,
Figure BDA0002408207130000024
the length and the width of the actual bounding box of the cast land mine are the same; />
Figure BDA0002408207130000025
Judging whether the center of the cast land mine falls in the ith grid or not, C i And->
Figure BDA0002408207130000026
The confidence of the boundary frame predicted by the passive night vision intelligent mine detection network model and the confidence of the actual boundary frame; p is p i (c) Class probability representing passive night vision intelligent mine detection network model prediction, < >>
Figure BDA0002408207130000027
Representing actual class probabilities;
3) And (2) evaluating the passive night vision intelligent mine detection network model constructed in the step (2) by adopting a recall rate and accuracy as network model evaluation standards, wherein the recall rate refers to the proportion of the correctly detected cast land mines to all cast land mines in the verification set, and the formula is as follows:
Figure BDA0002408207130000031
accuracy refers to the proportion of the correctly detected cast land mines to the detected cast land mines, and the formula is as follows:
Figure BDA0002408207130000032
wherein Recall is the Recall rate; precision is Precision; tureMines indicates that the target is the cast mine and is correctly detected as the cast mine; tureNMines indicates that the target is not the cast mine and is not misdetected as the cast mine; falseMines indicates that the target is not the cast mine, but is misdetected as the cast mine; false nmines means that the target is the cast mine but is not detected as the cast mine;
4) The method for optimizing the passive night vision intelligent lightning detection network model comprises the following specific implementation steps:
a) Pruning the passive night vision intelligent lightning detection network model, setting a threshold value, deleting the connection with lower weight, converting the dense network into a sparse network, retraining the weight of the rest sparse connection in the network model, learning the correct connection in the network parameters with continuous sparsity, and finding the minimum number of connections after multiple iterations;
b) The method comprises the steps of performing quantization processing on a passive night vision intelligent lightning detection network model by using weight sharing and weight reduction methods;
5) If the optimized passive night vision intelligent lightning detection network model reaches the network model evaluation standard, the passive night vision intelligent lightning detection network model is stored; if the optimized passive night vision intelligent lightning detection network model does not reach the network model evaluation standard, performing step 6);
6) Repeating the steps 4) to 5).
As a further preferable technical solution of the passive night vision intelligent mine detection system, in the step 1), a night vision image of the cast mine is photographed and collected by using the low-light camera, preferably 3000 night vision images of the cast mine are photographed and collected by using the low-light camera, in the collecting process, the exposed area of the cast mine changes from maximum to minimum, for each exposed area, the cast mine changes from 0 degree to 360 degrees clockwise at intervals of 10 degrees on a plane parallel to the ground, and changes from 0 degree to 360 degrees anticlockwise at intervals of 10 degrees on a plane perpendicular to the ground, so as to ensure the completeness of a sample to the maximum extent, after the acquisition of the cast mine night vision image is completed, a data set of the cast mine is manufactured according to a PASCAL VOC2007 format, and the data set of the cast mine comprises 4 folders:
an items folder, under which tag files in xml format are stored, each xml file corresponding to one of the cast land mine night vision pictures in the JPEGImages folder;
the JPEGImages folder is used for storing the data set pictures of the landmine, including training pictures and test pictures;
the imagesets folder is provided with a folder Main, and two files of a train. Txt and a train. Txt are mainly stored under the Main folder, so that the training set and the testing set are divided;
and d, a labels folder which is a file obtained by converting the file to adapt to a YOLO v2 algorithm, wherein each xml file in the above-mentioned anotions folder is converted into a corresponding txt label, and two files of tranval. Txt and val. Txt are generated.
As a preferable technical scheme of the invention, a passive night vision intelligent mine detection system is identified by adopting a passive night vision intelligent mine detection network model, and when the cast mine is identified, the distance of the cast mine is measured by adopting a passive night vision intelligent mine detection ranging model and utilizing the FPGA development board, and the construction and use method of the passive night vision intelligent mine detection ranging model comprises the following specific implementation steps:
i, converting the pixel coordinate values of the cast land mines into actual coordinate values according to a geometrical optical imaging method, and determining the actual distance from the cast land mines to the micro-light camera, wherein a calculation formula of the vertical distance between the cast land mines and the micro-light camera is as follows:
Figure BDA0002408207130000041
Figure BDA0002408207130000042
y＝H·tan(α+Δθ) (6)
wherein Height is the night vision image Height of the cast mine, (u, v) is the pixel coordinate value of the cast mine, (x, y) is the actual coordinate value of the cast mine, θ is the vertical field angle of the low-light camera, and Δθ is the step angle; the included angle formed by the pixel coordinates of the throwing land mine and the maximum pixel coordinates is H, the ground clearance height of the low-light camera, and alpha is the pitching angle of the low-light camera;
according to the geometrical optical imaging method and the actual vertical distance y between the landmine and the low-light camera calculated in the step I, converting the horizontal coordinates of the image pixels into the actual horizontal distance between the landmine and the low-light camera, wherein the calculation formula is as follows:
Figure BDA0002408207130000043
Figure BDA0002408207130000044
Figure BDA0002408207130000045
wherein l is the actual distance from an actual coordinate (0, y) to the micro-camera, H is the ground clearance height of the micro-camera, (u, v) is the pixel coordinate value of the cast mine, (x, y) is the actual coordinate value of the cast mine, b is the field radius at the y coordinate value corresponding to the cast mine, beta is the horizontal field angle of the micro-camera, and Width is the night vision image Width of the cast mine.
The invention relates to a passive night vision intelligent lightning detection method, which adopts a passive night vision intelligent lightning detection system and comprises the following specific implementation steps:
a. embedding a passive night vision intelligent mine detection network model and a passive night vision intelligent mine detection ranging model into the FPGA development board;
b. conveying the passive night vision intelligent mine detection system to a ground area of the ground to be detected at night for standby;
c. adjusting the focal length of the prime lens to ensure that the prime lens is positioned at the front 18m of the micro-light camera Jiao Zaisuo;
d. moving the low-light camera by using the instrument loading vehicle, shooting and collecting night vision images of the landmine;
e. transmitting the acquired night vision image of the landmine to the FPGA development board in real time through the Cameralink data transmission line;
f. identifying the night vision image of the landmine by using the FPGA development board, and judging whether the landmine is thrown or not;
g. if the landmine is identified, measuring and calculating the distance by using the FPGA development board, and displaying a detection result and a ranging result on the display screen; if the landmine is not identified, carrying out the following step h;
h. repeating the operations of the steps d to g.
Compared with the prior art, the invention has the following obvious prominent substantive features and obvious advantages:
1. the passive night vision intelligent mine detection system and the mine detection method show good detection prospect, wherein the imaging part adopts a low-light camera and a fixed focus lens, and the images of the landmine are shot and collected at night, so that the concealment is strong and the speed is high;
2. the cast land mine detection part adopts the YOLO v2 algorithm to construct a passive night vision intelligent mine detection network model, so that the intelligent detection of the cast land mine is realized, and the network model has strong stability and high accuracy; the passive night vision intelligent mine detection ranging part utilizes a geometrical optical imaging principle to finish mapping from pixel coordinates to actual coordinates, so that the actual distance between a cast mine and a low-light camera is measured;
3. the invention realizes intelligent and rapid detection of the cast land mines at night; the method is simple and easy to implement, has low device cost, is suitable for being used under various conditions, and has individual detection capability.
Drawings
Fig. 1 is a schematic diagram of a system for a rapid detection device for a mine thrown at night according to a preferred embodiment of the invention.
FIG. 2 is a block diagram of the steps for constructing a passive night vision intelligent lightning detection network model according to the preferred embodiment of the invention.
FIG. 3 is a schematic diagram of a passive night vision intelligent mine detection vertical distance measurement model according to a preferred embodiment of the present invention.
FIG. 4 is a schematic diagram of a passive night vision intelligent mine detection horizontal distance measurement model according to the preferred embodiment of the invention.
FIG. 5 is a block diagram showing the steps of a method for detecting and throwing a mine at night by the passive night vision intelligent mine detection system according to the preferred embodiment of the invention.
Detailed Description
The foregoing aspects are further described in conjunction with specific embodiments, and the following detailed description of preferred embodiments of the present invention is provided:
embodiment one:
in this embodiment, referring to fig. 1, a passive night vision intelligent lightning detection system includes a micro-camera 101, a fixed focus lens 102, a camera link data transmission line 103, an FPGA development board 104, a display screen 105, a support 106, a lead storage battery 107, and an instrument loading vehicle 108; the fixed focus lens 102 is connected with the micro-light camera 101 through an F bayonet, the micro-light camera 101 is placed on the supporting frame 106, the FPGA development board 103 is connected with the micro-light camera 101 through the Cameralink data transmission line 103, the FPGA development board 103 is connected with the display screen 105 through a data line, the lead storage battery 107 supplies power for the micro-light camera 101, the FPGA development board 104 and the display screen 105, and the supporting frame 106, the FPGA development board 104, the display screen 105 and the lead storage battery 107 are placed on the instrument loading vehicle 108; the micro-camera 101 collects night vision images of the land mines 109 through the fixed focus lens 102, transmits image data to the FPGA development board 104 through the camera link data transmission line 103, and the FPGA development board 104 identifies the detected night vision images of suspected targets of the land mines 109, judges whether the night vision images are the land mines 109, measures the distance of the land mines 109 through data processing, obtains distance data of the land mines 109 and the micro-camera 101, and displays detection results and distance measurement results on the display screen 105.
The micro-light camera used in this embodiment is a pco.edge4.2 high-speed camera manufactured by germany meta-aoku instruments, the fixed focus lens is a D-type AF nike lens manufactured by nikka corporation, the used camera link data transmission line is manufactured by CEI corporation, the used display screen is an Hp N246V monitor model display screen manufactured by hewlett packard corporation, the used mobile power supply is a 220V lead battery manufactured by beijing nikka corporation, and the used FPGA development board is sold by beijing peak technology limited.
In this embodiment, referring to fig. 2, the passive night vision intelligent mine detection system uses a passive night vision intelligent mine detection network model to identify a night vision image of the cast land mine 109, and determine whether the cast land mine 109 exists, and the construction and use method of the passive night vision intelligent mine detection network model specifically includes the following implementation steps:
1) The night vision image 3000 of the land mine throwing 109 is shot and acquired by using the low-light camera 101, in the acquisition process, the exposed area of the land mine throwing 109 is changed from the maximum to the minimum, for each exposed area, the land mine throwing 109 is changed from 0 degree to 360 degrees clockwise at intervals of 10 degrees on a plane parallel to the ground, and on a plane perpendicular to the ground, the included angle between the land mine throwing and the ground is changed from 0 degree to 360 degrees anticlockwise at intervals of 10 degrees, so that the completeness of a sample is ensured to the maximum, after the acquisition of the night vision image of the land mine throwing 109 is completed, a data set of the land mine throwing 109 is manufactured according to the PASCAL VOC2007 format, and the data set of the land mine throwing 109 comprises the following 4 folders:
an items folder, under which tag files in xml format are stored, each xml file corresponding to one of the cast land mine 109 night vision pictures in the JPEGImages folder;
the JPEGImages folder is used for storing the data set pictures of the land mines 109, including training pictures and test pictures;
the imagesets folder is provided with a folder Main, and two files of a train. Txt and a train. Txt are mainly stored under the Main folder, so that the training set and the testing set are divided;
a labels folder, which is a file obtained by converting a file to adapt to a YOLO v2 algorithm, wherein each xml file in the above-mentioned Annogens folder is converted into a corresponding txt label, and two files of tranval. Txt and val. Txt are generated;
2) Constructing a passive night vision intelligent mine detection network model by utilizing a YOLO v2 algorithm, adopting DarkNet-19 as a basic network structure, training the passive night vision intelligent mine detection network model by utilizing the data set of the cast land mine 109 generated in the step 1), evaluating the passive night vision intelligent mine detection network model by adopting a loss function, and storing a weight model when the loss value of the loss function is reduced to 0.05, wherein the loss function formula is as follows:
Figure BDA0002408207130000071
where loss is the loss function value, lambda coord For locating the weighting coefficient of the prediction error, lambda noobj For classifying the weight coefficient of the prediction error, S 2 The night vision image of the shed mines 109 is shown divided into S x S boxes, B is the number of bounding boxes in each box,
Figure BDA0002408207130000072
judging whether an ith grid and a jth bounding box in the night vision image are responsible for throwing the land mine 109; (x) i ,y i ) Refers to the center coordinates of the object boundary frame predicted by the passive night vision intelligent mine detection network model,
Figure BDA0002408207130000073
center coordinates of an actual bounding box of the cast land mine 109, (w) i ,h i ) Length and width of bounding box for ith grid, +.>
Figure BDA0002408207130000074
The length and width of the actual bounding box of the cast land mine 109; />
Figure BDA0002408207130000075
Judging whether the center of the cast land mine 109 falls in the ith grid, C i And->
Figure BDA0002408207130000076
Is passiveConfidence of boundary frame predicted by the night vision intelligent mine detection network model and confidence of actual boundary frame; p is p i (c) Class probability representing passive night vision intelligent mine detection network model prediction, < >>
Figure BDA0002408207130000077
Representing actual class probabilities;
3) And (2) evaluating the passive night vision intelligent mine detection network model constructed in the step (2) by adopting a recall rate and an accuracy as network model evaluation standards, wherein the recall rate refers to the proportion of the correctly detected cast land mines 109 to all cast land mines 109 in the verification set, and the formula is as follows:
Figure BDA0002408207130000081
accuracy refers to the proportion of correctly detected cast mines 109 to the detected cast mines 109, and the formula is:
Figure BDA0002408207130000082
wherein Recall is the Recall rate; precision is Precision; tureMines represents that the target is the cast land mine 109 and is correctly detected as the cast land mine 109; tureNMines indicates that the target is not the cast land mine 109 and is not misdetected as the cast land mine 109; falseMines indicates that the target is not the cast land mine 109, but is misdetected as the cast land mine 109; false nmines means that the target is the cast land mine 109, but is not detected as the cast land mine 109;
4) The method for optimizing the passive night vision intelligent lightning detection network model comprises the following specific implementation steps:
a) Pruning the passive night vision intelligent lightning detection network model, setting a threshold value, deleting the connection with lower weight, converting the dense network into a sparse network, retraining the weight of the rest sparse connection in the network model, learning the correct connection in the network parameters with continuous sparsity, and finding the minimum number of connections after multiple iterations;
b) The method comprises the steps of performing quantization processing on a passive night vision intelligent lightning detection network model by using weight sharing and weight reduction methods;
5) If the optimized passive night vision intelligent lightning detection network model reaches the network model evaluation standard, the passive night vision intelligent lightning detection network model is stored; if the optimized passive night vision intelligent lightning detection network model does not reach the network model evaluation standard, performing step 6);
6) Repeating the steps 4) to 5).
In this embodiment, referring to fig. 3 and 4, the passive night vision intelligent mine detection system adopts a passive night vision intelligent mine detection network model to identify, and when the thrown mine 109 is identified, adopts a passive night vision intelligent mine detection ranging model, and uses the FPGA development board 104 to measure and calculate the distance, and the specific implementation steps of the construction and use method of the passive night vision intelligent mine detection ranging model are as follows:
i, converting pixel coordinate values of the cast land mines 109 into actual coordinate values according to a geometrical optical imaging method, and determining the actual distance from the cast land mines 109 to the low-light level camera 101, wherein a measuring and calculating formula of the vertical distance between the cast land mines 109 and the low-light level camera 101 is as follows:
Figure BDA0002408207130000083
Figure BDA0002408207130000084
y＝H·tan(α+Δθ) (6)
wherein Height is a night vision image Height of the cast land mine 109, (u, v) is a pixel coordinate value of the cast land mine 109, (x, y) is an actual coordinate value of the cast land mine 109, θ is a vertical field angle of the low-light camera 101, and Δθ is a step angle; an included angle formed by the pixel coordinates of the throwing land mine 109 and the maximum pixel coordinates is H, which is the ground clearance height of the low-light camera 101, and α is the pitch angle of the low-light camera 109;
according to the geometrical optical imaging method and the actual vertical distance y between the land mine 109 and the low-light camera 101 calculated in the step I, converting the horizontal coordinates of the image pixels into the actual horizontal distance between the land mine 109 and the low-light camera 101, wherein the calculation formula is as follows:
Figure BDA0002408207130000091
/>
Figure BDA0002408207130000092
Figure BDA0002408207130000093
where l is the actual distance from the actual coordinate (0, y) to the micro camera 101, H is the ground clearance of the micro camera 101, (u, v) is the pixel coordinate value of the cast mine 109, (x, y) is the actual coordinate value of the cast mine 109, b is the field radius at the y coordinate value corresponding to the cast mine 109, β is the horizontal field angle of the micro camera 101, and Width is the night vision image Width of the cast mine 109.
In this embodiment, referring to fig. 1 and 5, a passive night vision intelligent lightning detection method, adopting the passive night vision intelligent lightning detection system of this embodiment, comprises the following specific implementation steps:
a. embedding a passive night vision intelligent mine detection network model and a passive night vision intelligent mine detection ranging model into the FPGA development board 104;
b. conveying the passive night vision intelligent mine detection system to a ground area of the land mine to be detected at night for later use;
c. adjusting the focal length of the prime lens 102 to ensure that the prime lens is positioned at the front 18m of the micro-light camera 101 at Jiao Zaisuo;
d. moving the low-light camera 101 by using the instrument loading vehicle 108, shooting and collecting night vision images of the cast land mines 109;
e. transmitting the acquired night vision image of the land mine shed 109 to the FPGA development board 104 in real time through the Cameralink data transmission line 103;
f. identifying night vision images of the cast land mines 109 by using the FPGA development board 104, and judging whether the cast land mines 109 exist or not;
g. if the landmine 109 is identified, measuring and calculating the distance by using the FPGA development board 104, and displaying a detection result and a ranging result on the display screen 105; if the landmine 109 is not identified, performing the following step h;
h. repeating the operations of the steps d to g.
The cast land mine 109 used in this embodiment adopts a six-nine type back-stepping plastic mine, a seven-two type antitank metal mine, and a five-eight type back-stepping plastic mine.
According to the passive night vision intelligent mine detection system and the mine detection method, a passive night vision intelligent mine detection network model is built by means of a YOLOv2 algorithm, a data set of a cast land mine 109 is built, the passive night vision intelligent mine detection network is trained, evaluated and optimized, and the trained passive night vision intelligent mine detection network model is embedded into an FPGA development board; a passive night vision intelligent exploratory ranging model is built by using the imaging principle of geometrical optics and is embedded into an FPGA development board; the device and the method realize intelligent and rapid detection of the cast land mines at night.
Embodiment two:
this embodiment is substantially the same as the first embodiment, and is characterized in that:
in this embodiment, a passive night vision intelligent mine detection method is adopted, and when the cast land mine 109 is not identified after the identification of the set times or the set time length, the area to be detected can be set as a non-obstacle area or a low-alarm area, and then the map is marked; and the passive night vision intelligent mine detection system is powered off manually, or a wireless or wired signal module is improved to control the passive night vision intelligent mine detection system to be powered off, so that the detection tasks of identifying and ranging the cast mines 109 in the area to be detected are completed.
The embodiments of the present invention have been described above with reference to the accompanying drawings, but the present invention is not limited to the embodiments described above, and various changes, modifications, substitutions, combinations or simplifications can be made according to the purposes of the present invention, which are the same as the technical spirit and principles of the technical solution of the present invention, so long as the present invention meets the purpose of the present invention, and the present invention does not deviate from the technical principles and the inventive concepts of the passive night vision intelligent lightning detection system and the intelligent lightning detection method.

Claims (7)

1. A passive night vision intelligent lightning detection system is characterized in that; the system comprises a low-light camera (101), a prime lens (102), a camera link data transmission line (103), an FPGA development board (104), a display screen (105), a supporting frame (106), a lead storage battery (107) and an instrument loading vehicle (108); the fixed focus lens (102) is connected with the micro-light camera (101) through an F bayonet, the micro-light camera (101) is arranged on the supporting frame (106), the FPGA development board (103) is connected with the micro-light camera (101) through the Camellia link data transmission line (103), the FPGA development board (103) is connected with the display screen (105) through a data line, the lead storage battery (107) is used for supplying power for the micro-light camera (101), the FPGA development board (104) and the display screen (105), and the supporting frame (106), the FPGA development board (104), the display screen (105) and the lead storage battery (107) are arranged on the instrument loading vehicle (108); the micro-camera (101) collects night vision images of the land mines (109) through the prime lens (102), transmits image data to the FPGA development board (104) through the camera link data transmission line (103), and the FPGA development board (104) identifies the detected night vision images of suspected targets of the land mines (109), judges whether the night vision images are the land mines (109), measures the distance of the land mines (109) through data processing, obtains distance data of the land mines (109) and the micro-camera (101), and displays detection results and distance measurement results on the display screen (105).
2. The passive night vision intelligent lightning detection system of claim 1, wherein: identifying night vision images of the cast mines (109) by using a passive night vision intelligent mine detection network model, and judging whether the cast mines (109) exist or not, wherein the construction and use method of the passive night vision intelligent mine detection network model comprises the following specific implementation steps:
1) Shooting and collecting not less than 3000 night vision images of the cast land mines (109) by using the low-light camera (101), and manufacturing a data set of the cast land mines (109) according to a PASCAL VOC2007 format;
2) Constructing a passive night vision intelligent mine detection network model by utilizing a YOLO v2 algorithm, adopting DarkNet-19 as a basic network structure, training the passive night vision intelligent mine detection network model by utilizing the data set of the cast land mine (109) generated in the step 1), evaluating the passive night vision intelligent mine detection network model by adopting a loss function, and storing a weight model when the loss value of the loss function is reduced to 0.05, wherein the loss function formula is as follows:
Figure QLYQS_1
where loss is the loss function value, lambda coord For locating the weighting coefficient of the prediction error, lambda noobj For classifying the weight coefficient of the prediction error, S 2 Representing the division of the night vision image of the cast mine (109) into S x S boxes, B representing the number of bounding boxes in each box,
Figure QLYQS_2
judging whether an ith grid and a jth bounding box in the night vision image are responsible for throwing land mines (109); (x) i ,y i ) Refers to the center coordinates of the object boundary frame predicted by the passive night vision intelligent mine detection network model,
Figure QLYQS_3
for the center coordinates of the actual bounding box of the cast land mine (109), (w) i ,h i ) Length and width of bounding box for ith grid, +.>
Figure QLYQS_4
-length and width of the actual bounding box of the cast land mine (109); />
Figure QLYQS_5
Judging whether the center of the cast land mine (109) falls in the ith grid, C i And->
Figure QLYQS_6
The confidence of the boundary frame predicted by the passive night vision intelligent mine detection network model and the confidence of the actual boundary frame; p is p i (c) Class probability representing passive night vision intelligent mine detection network model prediction, < >>
Figure QLYQS_7
Representing actual class probabilities;
3) And (2) evaluating the passive night vision intelligent mine detection network model constructed in the step (2) by adopting a recall rate and an accuracy as network model evaluation standards, wherein the recall rate refers to the proportion of the correctly detected cast mines (109) to all cast mines (109) in the verification set, and the formula is as follows:
Figure QLYQS_8
accuracy refers to the proportion of correctly detected cast mines (109) to the detected cast mines (109), and the formula is:
Figure QLYQS_9
wherein Recall is the Recall rate; precision is Precision; tureMines represents that the target is the cast mine (109) and is correctly detected as the cast mine (109); tureNMines means that the target is not the cast land mine (109) and is not misdetected as the cast land mine (109); falseMines indicates that the target is not the cast land mine (109), but is misdetected as the cast land mine (109); falseNMines means that the target is the cast land mine (109), but is not detected as the cast land mine (109);
4) The method for optimizing the passive night vision intelligent lightning detection network model comprises the following specific implementation steps:
a) Pruning the passive night vision intelligent lightning detection network model, setting a threshold value, deleting the connection with lower weight, converting the dense network into a sparse network, retraining the weight of the rest sparse connection in the network model, learning the correct connection in the network parameters with continuous sparsity, and finding the minimum number of connections after multiple iterations;
b) The method comprises the steps of performing quantization processing on a passive night vision intelligent lightning detection network model by using weight sharing and weight reduction methods;
5) If the optimized passive night vision intelligent lightning detection network model reaches the network model evaluation standard, the passive night vision intelligent lightning detection network model is stored; if the optimized passive night vision intelligent lightning detection network model does not reach the network model evaluation standard, performing step 6);
6) Repeating the steps 4) to 5).
3. The passive night vision intelligent lightning detection system of claim 2, wherein: in the step 1), a night vision image of the cast land mine (109) is shot and acquired by using the low-light camera (101), the exposed area of the cast land mine is changed from the maximum to the minimum in the acquisition process, the cast land mine is changed from 0 degree to 360 degrees clockwise at intervals of 10 degrees on a plane parallel to the ground for each exposed area, and the included angle between the cast land mine and the ground is changed from 0 degree to 360 degrees anticlockwise at intervals of 10 degrees on a plane perpendicular to the ground, so that the completeness of a sample is ensured to the maximum, and a data set of the cast land mine is manufactured according to a PASCAL VOC2007 format after the acquisition of the cast land mine is completed, wherein the data set of the cast land mine comprises the following 4 folders:
an items folder, under which tag files in xml format are stored, each xml file corresponding to one of the cast land mine night vision pictures in the JPEGImages folder;
the JPEGImages folder is used for storing the data set pictures of the landmine, including training pictures and test pictures;
the imagesets folder is provided with a folder Main, and two files of a train. Txt and a train. Txt are mainly stored under the Main folder, so that the training set and the testing set are divided;
and d, a labels folder which is a file obtained by converting the file to adapt to a YOLO v2 algorithm, wherein each xml file in the above-mentioned anotions folder is converted into a corresponding txt label, and two files of tranval. Txt and val. Txt are generated.
4. The passive night vision intelligent mine detection system according to claim 2, wherein a passive night vision intelligent mine detection network model is adopted for identification, and when the cast mine (109) is identified, a passive night vision intelligent mine detection ranging model is adopted, the distance is measured and calculated by using the FPGA development board (104), and the method for constructing and using the passive night vision intelligent mine detection ranging model comprises the following specific implementation steps:
according to a geometrical optical imaging method, converting pixel coordinate values of the cast land mines (109) into actual coordinate values, and determining the actual distance from the cast land mines (109) to the micro-light camera (101), wherein a measuring and calculating formula of the vertical distance between the cast land mines (109) and the micro-light camera (101) is as follows:
Figure QLYQS_10
Figure QLYQS_11
y＝H·tan(α+Δθ) (6)
wherein Height is a night vision image Height of the cast land mine (109), (u, v) is a pixel coordinate value of the cast land mine (109), (x, y) is an actual coordinate value of the cast land mine (109), θ is a vertical field angle of the low-light camera (101), and Δθ is a step angle; an included angle formed by the pixel coordinates of the throwing land mine (109) and the maximum pixel coordinates is H, the ground clearance of the low-light camera (101), and alpha is the pitching angle of the low-light camera (101);
according to the geometrical optical imaging method and the actual vertical distance y between the throwing land mine (109) and the low-light camera (101) calculated in the step I, converting the horizontal coordinates of the image pixels into the actual horizontal distance between the throwing land mine (109) and the low-light camera (101), wherein the calculation formula is as follows:
Figure QLYQS_12
Figure QLYQS_13
Figure QLYQS_14
wherein l is the actual distance from an actual coordinate (0, y) to the micro camera (101), H is the ground clearance of the micro camera (101), u, v is the pixel coordinate value of the cast mine (109), x, y is the actual coordinate value of the cast mine (109), b is the field radius at the y coordinate value corresponding to the cast mine (109), beta is the horizontal field angle of the micro camera (101), and Width is the night vision image Width of the cast mine (109).
5. The passive night vision intelligent lightning detection method adopts the passive night vision intelligent lightning detection system as claimed in claim 1, and is characterized by comprising the following specific implementation steps:
a. embedding a passive night vision intelligent mine detection network model and a passive night vision intelligent mine detection ranging model into the FPGA development board (104);
b. conveying the passive night vision intelligent mine detection system to a ground area of a cast mine (109) to be detected at night for later use;
c. adjusting the focal length of the prime lens (102) to ensure that the prime lens is positioned at the front 18m of the micro-light camera (101) Jiao Zaisuo;
d. moving the low-light camera (101) by means of the instrument loading vehicle (108), capturing and acquiring night vision images of the cast land mines (109);
e. transmitting the acquired night vision image of the cast land mine (109) to the FPGA development board (104) in real time through the Cameralink data transmission line (103);
f. identifying the night vision image of the cast land mine (109) by utilizing the FPGA development board (104), and judging whether the cast land mine (109) exists or not;
g. if the landmine (109) is identified, measuring and calculating the distance by utilizing the FPGA development board (104), and displaying a detection result and a ranging result on the display screen (105); if the landmine (109) is not identified, carrying out the following step h;
h. repeating the operations of the steps d to g.
6. The passive night vision intelligent lightning detection method according to claim 5, wherein: identifying night vision images of the cast mines (109) by using a passive night vision intelligent mine detection network model, and judging whether the cast mines (109) exist or not, wherein the construction and use method of the passive night vision intelligent mine detection network model comprises the following specific implementation steps:
1) Shooting and collecting not less than 3000 night vision images of the cast land mines (109) by using the low-light camera (101), and manufacturing a data set of the cast land mines (109) according to a PASCAL VOC2007 format;
2) Constructing a passive night vision intelligent mine detection network model by utilizing a YOLO v2 algorithm, adopting DarkNet-19 as a basic network structure, training the passive night vision intelligent mine detection network model by utilizing the data set of the cast land mine (109) generated in the step 1), evaluating the passive night vision intelligent mine detection network model by adopting a loss function, and storing a weight model when the loss value of the loss function is reduced to 0.05, wherein the loss function formula is as follows:
Figure QLYQS_15
where loss is the loss function value, lambda coord For locating the weighting coefficient of the prediction error, lambda noobj For classifying the weight coefficient of the prediction error, S 2 Representing the division of the night vision image of the cast mine (109) into S x S boxes, B representing the number of bounding boxes in each box,
Figure QLYQS_16
judging whether an ith grid and a jth bounding box in the night vision image are responsible for throwing land mines (109); (x) i ,y i ) Refers to the center coordinates of the object boundary frame predicted by the passive night vision intelligent mine detection network model,
Figure QLYQS_17
for the center coordinates of the actual bounding box of the cast land mine (109), (w) i ,h i ) Length and width of bounding box for ith grid, +.>
Figure QLYQS_18
-length and width of the actual bounding box of the cast land mine (109); />
Figure QLYQS_19
Judging whether the center of the cast land mine (109) falls in the ith grid, C i And->
Figure QLYQS_20
Confidence of boundary box predicted for passive night vision intelligent mine detection network model and confidence of actual boundary box；p i (c) Class probability representing passive night vision intelligent mine detection network model prediction, < >>
Figure QLYQS_21
Representing actual class probabilities;
3) And (2) evaluating the passive night vision intelligent mine detection network model constructed in the step (2) by adopting a recall rate and an accuracy as network model evaluation standards, wherein the recall rate refers to the proportion of the correctly detected cast mines (109) to all cast mines (109) in the verification set, and the formula is as follows:
Figure QLYQS_22
accuracy refers to the proportion of correctly detected cast mines (109) to the detected cast mines (109), and the formula is:
Figure QLYQS_23
wherein Recall is the Recall rate; precision is Precision; tureMines represents that the target is the cast mine (109) and is correctly detected as the cast mine (109); tureNMines means that the target is not the cast land mine (109) and is not misdetected as the cast land mine (109); falseMines indicates that the target is not the cast land mine (109), but is misdetected as the cast land mine (109); falseNMines means that the target is the cast land mine (109), but is not detected as the cast land mine (109);
4) The method for optimizing the passive night vision intelligent lightning detection network model comprises the following specific implementation steps:
a) Pruning the passive night vision intelligent lightning detection network model, setting a threshold value, deleting the connection with lower weight, converting the dense network into a sparse network, retraining the weight of the rest sparse connection in the network model, learning the correct connection in the network parameters with continuous sparsity, and finding the minimum number of connections after multiple iterations;
b) The method comprises the steps of performing quantization processing on a passive night vision intelligent lightning detection network model by using weight sharing and weight reduction methods;
5) If the optimized passive night vision intelligent lightning detection network model reaches the network model evaluation standard, the passive night vision intelligent lightning detection network model is stored; if the optimized passive night vision intelligent lightning detection network model does not reach the network model evaluation standard, performing step 6);
6) Repeating the steps 4) to 5).
7. The passive night vision intelligent lightning detection method according to claim 5, wherein: the passive night vision intelligent mine detection network model is adopted for identification, when the cast land mine (109) is identified, the passive night vision intelligent mine detection ranging model is adopted, the FPGA development board (104) is utilized for measuring and calculating the distance, and the construction and use method of the passive night vision intelligent mine detection ranging model comprises the following specific implementation steps:
according to a geometrical optical imaging method, converting pixel coordinate values of the cast land mines (109) into actual coordinate values, and determining the actual distance from the cast land mines (109) to the micro-light camera (101), wherein a measuring and calculating formula of the vertical distance between the cast land mines (109) and the micro-light camera (101) is as follows:
Figure QLYQS_24
Figure QLYQS_25
y＝H·tan(α+Δθ) (6)
wherein Height is a night vision image Height of the cast land mine (109), (u, v) is a pixel coordinate value of the cast land mine (109), (x, y) is an actual coordinate value of the cast land mine (109), θ is a vertical field angle of the low-light camera (101), and Δθ is a step angle; an included angle formed by the pixel coordinates of the throwing land mine (109) and the maximum pixel coordinates is H, the ground clearance of the low-light camera (101), and alpha is the pitching angle of the low-light camera (101);
according to the geometrical optical imaging method and the actual vertical distance y between the throwing land mine (109) and the low-light camera (101) calculated in the step I, converting the horizontal coordinates of the image pixels into the actual horizontal distance between the throwing land mine (109) and the low-light camera (101), wherein the calculation formula is as follows:
Figure QLYQS_26
/>
Figure QLYQS_27
Figure QLYQS_28
wherein l is the actual distance from an actual coordinate (0, y) to the micro camera (101), H is the ground clearance of the micro camera (101), u, v is the pixel coordinate value of the cast mine (109), x, y is the actual coordinate value of the cast mine (109), b is the field radius at the y coordinate value corresponding to the cast mine (109), beta is the horizontal field angle of the micro camera (101), and Width is the night vision image Width of the cast mine (109).
CN202010168178.2A 2020-03-11 2020-03-11 Passive night vision intelligent lightning detection system and intelligent lightning detection method Active CN111445522B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010168178.2A CN111445522B (en) 2020-03-11 2020-03-11 Passive night vision intelligent lightning detection system and intelligent lightning detection method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010168178.2A CN111445522B (en) 2020-03-11 2020-03-11 Passive night vision intelligent lightning detection system and intelligent lightning detection method

Publications (2)

Publication Number Publication Date
CN111445522A CN111445522A (en) 2020-07-24
CN111445522B true CN111445522B (en) 2023-05-23

Family

ID=71654002

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010168178.2A Active CN111445522B (en) 2020-03-11 2020-03-11 Passive night vision intelligent lightning detection system and intelligent lightning detection method

Country Status (1)

Country Link
CN (1) CN111445522B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111899525A (en) * 2020-08-18 2020-11-06 重庆紫光华山智安科技有限公司 Distance measuring method, distance measuring device, electronic device, and storage medium
CN112666214A (en) * 2020-12-24 2021-04-16 上海大学 Infrared overtime-phase mine detection system and method based on computer vision
CN113534287A (en) * 2021-06-23 2021-10-22 上海大学 All-weather sound-light mine detection device and method
CN113514871A (en) * 2021-06-23 2021-10-19 上海大学 Vehicle-mounted landmine acoustic vibration mode measurement device and method

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2004092826A1 (en) * 2003-04-18 2004-10-28 Appro Technology Inc. Method and system for obtaining optical parameters of camera
CN108680137A (en) * 2018-04-24 2018-10-19 天津职业技术师范大学 Earth subsidence detection method and detection device based on unmanned plane and Ground Penetrating Radar
CN109632822A (en) * 2018-12-25 2019-04-16 东南大学 A kind of quasi-static high-precision road surface breakage intelligent identification device and its method
CN109923372A (en) * 2016-10-25 2019-06-21 特里纳米克斯股份有限公司 Using the infrared optics detector of integrated filter
CN110261888A (en) * 2019-04-02 2019-09-20 上海大学 A kind of the fast sound-light detection device and detection method of mine
CN110458129A (en) * 2019-08-16 2019-11-15 电子科技大学 Nonmetallic mine recognition methods based on depth convolutional neural networks

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2004092826A1 (en) * 2003-04-18 2004-10-28 Appro Technology Inc. Method and system for obtaining optical parameters of camera
CN109923372A (en) * 2016-10-25 2019-06-21 特里纳米克斯股份有限公司 Using the infrared optics detector of integrated filter
CN108680137A (en) * 2018-04-24 2018-10-19 天津职业技术师范大学 Earth subsidence detection method and detection device based on unmanned plane and Ground Penetrating Radar
CN109632822A (en) * 2018-12-25 2019-04-16 东南大学 A kind of quasi-static high-precision road surface breakage intelligent identification device and its method
CN110261888A (en) * 2019-04-02 2019-09-20 上海大学 A kind of the fast sound-light detection device and detection method of mine
CN110458129A (en) * 2019-08-16 2019-11-15 电子科技大学 Nonmetallic mine recognition methods based on depth convolutional neural networks

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Winter EM 等.spectral methods to detect surface mines.proceedings of SPIE.2008,第6953卷 *
王海馨.抛撒地雷的光学图像探测技术研究.西安电子科技大学硕士论文.2018,(第07期),19-42. *

Also Published As

Publication number Publication date
CN111445522A (en) 2020-07-24

Similar Documents

Publication Publication Date Title
CN111445522B (en) Passive night vision intelligent lightning detection system and intelligent lightning detection method
CN108106801B (en) Bridge and tunnel disease non-contact detection system and detection method
CN105652154B (en) Contact Running State security auditing system
CN103808723A (en) Exhaust gas blackness automatic detection device for diesel vehicles
CN104122560B (en) Electric transmission line wide area ice condition monitoring method
CN111045000A (en) Monitoring system and method
CN109979468B (en) Lightning stroke optical path monitoring system and method
CN104410839B (en) A kind of mobile power transmission line bar tower region mountain fire and disaster of mountain massif coast on-line monitoring system and monitoring method
CN113359097A (en) Millimeter wave radar and camera combined calibration method
CN112487900B (en) SAR image ship target detection method based on feature fusion
CN112348882A (en) Low-altitude target tracking information fusion method and system based on multi-source detector
CN110162735B (en) Ballistic trajectory calculation method and system based on laser ranging telescope
CN111830470B (en) Combined calibration method and device, target object detection method, system and device
CN112802004B (en) Portable intelligent video detection device for health of power transmission line and pole tower
CN116448773B (en) Pavement disease detection method and system with image-vibration characteristics fused
CN115620239B (en) Point cloud and video combined power transmission line online monitoring method and system
CN114898238A (en) Wild animal remote sensing identification method and device
CN211477203U (en) Refined monitoring equipment system based on high-resolution remote sensing image
CN100421456C (en) All-weather two camera shooting laser speed measuring testification apparatus
CN112066882A (en) Detection device, system and method for contact network
CN110989645A (en) Target space attitude processing method based on compound eye imaging principle
CN116343534A (en) Airplane berth guiding system and method
CN112700423B (en) Deep learning-based automatic detection method and system for surface damage defects of airframe
CN111089607B (en) Automatic calibration method for detection capability of telescope system
CN211527336U (en) Tramcar trapezoidal turnout deformation monitoring system

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