CN104915962A - Real-time point target detection method in ultraviolet visible light video - Google Patents
Real-time point target detection method in ultraviolet visible light video Download PDFInfo
- Publication number
- CN104915962A CN104915962A CN201510336423.5A CN201510336423A CN104915962A CN 104915962 A CN104915962 A CN 104915962A CN 201510336423 A CN201510336423 A CN 201510336423A CN 104915962 A CN104915962 A CN 104915962A
- Authority
- CN
- China
- Prior art keywords
- target
- information table
- image
- ultraviolet
- point
- 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
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
- G06T7/77—Determining position or orientation of objects or cameras using statistical methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10052—Images from lightfield camera
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a real-time point target detection method in an ultraviolet visible light video, which comprises steps: (1) a target position information table is built; (2) search parameters are initialized; (3) an image is read and filtering is carried out; (4) binarization processing is carried out; (5) whether the information table is empty is judged, and if yes, an ultraviolet target is searched in the whole image, and the information table is updated; (6) the center position of a candidate target point in the information table serves as a search center, searching is carried out in a certain range, the information table is updated when a target exists, and the seventh step is carried out, and the candidate target is cancelled in the information table without a target, and the third step is carried out; (7) if the target point in the information table appears for more times than a preset threshold n, an alarm is given, and fusion with the frame image of the visible light is carried out; and (8) steps from (3) to (7) are repeated until all image frames in the video are processed. The false alarm probability can be effectively reduced, the target detection probability is improved, and the target can be accurately positioned and analyzed.
Description
Technical field
The invention belongs to video information process field, relate to real-time detection and the fusion method of point target in a kind of ultraviolet-visible video, can be used for the detection identification to target in ultraviolet-visible video.
Background technology
Traditional infrared search and rescue instrument carry out airborne search and rescue on a large scale time, infrared target is submerged in noise background substantially, and signal to noise ratio (S/N ratio) is lower, brings very large difficulty to search-and-rescue work.Because earth background is few without clutter, jamming target source at ultraviolet band, the UV signal signal to noise ratio (S/N ratio) that detector receives is higher.Therefore, the mode adopting ultraviolet to carry out searching and rescuing is arisen at the historic moment.
In ultraviolet detection, target is point-like, without obvious shape information, due to the high-gain of system noise and ultraviolet imaging enhancer, may cause false-alarm, needs to take measures to reduce false alarm rate.In addition, ultraviolet image does not carry background information substantially, therefore has difficulties to the location of target.
Summary of the invention
The technical matters that the present invention solves is: overcome the deficiencies in the prior art, provide a kind of day blind ultraviolet and visible light video in the real-time detection of point target and fusion method, the false-alarm that the high-gain due to system noise and ultraviolet imaging enhancer causes can be reduced, improve the reliability of system early warning, and can by the location of visible images realization to target.
Technical solution of the present invention is: the real-time detection method of point target in a kind of ultraviolet-visible video, comprises the steps:
(1) target position information table is set up, the number of times that geographic coordinate and this target for record object occur;
(2) adopt Uv and visible light two kinds of modes to make a video recording to same search and rescue region simultaneously, shooting m frame per second, image resolution ratio is p, determines to search and rescue speed v and region of search radius size R=[v/ (p × m)] simultaneously, and wherein symbol [] represents round numbers;
(3) from ultraviolet video, extract a frame ultraviolet image of t, filtering is carried out to described ultraviolet image, filter out background noise and noise of detector, then obtain the ultraviolet image after Image semantic classification;
(4) adopt the method for global statistics segmentation to carry out binary conversion treatment to the image that step (3) obtains, obtain bianry image I
b;
(5) judge whether target position information table is empty, if it is empty then to the bianry image I that step (4) obtains
bcarry out connected component labeling, and the connected domain alternatively target that will mark, then by the position of all candidate targets stored in target position information table, the number of times that each candidate target occurs is set to 1 simultaneously, then enters step (6); Then directly enter step (6) if not empty;
(6) positional information of each candidate target point in target position information table is extracted, calculate the center of each candidate target point, and respectively using the center of each candidate target point as search center, with hunting zone R for radius is searched in present frame ultraviolet image, and connected component labeling is carried out in hunting zone, if there is candidate target point in hunting zone, then upgrade target position information table, the number of times that candidate target point occurs is added 1, and upgrade the positional information of candidate target point, enter step (7); If there is not candidate target point in hunting zone, then in target position information table, delete the candidate target point corresponding with search center, and enter step (3);
(7) add up the times N that in target position information table, each candidate target occurs, if N is greater than the threshold value n preset, report to the police, export corresponding target location and superpose this ultraviolet target image on the visible ray two field picture of t;
(8) step (3) is repeated to step (7), until complete the process to picture frames all in video.
Filtering method in described step (3) is:
(31) the frame ultraviolet image supposing t is I, and pixel is wherein designated as I (i, j), centered by I (i, j), gets the neighborhood f not comprising I (i, j), and adds up the sum of all pixels M comprised in neighborhood f;
(32) according to neighborhood averaging formula, I (i, j) is processed, obtains the filter value I ' (i, j) of I (i, j):
The size of described neighborhood f is 1 or radical sign 2.
The present invention's advantage is compared with prior art: because noise is normally discrete, and aiming spot is relatively fixing, and the inventive method utilizes a certain image-region to export UV signal continuously to detect target, and false alarm rate will significantly reduce.Record target location after finding target, and ultraviolet target image and visible ray background image are carried out image co-registration, utilize visible ray background can position analysis to target accurately.The inventive method only reads in a frame ultraviolet image and a frame visible images at every turn, and Processing Algorithm is the mode of going forward one by one, and committed memory is few, and method only used comparison operation, and algorithm is simple, is convenient to hardware implementing and detects in real time.
Accompanying drawing explanation
Fig. 1 is the FB(flow block) of the inventive method;
Fig. 2 is the target search schematic diagram of the inventive method.
Embodiment
As shown in Figure 1, be the FB(flow block) of the inventive method.Fig. 2 is the basic thought of the inventive method, wherein (a), (b), (c) represent continuous print three two field picture, in figure, the point of black represents UV signal, when there being signal in (a), get the region centered by signal, in (b) with this region for hunting zone, judge whether UV signal, corresponding region search signal in (c) is got centered by signal afterwards in (b), if there is signal, trigger alarm.
The key step of the inventive method is as follows:
(1) target position information table is set up, the number of times that geographic coordinate and this target for record object occur;
(2) adopt Uv and visible light two kinds of modes to make a video recording to same search and rescue region simultaneously, shooting m frame per second, image resolution ratio is p, determines to search and rescue speed v and region of search radius size R=[v/ (p × m)] simultaneously, and wherein symbol [] represents round numbers.
Such as searching and rescuing fastest is 50m/s, searching and rescuing instrument frame frequency is 25fps, the displacement that two two field picture correspondences search and rescue instrument is 2m, image resolution ratio is 0.088m, and in two two field pictures, target moves 23 pixels, because target has movement, scope is expanded a little, therefore, if there is target, then on continuous print two two field picture, the distance of UV signal should be less than 25 pixels.
(3) from ultraviolet video, extract a frame ultraviolet image of t, filtering is carried out to described ultraviolet image, filter out background noise and noise of detector, then obtain the ultraviolet image after Image semantic classification.
Be the Neighborhood Filtering of 1 for radius, the ultraviolet image of t is I, and centered by I (i, j), its neighborhood is expressed as f={ (i, j-1), (i, j+1), (i-1, j), (i+1, j) }; The sum of all pixels comprised in neighborhood f is 4, processes, obtain the filter value I ' (i, j) of I (i, j) according to neighborhood averaging formula to I (i, j):
Filtering is carried out to each pixel of full figure, obtains the pretreated result I ' of present frame ultraviolet image.
(4) adopt the method for global statistics segmentation to carry out binary conversion treatment to the image that step (3) obtains, obtain bianry image I
b; Computation process is as follows:
(5) judge whether target position information table is empty, if it is empty then to the bianry image I that step (4) obtains
bcarry out connected component labeling, the pixel value of each connection be 1 region be a connected domain, be a little a candidate target point set in each connected domain.Wherein each candidate target is designated as A
i, A
ifor the set A of the point of connection
i={ p
1, p
2.p
m, there is a p by each candidate target
1, p
2.p
mposition stored in the geographic coordinate as target in target position information table, the number of times that each candidate target occurs is set to 1 simultaneously, then enters step (6); Then directly enter step (6) if not empty.
(6) to extract in target position information table in each candidate target positional information a little, calculate the center of each candidate target point respectively, successively using each center as search center, with hunting zone R for radius is searched in present frame ultraviolet image, connected component labeling is carried out in hunting zone, if have candidate target in hunting zone, then upgrade target position information table, the number of times that candidate target occurs is added 1, and upgrade the positional information (in hunting zone emerging position) of candidate target, enter step (7); If not there is candidate target in hunting zone, then in target position information table, delete this candidate target, and enter step (3);
(7) add up the times N that in target position information table, each candidate target point occurs, if N is greater than the threshold value n preset, report to the police, if n is 3, then represent that continuous 3 two field pictures have ultraviolet target.Export corresponding target location A
iand this ultraviolet target image is superposed on the visible ray two field picture of t;
The process of superposition ultraviolet target is as follows:
(1) from target position information table, read ultraviolet point target collection A
i={ p
1, p
2.p
m;
(2) point of relevant position on visible ray frame is demarcated as redness
Wherein R, G, B represent red, green, blue triple channel respectively.
The content be not described in detail in instructions of the present invention belongs to the known technology of those skilled in the art.
Claims (3)
1. the real-time detection method of point target in ultraviolet-visible video, is characterized in that comprising the steps:
(1) target position information table is set up, the number of times that geographic coordinate and this target for record object occur;
(2) adopt Uv and visible light two kinds of modes to make a video recording to same search and rescue region simultaneously, shooting m frame per second, image resolution ratio is p, determines to search and rescue speed v and region of search radius size R=[v/ (p × m)] simultaneously, and wherein symbol [] represents round numbers;
(3) from ultraviolet video, extract a frame ultraviolet image of t, filtering is carried out to described ultraviolet image, filter out background noise and noise of detector, then obtain the ultraviolet image after Image semantic classification;
(4) adopt the method for global statistics segmentation to carry out binary conversion treatment to the image that step (3) obtains, obtain bianry image I
b;
(5) judge whether target position information table is empty, if it is empty then to the bianry image I that step (4) obtains
bcarry out connected component labeling, and the connected domain alternatively target that will mark, then by the position of all candidate targets stored in target position information table, the number of times that each candidate target occurs is set to 1 simultaneously, then enters step (6); Then directly enter step (6) if not empty;
(6) positional information of each candidate target point in target position information table is extracted, calculate the center of each candidate target point, and respectively using the center of each candidate target point as search center, with hunting zone R for radius is searched in present frame ultraviolet image, and connected component labeling is carried out in hunting zone, if there is candidate target point in hunting zone, then upgrade target position information table, the number of times that candidate target point occurs is added 1, and upgrade the positional information of candidate target point, enter step (7); If there is not candidate target point in hunting zone, then in target position information table, delete the candidate target point corresponding with search center, and enter step (3);
(7) add up the times N that in target position information table, each candidate target occurs, if N is greater than the threshold value n preset, report to the police, export corresponding target location and superpose this ultraviolet target image on the visible ray two field picture of t;
(8) step (3) is repeated to step (7), until complete the process to picture frames all in video.
2. the real-time detection method of point target in a kind of ultraviolet-visible video according to claim 1, is characterized in that: the filtering method in described step (3) is:
(31) the frame ultraviolet image supposing t is I, and pixel is wherein designated as I (i, j), centered by I (i, j), gets the neighborhood f not comprising I (i, j), and adds up the sum of all pixels M comprised in neighborhood f;
(32) according to neighborhood averaging formula, I (i, j) is processed, obtains the filter value I ' (i, j) of I (i, j):
3. the real-time detection method of point target in a kind of ultraviolet-visible video according to claim 2, is characterized in that: the size of described neighborhood f is 1 or radical sign 2.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510336423.5A CN104915962B (en) | 2015-06-17 | 2015-06-17 | The real-time detection method of point target in a kind of ultraviolet-visible video |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510336423.5A CN104915962B (en) | 2015-06-17 | 2015-06-17 | The real-time detection method of point target in a kind of ultraviolet-visible video |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104915962A true CN104915962A (en) | 2015-09-16 |
CN104915962B CN104915962B (en) | 2018-02-06 |
Family
ID=54085000
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510336423.5A Active CN104915962B (en) | 2015-06-17 | 2015-06-17 | The real-time detection method of point target in a kind of ultraviolet-visible video |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104915962B (en) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101363918A (en) * | 2008-08-12 | 2009-02-11 | 阎锋 | Method for positioning and identifying objects by solar blind UV |
CN102574569A (en) * | 2009-09-04 | 2012-07-11 | 雷声公司 | Search and rescue using ultraviolet radiation |
CN103413138A (en) * | 2013-07-18 | 2013-11-27 | 航天恒星科技有限公司 | Method for detecting point target in infrared image sequence |
-
2015
- 2015-06-17 CN CN201510336423.5A patent/CN104915962B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101363918A (en) * | 2008-08-12 | 2009-02-11 | 阎锋 | Method for positioning and identifying objects by solar blind UV |
CN102574569A (en) * | 2009-09-04 | 2012-07-11 | 雷声公司 | Search and rescue using ultraviolet radiation |
CN103413138A (en) * | 2013-07-18 | 2013-11-27 | 航天恒星科技有限公司 | Method for detecting point target in infrared image sequence |
Non-Patent Citations (1)
Title |
---|
赵小明 等: "基于移动式管道滤波的红外小目标检测方法研究", 《红外技术》 * |
Also Published As
Publication number | Publication date |
---|---|
CN104915962B (en) | 2018-02-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Siogkas et al. | Traffic lights detection in adverse conditions using color, symmetry and spatiotemporal information | |
De Charette et al. | Real time visual traffic lights recognition based on spot light detection and adaptive traffic lights templates | |
CN108152808A (en) | A kind of circumference intelligent predicting method for early warning based on millimetre-wave radar | |
EP3098753A1 (en) | Lane detection | |
RU2017100468A (en) | LOW AND HIGH QUALITY CLASSIFIERS APPLICABLE TO IMAGES OF ROAD SCENES | |
CN106845346A (en) | A kind of image detecting method for airfield runway foreign bodies detection | |
CN103049909B (en) | A kind of be focus with car plate exposure method | |
CN104282020A (en) | Vehicle speed detection method based on target motion track | |
WO2003001473A1 (en) | Vision-based collision threat detection system_ | |
JP2008187347A (en) | On-vehicle navigation system, road marking identifying program and road marking identifying method | |
CN103034843B (en) | Method for detecting vehicle at night based on monocular vision | |
CN104063882B (en) | Vehicle video speed measuring method based on binocular camera | |
CN107516423B (en) | Video-based vehicle driving direction detection method | |
CN111781600A (en) | Vehicle queuing length detection method suitable for signalized intersection scene | |
CN107729843A (en) | The low-floor tramcar pedestrian recognition method merged based on radar with visual information | |
Wu et al. | Traffic monitoring and vehicle tracking using roadside cameras | |
CN103679748A (en) | Dim point target extraction method and device of infrared remote sensing image | |
Kumar et al. | Vehicle speed detection using corner detection | |
CN103824067B (en) | The location of a kind of image main target and recognition methods | |
CN108629225B (en) | Vehicle detection method based on multiple sub-images and image significance analysis | |
CN108257152A (en) | A kind of road intrusion detection method and system based on video | |
Alkawsi et al. | Arabic vehicle licence plate recognition using deep learning methods | |
CN104008396A (en) | In and out people flow statistical method based on people head color and shape features | |
CN103413138A (en) | Method for detecting point target in infrared image sequence | |
CN104915962A (en) | Real-time point target detection method in ultraviolet visible light video |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |