CN107945215A - High-precision infrared image tracker and a kind of target fast tracking method - Google Patents
High-precision infrared image tracker and a kind of target fast tracking method Download PDFInfo
- Publication number
- CN107945215A CN107945215A CN201711343812.6A CN201711343812A CN107945215A CN 107945215 A CN107945215 A CN 107945215A CN 201711343812 A CN201711343812 A CN 201711343812A CN 107945215 A CN107945215 A CN 107945215A
- Authority
- CN
- China
- Prior art keywords
- target
- real
- image
- template
- time image
- 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
- G06T7/20—Analysis of motion
- G06T7/207—Analysis of motion for motion estimation over a hierarchy of resolutions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/38—Registration of image sequences
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
-
- 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
- G06T2207/10048—Infrared image
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Signal Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of high-precision infrared image tracker and a kind of target fast tracking method, it is frame that the tracker, which is based on programmable logic array and digital processing chip, including real time image collection processing module, realtime graphic memory module, control module FPGA, output image memory module, communication module, digital signal processing module DSP and dynamic memory module SDRAM.The target realtime graphic changed by control module FPGA collections through A/D, and the image is stored in realtime graphic memory module by table tennis technology, then by the information of image and tracing positional is transferred to system by communication module after digital signal processing module DSP processing.Tracking of the present invention includes the technologies such as target Closing Binary Marker, Fast Search, the Adaptive template-updating of binary map mask and template amendment, the accuracy of template matching algorithm can be improved, effectively avoid false target, interference of the false target to template is reduced, has the advantages that target recognition and tracking is reliably low with to hardware requirement etc..
Description
Technical Field
The invention relates to a target tracker and a target tracking and identifying method, in particular to a high-precision infrared image tracker for a guided weapon system to accurately strike a target and a tracking method for coarsely aiming the accurately strike target by using the tracker.
Background
In a weapon system that accurately strikes a target, noise is caused by environmental influences, such as wind blowing, interference of leaves of a bird, and shaking of a photoelectric imaging system, and is recognized as a target in a general detection method. These false targets, which act as disturbances, can affect the tracking accuracy, resulting in inaccurate targeting.
Early trackers were simple in algorithm, and often required algorithmic selection to fit the tracked target according to a priori knowledge of the target before performing the task. Meanwhile, due to the limited hardware resources, the tracker is difficult to complete the processing of a large target surface and multi-target tracking in real time. In addition, the adopted strategy is simple, the hardware platform efficiency is not high enough, and the stable tracking of the target under the complex state change is still difficult to complete.
The prior art generally adopts a gray-scale correlation matching tracking technology, which has the advantages of easy hardware realization and excellent noise resistance, thereby being widely applied. However, in the conventional gray-scale correlation matching tracking, accumulated errors often occur, and the template is usually replaced at an appropriate time to effectively overcome the accumulated errors. When the template is updated, the simple weighted combination of the old template and the new template is generally adopted, the target template is easily damaged by background clutter (such as target shielding) because the background and the hit target are not distinguished, and the conventional template matching and tracking technology lacks the perception capability of target deformation under the clutter background, so that the target is easily drifted out of the template for matching and tracking, and finally the target tracking fails.
In addition, in the gray scale correlation matching tracking technology, the matching speed when hitting the target is a key factor, which affects the tracking target accuracy of the tracking system, and further affects the hitting target accuracy. However, the main factor influencing the matching speed is the number of times of calculating the correlation (i.e. the similarity between the template and the real-time image during matching), and at present, a layering strategy and an improved search strategy are often adopted to improve the matching speed, but the existing algorithms still cannot meet the requirements of the system on precision, real-time performance and low cost at the same time.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide an infrared image tracker which has the advantages of low cost, high reliability and the like and is suitable for a weapon system to accurately strike a target, and a tracking method for roughly aiming at the accurately struck target by using the tracker. The tracking method comprises the technologies of target binary marking, quick search strategy, adaptive template updating and template correction of a binary image mask and the like, can improve the accuracy of a template matching algorithm, inhibits tracking drift, realizes adaptive sensing of target deformation and plays a role in quick identification and tracking.
One of the technical schemes of the invention is as follows: a high-precision infrared image tracker is characterized in that the tracker is based on a programmable logic array and a digital processing chip as a framework and comprises a real-time image acquisition processing module, a real-time image storage module, a control module FPGA, an output image storage module, a communication module, a digital processing module DSP and a dynamic storage module SDRAM; the real-time image acquisition processing module comprises an analog-to-digital conversion module A/D and a programmable logic array FPGA, the real-time image storage module comprises static memories SRAM1 and SRAM2, the output image storage module comprises static memories SRAM3 and SRAM4, the control module FPGA is configured with a power supply and an erasable memory EPROM, the digital processing module DSP is configured with a power supply and FLASH, and the communication module is configured with RS232 and RS422 communication serial ports; the erasable memory EPROM and the FLASH are used for storing programs for the chip to run, and the dynamic storage module SDRAM provides data storage for the digital processing module DSP to run.
The real-time image acquisition module acquires a target real-time image subjected to A/D conversion through the control module FPGA, the image is stored in the real-time image storage module through a ping-pong technology, namely, the control module FPGA firstly acquires a frame of image and stores the frame of image in the SRAM1 of the real-time image storage module to generate image interruption, the digital processing module DSP responds to the interruption, image data in the SRAM1 is taken away through the EDMA to be analyzed and processed, then the processed image is transmitted to the SRAM3 in the output image storage module, and the D/A video display is carried out under the control of the control module FPGA. When the digital processing module DSP generates interruption, the control module FPGA stores the next acquired frame image into the SRAM2 in the real-time image storage module, when the digital processing module DSP finishes processing and analyzing the frame data and the control module FPGA finishes acquiring the next frame image and generates interruption, the digital processing module DSP takes the data in the SRAM2 away through an EDMA, transmits the data to an SRAM4 in the output image storage module after the data is analyzed and processed, and displays the data through a D/A video under the control of the control module FPGA; and repeating the steps in such a circulating way, and after the digital processing module DSP processes one frame of image, transmitting the image and the information of the tracking position to the fire control system through the communication module.
Furthermore, the control module FPGA roughly selects a target area from the real-time image acquired by the real-time image acquisition module and stores the target area into the real-time image storage module.
Further, the analysis processing of the digital processing module DSP is to carry out binarization processing on the target area stored in the real-time image storage module, and obtain the real centroid position of the target area by using an eight neighborhood marking method; and selecting a target area with the same size as that stored in a real-time image storage module as an original target template picture according to the real centroid position, and carrying out related matching on the original target template picture and a target real-time image until the real position of the target in the real-time image is obtained.
The second technical scheme of the invention is as follows: a rapid tracking method for performing coarse aiming and accurate target hitting based on the high-precision infrared image tracker is characterized by comprising the following specific implementation steps:
step one, roughly selecting a target area in a collected real-time image;
secondly, performing binarization processing on the target area selected in the first step, obtaining the centroid position of the target area by using an eight-neighborhood labeling method, and judging whether the centroid position is the real centroid position of the target or not; if yes, selecting the target area with the same size as the step one as the original target template picture by using the real centroid position; if not, selecting the target area by the method of the step, and carrying out binarization processing and marking until the real centroid position of the target is obtained;
step three, the original target template image in the step two is matched with the target real-time image in a relevant manner, if the matching is successful, the real position of the target in the real-time image can be obtained, the target template is updated according to the matching result, and the target tracking is switched in; and if the matching fails, performing Kalman or attitude information target prediction, and acquiring a target real-time image again to perform target matching identification so as to obtain a real target.
Further, when the target region is subjected to target extraction, a gray value which enables the variance between the target and the background to be maximum is obtained by adopting an improved maximum inter-class variance method.
Further, the gray scale larger than the gray scale value in the target region is 255, otherwise is 0.
Further, eight neighborhood zone marking is carried out on a plurality of connected zones appearing after the target zone is subjected to binarization processing to distinguish a real target from a false target, the connected zones which pass the eight neighborhood zone marking are combined, finally, the connected zone with the area closest to the pixel imaged by the target is selected, the centroid position of the connected zone is calculated, the centroid position is taken as the reference, then, the zone with the same size as that in the first step is selected from the target real-time image as the target zone, and the target zone obtained at the moment is the target template image, namely the target needing accurate striking.
Further, in step three, the target real-time image and the original target template image are represented by gray value matrix as S (M, N) and T (M, N), respectively, where M is>m,N>, n; the target template T (M, N) is translated on the real-time image S (M, N), and the area of the real-time image covered by the template is a subgraph S u,v Wherein (u)V) the coordinates of the subgraph with the upper left corner as the origin, called the reference point, and a measure is calculated on the real-time image reference point to characterize the size of the difference program between the target template graph and the corresponding subgraph, called the correlation value; the greater the correlation value, the higher the similarity between the target template graph and the corresponding subgraph, and if the correlation value is greater than a threshold value T established in advance, the matching is considered to be successful.
Further, in the third step, the correlation matching between the original target template image and the target real-time image is realized by adopting an improved hill-climbing search strategy and a simplified correlation measurement mode, and the specific method adopts a mean normalization algorithm, wherein the algorithm is as follows:
simplification of the above equation yields:
wherein:
and if the correlation value is larger than the set threshold value T, the matching is considered to be successful, the position and the correlation value of the target in the real-time image can be obtained, the target tracking process is switched to according to the updating of the target template, otherwise, the target position prediction is carried out by utilizing Kalman or attitude information, and then the target tracking is carried out by re-acquiring the real-time image of the target.
Further, in the third step, firstly, the original target template image is binarized and 3 × 3 expansion is performed to adapt to the small deformation capability of the target, and when the template is updated, the target template binary mask image is subjected to template updatingNon-zero for alpha 1 Updating the weight value, and adopting alpha for the pixel point with the value of zero 2 And updating the weight value to achieve the aims of delaying the background to carry out the target template and quickly updating the real target template graph.
Wherein Bin k (x, y) represents the mask map after binary expansion of the target template, T k+1 For updated target template graphs, T k Target template graph used for current matching, I k Obtaining an optimal target template graph for the current matching; alpha (alpha) ("alpha") 1 =λ 1 A,α 2 =λ 2 A, and satisfies λ 1 >λ 2 And A is an interframe confidence coefficient calculated according to a continuous multi-frame matching result:
and matching the next frame of image by using the updated new target template, and then repeatedly entering the cyclic step of template matching, calculating a correlation value, obtaining the current best matching position, target template updating and template matching to realize target tracking.
Compared with the prior art, the invention has the advantages that:
1. the tracker provided by the invention has low cost and high reliability.
2. The invention provides a template target binary mark, which can effectively mark a real target area and avoid a false target.
3. The measurement formula method for improving the hill climbing search strategy, sorting the related values, marking the positions and simplifying the related values solves the problems of more calculation times of the target tracking related values, large calculation amount, incapability of real-time tracking and poor precision.
4. The invention provides a self-adaptive template updating method of a binary image mask, which solves the problem of unreasonable template updating under the condition of not distinguishing a target from a background, reduces the interference of a false target on the template, has the characteristics of accurate identification and real-time tracking, has the advantage of low requirement on hardware, and has good application prospect in accurate striking weapons.
Drawings
FIG. 1 is a schematic diagram of a target tracker
FIG. 2 is a diagram illustrating binarization and labeling of a target region
FIG. 3 is a schematic diagram of the correlation matching principle
FIG. 4 is a flow chart of a hill-climbing search strategy
Fig. 5 is a target matching tracking flow chart.
Detailed Description
The specific embodiments of the present invention will now be further described with reference to the following examples.
As shown in fig. 1, a high-precision infrared image tracker is based on a programmable logic array and a digital processing chip as a framework, and comprises a real-time image acquisition processing module, a real-time image storage module, a control module FPGA, an output image storage module, a communication module, a digital processing module DSP and a dynamic storage module SDRAM; the real-time image acquisition processing module comprises an analog-to-digital conversion module A/D and a programmable logic array FPGA, the real-time image storage module comprises static memories SRAM1 and SRAM2, the output image storage module comprises static memories SRAM3 and SRAM4, the control module FPGA is configured with a power supply and an erasable memory EPROM, the digital processing module DSP is configured with a power supply and FLASH, and the communication module is configured with RS232 and RS422 communication serial ports; the erasable memory EPROM and the FLASH are used for storing programs for the chip to run, and the dynamic storage module SDRAM provides data storage for the digital processing module DSP to run.
The real-time image acquisition module acquires a target real-time image subjected to A/D conversion through the control module FPGA, the image is stored in the real-time image storage module through a ping-pong technology, namely, the control module FPGA firstly acquires a frame of image and stores the frame of image in the SRAM1 of the real-time image storage module to generate image interruption, the DSP responds to the interruption, image data in the SRAM1 is taken away through the EDMA to be analyzed and processed, then the processed image is transmitted to the SRAM3 in the output image storage module, and the image is displayed through a D/A video under the control of the control module FPGA. When the digital processing module DSP generates interruption, the control module FPGA stores the next acquired frame image into the SRAM2 in the real-time image storage module, when the digital processing module DSP finishes processing and analyzing the frame data and the control module FPGA finishes acquiring the next frame image and generates interruption, the digital processing module DSP takes the data in the SRAM2 away through an EDMA, transmits the data to an SRAM4 in the output image storage module after the data is analyzed and processed, and displays the data through a D/A video under the control of the control module FPGA; and repeating the steps in a circulating way, and after the digital processing module DSP processes one frame of image, transmitting the image and the information of the tracking position to the fire control system through the communication module.
And the control module FPGA roughly selects a target area from the real-time image acquired by the real-time image acquisition module and stores the target area into the real-time image storage module.
The analysis processing of the digital processing module DSP is to carry out binarization processing on the target area stored in the real-time image storage module and obtain the real centroid position of the target area by applying an eight neighborhood marking method; and selecting a target area with the same size as that stored in a real-time image storage module as an original target template picture according to the real centroid position, and carrying out related matching on the original target template picture and a target real-time image until the real position of the target in the real-time image is obtained.
A fast tracking method for performing rough aiming and accurate hitting target based on the high-precision infrared image tracker is shown in fig. 5, and comprises the following specific implementation steps:
step one, roughly selecting a target area in an acquired real-time image, wherein false targets generated by various environmental influences can occur in the target area.
Secondly, performing binarization processing on the target area selected in the first step, obtaining the centroid position of the target area by using an eight-neighborhood labeling method, and judging whether the centroid position is the real centroid position of the target or not; if yes, selecting the target area with the same size as the step one as the original target template picture by using the real centroid position; and if not, selecting the target area by the method of the steps again, and carrying out binarization processing and marking until the real centroid position of the target is obtained.
When the target region is subjected to target extraction, a gray value which enables the variance between the target and the background to be maximum is obtained by adopting an improved maximum inter-class variance method, the gray value which is larger than the gray value in the target region is 255, and otherwise, the gray value is 0. After binarization processing, because of a plurality of connected regions which can appear in the false target, eight-neighborhood labeling is needed to distinguish the real target from the false target, the connected regions which pass the eight-neighborhood labeling are combined, and finally the connected region with the area closest to the pixel imaged by the target is selected to calculate the centroid position of the connected region. And taking the centroid position as a reference, selecting an area with the same size as that in the first step from the target real-time image as a target area, wherein the obtained target area is a target template picture, namely a target needing accurate striking. The target area binarization and labeling schematic diagram is shown in fig. 2, and it is very likely that the beating target does not hit the real target 3 as the centroid detected by the conventional method.
Step three, the original target template image in the step two is matched with the target real-time image in a relevant manner, if the matching is successful, the real position of the target in the real-time image can be obtained, the target template is updated according to the matching result, and the target tracking is switched in; and if the matching fails, performing Kalman or attitude information target prediction, and acquiring a target real-time image again to perform target matching identification so as to obtain a real target.
As shown in FIG. 3, let the gray value matrixes for the real-time target image and the original target template image be S (M, N) and T (M, N), respectively, where M is>m,N>, n. The target template T (M, N) is translated on the real-time image S (M, N), and the area of the real-time image covered by the template is a subgraph S u,v (u, v) is the coordinate of the subgraph with the top left corner as the origin,referred to as reference points. A metric is computed at the real-time image reference points to characterize the size of the difference program between the target template graph and the corresponding subgraph, called the correlation value. The greater the correlation value is, the higher the similarity degree between the target template graph and the corresponding subgraph is, and if the correlation value is greater than a threshold value T established in advance, the matching is considered to be successful.
In order to meet the requirements of real-time performance and precision, an original target template graph and a target real-time image are subjected to relevant matching by adopting an improved hill-climbing search strategy and a simplified relevance measurement mode. The improved hill-climbing search strategy mainly adjusts the hill-climbing sequence according to the priority and reduces the calculation of the correlation degree by using samples. The adjustment of the hill climbing sequence according to the priority also solves many unnecessary calculation search processes on the premise of ensuring the matching precision, such as: a plurality of climbers search the point with the maximum correlation value and a plurality of climbers search the local maximum value, and certainly, the calculation efficiency can be further improved and the redundant calculation can be avoided by setting the correlation matrix table and the search position matrix table in the search process.
The search process from coarse to fine can be realized by setting initial hill climbing points at equal intervals under the condition of ensuring the precision, which is inspired by a hierarchical matching algorithm, and the specific flow is shown in fig. 4: the first step, initializing a position table and a relevance table, setting an initial climbing point with equal step length, and calculating subgraphs S of a template graph and a real-time graph u,v (u, v) correlation values, a record location table and a correlation table; step two, sorting the related tables from big to small, judging whether the corresponding position table is marked, if so, not calculating the subgraph S of the template graph and the real-time graph u,v (u, v) correlation values; otherwise, performing the next mountain climbing point at the adjacent point of the position table; thirdly, if the correlation value of the adjacent points is smaller than that of the current point, recording the maximum correlation value and the corresponding position, otherwise, updating the position of the current mountain climbing point by the next position, and calculating the correlation value; and finally, finishing all initial mountain climbing point searches to obtain the maximum position of the correlation value and the correlation value. The matching adopts a mean value removing normalization correlation algorithm as follows:
simplification of the above equation yields:
wherein:
and if the correlation value is larger than the set threshold value T, the matching is considered to be successful, the position and the correlation value of the target in the real-time image can be obtained, the target tracking process is switched to according to the updating of the target template, otherwise, the target position prediction is carried out by utilizing Kalman or attitude information, and then the target tracking is carried out by re-acquiring the real-time image of the target.
In order to improve the tracking stability, reduce the tracking accumulated error and suppress the tracking drift, the current matching optimal region and the current target template are weighted and combined according to a certain criterion according to the similarity degree of the current matching optimal region and the target template to form a new template for matching the next frame of image, so as to achieve a better tracking effect. The target image part should be focused during the target updating process, and the traditional target template updating does not distinguish the target from the background. In practical implementation, firstly, the template image is binarized and 3 × 3 expansion is performed to adapt to the small deformation capability of the target, and when the template is updated, alpha is performed on the non-zero point of the target template binary mask image 1 Updating the weight value, and adopting alpha for the pixel point with the value of zero 2 And updating the weight value to achieve the aims of delaying the background to carry out the target template and quickly updating the real target template graph.
Wherein Bin k (x, y) denotes the mask map after binary expansion of the target template, T k+1 For updated target template graphs, T k Target template graph used for current matching, I k Obtaining an optimal target template graph for the current matching; alpha is alpha 1 =λ 1 A,α 2 =λ 2 A, and satisfies λ 1 >λ 2 And A is an interframe confidence coefficient calculated according to a continuous multiframe matching result:
and matching the next frame of image by using the updated new target template, and then realizing target tracking in the cyclic step of repeatedly entering 'template matching-correlation value calculation-current best matching position acquisition-target template updating-template matching'.
The method can effectively avoid the false target, meet the real-time tracking, reduce the interference of the false target to the template, and has the advantages of reliable target identification and tracking, low requirement on hardware and the like.
Claims (10)
1. The high-precision infrared image tracker is characterized by taking a programmable logic array and a digital processing chip as a framework, and comprising a real-time image acquisition processing module, a real-time image storage module, a control module FPGA, an output image storage module, a communication module, a digital processing module DSP and a dynamic storage module SDRAM; the real-time image acquisition processing module comprises an analog-to-digital conversion module A/D and a programmable logic array FPGA, the real-time image storage module comprises static memories SRAM1 and SRAM2, the output image storage module comprises static memories SRAM3 and SRAM4, the control module FPGA is configured with a power supply and an erasable memory EPROM, the digital processing module DSP is configured with a power supply and FLASH, and the communication module is configured with RS232 and RS422 communication serial ports; the erasable memory EPROM and the FLASH are used for storing programs for running the chip, and the dynamic storage module SDRAM provides data storage for the running of the digital processing module DSP;
the real-time image acquisition module acquires a target real-time image subjected to A/D conversion through the control module FPGA, the image is stored in the real-time image storage module through a ping-pong technology, namely, the control module FPGA firstly acquires a frame of image and stores the frame of image in the SRAM1 of the real-time image storage module to generate image interruption, the DSP responds to the interruption, image data in the SRAM1 is taken away through the EDMA to be analyzed and processed, then the processed image is transmitted to the SRAM3 in the output image storage module, and the image is displayed through a D/A video under the control of the control module FPGA. When the digital processing module DSP generates interruption, the control module FPGA stores the next acquired frame image into the SRAM2 in the real-time image storage module, when the digital processing module DSP finishes processing and analyzing the frame data and the control module FPGA finishes acquiring the next frame image and generates interruption, the digital processing module DSP takes the data in the SRAM2 away through an EDMA, transmits the data to an SRAM4 in the output image storage module after the data is analyzed and processed, and displays the data through a D/A video under the control of the control module FPGA; and repeating the steps in a circulating way, and after the digital processing module DSP processes one frame of image, transmitting the image and the information of the tracking position to the fire control system through the communication module.
2. The high-precision infrared image tracker according to claim 1, wherein said control module FPGA roughly selects a target region from the real-time image collected by said real-time image collecting module and stores the selected target region in said real-time image storing module.
3. The high-precision infrared image tracker according to claim 1 or 2, characterized in that said digital processing module DSP performs an analysis process by performing a binarization process on a target area stored in a real-time image storage module, and using an eight-neighborhood labeling method to obtain a true centroid position of the target area; and selecting a target area with the same size as that stored in the real-time image storage module as an original target template drawing according to the real centroid position, and carrying out correlation matching on the original target template drawing and the target real-time image until the real position of the target in the real-time image is obtained.
4. The target fast tracking method based on the high-precision infrared image tracker of claim 1 is characterized by comprising the following specific implementation steps:
step one, roughly selecting a target area in a collected real-time image;
secondly, performing binarization processing on the target area selected in the first step, obtaining the centroid position of the target area by using an eight-neighborhood labeling method, and judging whether the centroid position is the real centroid position of the target or not; if yes, selecting the target area with the same size as the step one as the original target template picture by using the real centroid position; if not, selecting the target area by the method of the step again, and carrying out binarization processing and marking until the real centroid position of the target is obtained;
step three, the original target template image in the step two is matched with the target real-time image in a relevant manner, if the matching is successful, the real position of the target in the real-time image can be obtained, the target template is updated according to the matching result, and the target tracking is switched in; and if the matching fails, performing Kalman or attitude information target prediction, and acquiring a target real-time image again to perform target matching identification so as to obtain a real target.
5. The method for rapidly tracking the target according to claim 4, wherein when the target region is subjected to target extraction, a gray value which maximizes the variance between the target and the background is obtained by using an improved maximum inter-class variance method.
6. The method of claim 5, wherein the gray level in the target region greater than the gray level is 255 and vice versa is 0.
7. The method for quickly tracking the target according to claim 4, wherein eight neighborhood markers are applied to a plurality of connected regions appearing after the binarization processing of the target region to distinguish a real target from a false target, the connected regions which have passed the eight neighborhood markers are combined, finally, the connected region whose area is closest to the pixel imaged by the target is selected, the centroid position of the connected region is calculated, the centroid position is taken as a reference, a region with the same size as that in the step one is selected as the target region from the real-time image of the target, and the target region obtained at this time is the target template map, i.e. the target which needs to be accurately hit.
8. The method for fast tracking the target of claim 4, wherein the real-time image of the target and the original target template map are represented by gray value matrix as S (M, N) and T (M, N), respectively, in step three, where M is M>m,N>, n; the target template T (M, N) is translated on the real-time image S (M, N), and the area of the real-time image covered by the template is a subgraph S u,v Wherein (u, v) is the coordinate of the subgraph with the top left corner as the origin, called the reference point, and a measure is calculated on the real-time image reference point to characterize the size of the difference program between the target template graph and the corresponding subgraph, called the correlation value; the greater the correlation value is, the higher the similarity degree between the target template graph and the corresponding subgraph is, and if the correlation value is greater than a threshold value T established in advance, the matching is considered to be successful.
9. The method for fast tracking of the target according to claim 8, wherein the correlation matching between the original target template map and the target real-time image in step three is implemented by using an improved hill-climbing search strategy and a simplified correlation metric, and the specific method is a mean normalization algorithm, which is:
a simplification of the above equation can be obtained:
wherein:
and if the correlation value is greater than the set threshold value T, the matching is considered to be successful, the position and the correlation value of the target in the real-time image can be obtained, the target tracking process is switched to according to the updating of the target template, otherwise, the target position prediction is carried out by using Kalman or attitude information, and then the target tracking is carried out by switching to the target real-time image acquisition again.
10. The method for fast tracking of the target as claimed in claim 4, characterized in that in step three, the original target template image is first binarized and 3 x 3 dilated to adapt to the small deformation capability of the target, and when updating the template, the non-zero point of the target template binary mask image is alpha-updated 1 Updating the weight value, and adopting alpha for the pixel point with the value of zero 2 Updating the weight value to achieve the purposes of delaying the background to carry out the target template and quickly updating the real target template graph;
wherein Bin k (x, y) denotes the mask map after binary expansion of the target template, T k+1 For updated target template graphs, T k Target template graph used for current matching, I k Obtaining an optimal target template graph for the current matching; alpha is alpha 1 =λ 1 A,α 2 =λ 2 A, and satisfies λ 1 >λ 2 And A is an interframe confidence coefficient calculated according to a continuous multiframe matching result:
and matching the next frame of image by using the updated new target template, and then repeatedly entering the cyclic step of template matching, calculating a correlation value, obtaining the current best matching position, target template updating and template matching to realize target tracking.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711343812.6A CN107945215B (en) | 2017-12-14 | 2017-12-14 | High-precision infrared image tracker and target rapid tracking method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711343812.6A CN107945215B (en) | 2017-12-14 | 2017-12-14 | High-precision infrared image tracker and target rapid tracking method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107945215A true CN107945215A (en) | 2018-04-20 |
CN107945215B CN107945215B (en) | 2021-07-23 |
Family
ID=61943355
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711343812.6A Active CN107945215B (en) | 2017-12-14 | 2017-12-14 | High-precision infrared image tracker and target rapid tracking method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107945215B (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109598683A (en) * | 2018-11-12 | 2019-04-09 | 平安科技(深圳)有限公司 | Image display method, apparatus and computer readable storage medium |
CN110084822A (en) * | 2019-05-05 | 2019-08-02 | 中国人民解放军战略支援部队航天工程大学 | A kind of target acquisition real time processing system and method towards the in-orbit application of satellite |
CN110130987A (en) * | 2019-04-19 | 2019-08-16 | 同济大学 | A kind of tunnel convergence deformation monitoring method based on image analysis |
CN110796687A (en) * | 2019-10-30 | 2020-02-14 | 电子科技大学 | Sky background infrared imaging multi-target tracking method |
CN110895120A (en) * | 2018-09-13 | 2020-03-20 | 中国舰船研究设计中心 | Image processing technology-based ship cannon system precision detection device and detection method |
CN112257505A (en) * | 2020-09-17 | 2021-01-22 | 昆明物理研究所 | False target identification method applied to infrared panoramic system and storage medium |
CN112287846A (en) * | 2020-10-30 | 2021-01-29 | 深圳市优必选科技股份有限公司 | Target person following method, device, mobile robot and readable storage medium |
CN112489091A (en) * | 2020-12-18 | 2021-03-12 | 湖南华南光电(集团)有限责任公司 | Full strapdown image seeker target tracking method based on direct-aiming template |
CN114782367A (en) * | 2022-04-24 | 2022-07-22 | 广东天太机器人有限公司 | Control system and method for mechanical arm |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2005031382A2 (en) * | 2003-05-23 | 2005-04-07 | Lockheed Martin Corporation | Real-time multistage infrared image-based tracking system |
CN101105836A (en) * | 2007-07-06 | 2008-01-16 | 哈尔滨工程大学 | Infrared image real-time target identification and tracking device for movement background, and the method |
CN101620671A (en) * | 2009-08-14 | 2010-01-06 | 华中科技大学 | Method for indirectly positioning and identifying three-dimensional buildings by using riverway landmarks |
CN102163280A (en) * | 2011-04-12 | 2011-08-24 | 华中科技大学 | Method for identifying, tracking and converting target based on confidence degree and multi-frame judgement |
CN201966922U (en) * | 2011-03-08 | 2011-09-07 | 上海鹏燕矿业安全设备制造有限公司 | Ultrasonic carrier information communication system |
CN102646279A (en) * | 2012-02-29 | 2012-08-22 | 北京航空航天大学 | Anti-shielding tracking method based on moving prediction and multi-sub-block template matching combination |
CN202422217U (en) * | 2011-12-20 | 2012-09-05 | 中国人民解放军总装备部军械技术研究所 | Infrared target acquisition and recognition tracker |
CN103391430A (en) * | 2013-07-22 | 2013-11-13 | 南通大学 | DSP (digital signal processor) based relevant tracking method and special device |
CN104038744A (en) * | 2013-07-22 | 2014-09-10 | 南通大学 | Correlation tracking method based on DSP (digital signal processor) with high tracking success rate |
-
2017
- 2017-12-14 CN CN201711343812.6A patent/CN107945215B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2005031382A2 (en) * | 2003-05-23 | 2005-04-07 | Lockheed Martin Corporation | Real-time multistage infrared image-based tracking system |
CN101105836A (en) * | 2007-07-06 | 2008-01-16 | 哈尔滨工程大学 | Infrared image real-time target identification and tracking device for movement background, and the method |
CN101620671A (en) * | 2009-08-14 | 2010-01-06 | 华中科技大学 | Method for indirectly positioning and identifying three-dimensional buildings by using riverway landmarks |
CN201966922U (en) * | 2011-03-08 | 2011-09-07 | 上海鹏燕矿业安全设备制造有限公司 | Ultrasonic carrier information communication system |
CN102163280A (en) * | 2011-04-12 | 2011-08-24 | 华中科技大学 | Method for identifying, tracking and converting target based on confidence degree and multi-frame judgement |
CN202422217U (en) * | 2011-12-20 | 2012-09-05 | 中国人民解放军总装备部军械技术研究所 | Infrared target acquisition and recognition tracker |
CN102646279A (en) * | 2012-02-29 | 2012-08-22 | 北京航空航天大学 | Anti-shielding tracking method based on moving prediction and multi-sub-block template matching combination |
CN103391430A (en) * | 2013-07-22 | 2013-11-13 | 南通大学 | DSP (digital signal processor) based relevant tracking method and special device |
CN104038744A (en) * | 2013-07-22 | 2014-09-10 | 南通大学 | Correlation tracking method based on DSP (digital signal processor) with high tracking success rate |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110895120A (en) * | 2018-09-13 | 2020-03-20 | 中国舰船研究设计中心 | Image processing technology-based ship cannon system precision detection device and detection method |
CN110895120B (en) * | 2018-09-13 | 2022-04-08 | 中国舰船研究设计中心 | Image processing technology-based ship cannon system precision detection device and detection method |
CN109598683A (en) * | 2018-11-12 | 2019-04-09 | 平安科技(深圳)有限公司 | Image display method, apparatus and computer readable storage medium |
CN110130987A (en) * | 2019-04-19 | 2019-08-16 | 同济大学 | A kind of tunnel convergence deformation monitoring method based on image analysis |
CN110084822A (en) * | 2019-05-05 | 2019-08-02 | 中国人民解放军战略支援部队航天工程大学 | A kind of target acquisition real time processing system and method towards the in-orbit application of satellite |
CN110796687A (en) * | 2019-10-30 | 2020-02-14 | 电子科技大学 | Sky background infrared imaging multi-target tracking method |
CN112257505A (en) * | 2020-09-17 | 2021-01-22 | 昆明物理研究所 | False target identification method applied to infrared panoramic system and storage medium |
CN112257505B (en) * | 2020-09-17 | 2023-07-04 | 昆明物理研究所 | False target identification method and storage medium applied to infrared panoramic system |
CN112287846A (en) * | 2020-10-30 | 2021-01-29 | 深圳市优必选科技股份有限公司 | Target person following method, device, mobile robot and readable storage medium |
CN112287846B (en) * | 2020-10-30 | 2024-05-07 | 深圳市优必选科技股份有限公司 | Target person following method, device, movable robot and readable storage medium |
CN112489091A (en) * | 2020-12-18 | 2021-03-12 | 湖南华南光电(集团)有限责任公司 | Full strapdown image seeker target tracking method based on direct-aiming template |
CN114782367A (en) * | 2022-04-24 | 2022-07-22 | 广东天太机器人有限公司 | Control system and method for mechanical arm |
Also Published As
Publication number | Publication date |
---|---|
CN107945215B (en) | 2021-07-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107945215B (en) | High-precision infrared image tracker and target rapid tracking method | |
CN112308881B (en) | Ship multi-target tracking method based on remote sensing image | |
CN104282020B (en) | A kind of vehicle speed detection method based on target trajectory | |
CN108921873B (en) | Markov decision-making online multi-target tracking method based on kernel correlation filtering optimization | |
CN105792353B (en) | Crowd sensing type WiFi signal fingerprint assisted image matching indoor positioning method | |
CN111781608B (en) | Moving target detection method and system based on FMCW laser radar | |
CN109636829A (en) | A kind of multi-object tracking method based on semantic information and scene information | |
CN104867137B (en) | A kind of method for registering images based on improvement RANSAC algorithms | |
CN114299417A (en) | Multi-target tracking method based on radar-vision fusion | |
CN114677554A (en) | Statistical filtering infrared small target detection tracking method based on YOLOv5 and Deepsort | |
CN107798691B (en) | A kind of unmanned plane independent landing terrestrial reference real-time detection tracking of view-based access control model | |
CN110084830B (en) | Video moving object detection and tracking method | |
CN101770583B (en) | Template matching method based on global features of scene | |
CN111340845B (en) | Automatic tracking method, system, terminal and medium based on depth vision sensor | |
CN108564598A (en) | A kind of improved online Boosting method for tracking target | |
CN109146918A (en) | A kind of adaptive related objective localization method based on piecemeal | |
CN104391294A (en) | Connection domain characteristic and template matching based radar plot correlation method | |
CN110826575A (en) | Underwater target identification method based on machine learning | |
CN112197705A (en) | Fruit positioning method based on vision and laser ranging | |
CN116381672A (en) | X-band multi-expansion target self-adaptive tracking method based on twin network radar | |
CN111744161A (en) | Table tennis falling point detection and edge ball wiping judgment system | |
CN113781523B (en) | Football detection tracking method and device, electronic equipment and storage medium | |
CN110956083A (en) | Bohai sea ice drift remote sensing detection method based on high-resolution four-signal optical satellite | |
CN104537690B (en) | Moving point target detection method based on maximum value-time index combination | |
CN108985216B (en) | Pedestrian head detection method based on multivariate logistic regression feature fusion |
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 |