CN112164109B - Coordinate correction method, coordinate correction device, storage medium and electronic device - Google Patents
Coordinate correction method, coordinate correction device, storage medium and electronic device Download PDFInfo
- Publication number
- CN112164109B CN112164109B CN202010653896.9A CN202010653896A CN112164109B CN 112164109 B CN112164109 B CN 112164109B CN 202010653896 A CN202010653896 A CN 202010653896A CN 112164109 B CN112164109 B CN 112164109B
- Authority
- CN
- China
- Prior art keywords
- original
- target
- picture
- target point
- 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.)
- Active
Links
- 238000012937 correction Methods 0.000 title description 35
- 238000000034 method Methods 0.000 title description 33
- 238000012545 processing Methods 0.000 description 17
- 238000013528 artificial neural network Methods 0.000 description 14
- 238000001514 detection method Methods 0.000 description 13
- 238000010586 diagram Methods 0.000 description 13
- 238000004590 computer program Methods 0.000 description 12
- 238000013527 convolutional neural network Methods 0.000 description 7
- 230000005540 biological transmission Effects 0.000 description 5
- 238000004891 communication Methods 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012805 post-processing Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 239000000779 smoke Substances 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
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
- G06T7/74—Determining position or orientation of objects or cameras using feature-based methods involving reference images or patches
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the invention provides a coordinate correction method, a device, a storage medium and an electronic device, wherein the method comprises the following steps: detecting coordinates of a target point included in a target picture, wherein the target picture is a picture with a size suitable for a target neural network, which is obtained by carrying out reduction processing on an original picture according to a reduction ratio; determining a feature map of a target area of the target point based on the coordinates of the target point; determining a regional feature map of a plurality of original points included in a neighborhood corresponding to the target point in the original picture; and carrying out coordinate correction on the original picture based on the characteristic diagram of the target point and the characteristic diagrams of the plurality of original points. The invention solves the problem of low coordinate correction precision in the related technology and achieves the effect of improving the correction precision.
Description
Technical Field
The embodiment of the invention relates to the field of communication, in particular to a coordinate correction method, a coordinate correction device, a storage medium and an electronic device.
Background
The image target detection technology has important roles in image processing neighborhood, and the basis is tamped for subsequent image recognition, behavior analysis and event analysis. It is particularly important today that artificial intelligence and image data grow up. Detection of targets involves many aspects, human, animal, plant, vehicle, fire, water, smoke, and the like. The accuracy of detection is just the gauge post of detection performance measurement, especially the detection of little target, and the accuracy problem is the decisive factor directly. In the related art, the coordinate precision of a detection network is improved, and network performance improvement, a subsequent correction module and early image processing are mainly considered.
In the related art, for the link made by correcting the network detection coordinates, the link of image input can be enhanced, so that the characteristic enhancement is achieved, and the target detection is more accurate. For example, various measures are added to the convolutional neural network (Convolutional Neural Networks, abbreviated as CNN) part to optimize the network and improve the performance. There are also post-processing modules that re-correct the coordinates after the detected coordinates are obtained. However, in practical applications, especially embedded devices, real-time networks are scaled to the CNN network because of time-consuming problems. The resulting detection coordinates are obtained at the resolution of the input network. And then the coordinates of the input network are converted into the coordinates of the original image. Since image scaling is lossy, this loss may cause a loss in detection accuracy.
As is clear from this, the related art has a problem of low coordinate correction accuracy.
In view of the above problems in the related art, no effective solution has been proposed at present.
Disclosure of Invention
The embodiment of the invention provides a coordinate correction method, a coordinate correction device, a storage medium and an electronic device, which are used for at least solving the problem of low coordinate correction precision in the related technology.
According to an embodiment of the present invention, there is provided a coordinate correcting method including: detecting coordinates of a target point included in a target picture, wherein the target picture is a picture with a size suitable for a target neural network, which is obtained by carrying out reduction processing on an original picture according to a reduction ratio; determining a feature map of a target area of the target point based on the coordinates of the target point; determining a regional feature map of a plurality of original points included in a neighborhood corresponding to the target point in the original picture; and carrying out coordinate correction on the original picture based on the characteristic diagram of the target point and the characteristic diagrams of the plurality of original points.
According to another embodiment of the present invention, there is provided a coordinate correcting apparatus including: the detection module is used for detecting coordinates of a target point included in a target picture, wherein the target picture is a picture with a size suitable for a target neural network, which is obtained by carrying out reduction processing on an original picture according to a reduction ratio; a first determining module, configured to determine a feature map of a target area of the target point based on coordinates of the target point; a second determining module, configured to determine a region feature map of a plurality of original points included in a neighborhood corresponding to the target point in the original picture; and the correction module is used for carrying out coordinate correction on the original picture based on the characteristic diagram of the target point and the characteristic diagrams of the plurality of original points.
According to a further embodiment of the invention, there is also provided a computer readable storage medium having stored therein a computer program, wherein the computer program is arranged to perform the steps of any of the method embodiments described above when run.
According to a further embodiment of the invention, there is also provided an electronic device comprising a memory having stored therein a computer program and a processor arranged to run the computer program to perform the steps of any of the method embodiments described above.
According to the method, after the coordinates of the target points in the image with the size suitable for the target neural network after the shrinking processing is detected, the characteristic images of the target points are determined, the characteristic images of a plurality of original points included in the field corresponding to the target points in the original image are determined, and the coordinates of the original image are corrected according to the characteristic images of the target points and the characteristic images of the plurality of original points. And carrying out coordinate correction on the original picture by utilizing the characteristic diagram of the target point and the characteristic diagrams of a plurality of original points, so that the coordinate error introduced by lossy scaling can be compensated. Therefore, the problem of low correction accuracy of the coordinates in the related art can be solved, and the effect of improving the correction accuracy can be achieved.
Drawings
Fig. 1 is a block diagram of a hardware structure of a mobile terminal of a coordinate correcting method according to an embodiment of the present invention;
FIG. 2 is a flow chart of a coordinate correction method according to an embodiment of the present invention;
FIG. 3 is a flow chart of a coordinate correcting method according to an embodiment of the present invention;
fig. 4 is a block diagram of a coordinate correcting apparatus according to an embodiment of the present invention.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings in conjunction with the embodiments.
It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order.
The method embodiments provided in the embodiments of the present application may be performed in a mobile terminal, a computer terminal or similar computing device. Taking the mobile terminal as an example, fig. 1 is a block diagram of a hardware structure of the mobile terminal according to an embodiment of the present application. As shown in fig. 1, a mobile terminal may include one or more (only one is shown in fig. 1) processors 102 (the processor 102 may include, but is not limited to, a microprocessor MCU or a processing device such as a programmable logic device FPGA) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 for communication functions and an input-output device 108. It will be appreciated by those skilled in the art that the structure shown in fig. 1 is merely illustrative and not limiting of the structure of the mobile terminal described above. For example, the mobile terminal may also include more or fewer components than shown in fig. 1, or have a different configuration than shown in fig. 1.
The memory 104 may be used to store a computer program, for example, a software program of application software and a module, such as a computer program corresponding to a coordinate correcting method in an embodiment of the present invention, and the processor 102 executes the computer program stored in the memory 104 to perform various functional applications and data processing, that is, implement the above-mentioned method. Memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include memory remotely located relative to the processor 102, which may be connected to the mobile terminal via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The transmission means 106 is arranged to receive or transmit data via a network. Specific examples of the network described above may include a wireless network provided by a communication provider of the mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, simply referred to as a NIC) that can connect to other network devices through a base station to communicate with the internet. In one example, the transmission device 106 may be a Radio Frequency (RF) module, which is used to communicate with the internet wirelessly.
In this embodiment, a coordinate correcting method is provided, fig. 2 is a flowchart of the coordinate correcting method according to an embodiment of the present invention, and as shown in fig. 2, the flowchart includes the following steps:
step S202, detecting coordinates of a target point included in a target picture, wherein the target picture is a picture with a size suitable for a target neural network, which is obtained by performing reduction processing on an original picture according to a reduction ratio;
step S204, determining a feature map of a target area of the target point based on the coordinates of the target point;
step S206, determining a regional feature map of a plurality of original points included in a neighborhood corresponding to the target point in the original picture;
And step S208, carrying out coordinate correction on the original picture based on the characteristic diagram of the target point and the characteristic diagrams of a plurality of original points.
In the above embodiment, the original picture may be a picture taken by the photographing apparatus, or an input picture, or a picture extracted from a multimedia file. The target neural network may be a convolutional neural network CNN, but may also be other neural networks.
In the above embodiment, the resolution of the original picture may be w ori,hori, the size input to the target neural network may be w net,hnet, the scale may be ratio x,ratioy, and the coordinates of the target point included in the detected target picture may be x no,yno,wno,hno. Then, a feature map Feat n of the target point can be determined according to the coordinates of the target point, feature maps of a plurality of original points included in a neighborhood corresponding to the target point in the original picture are determined, and the original picture is subjected to coordinate correction according to the two feature maps, namely, the coordinates of the target point included in the detected target picture are converted into the coordinates x oo,yoo,woo,hoo of the original picture resolution. The coordinates of the target points may be coordinates of one fixed point, and since the offsets of all the points are practically identical, it is not necessary to calculate the coordinates of a plurality of points, and only one fixed point is used for coordinate mapping, so that the calculation time is saved. When determining the feature, the whole area or the local area where the target is located can be used as the feature calculation range.
Alternatively, the main body of execution of the above steps may be a background processor, or other devices with similar processing capability, and may also be a machine integrated with at least an image acquisition device and a data processing device, where the image acquisition device may include a graphics acquisition module such as a camera, and the data processing device may include a terminal such as a computer, a mobile phone, and the like, but is not limited thereto.
According to the method and the device, after the coordinates of the target points in the image with the size suitable for the target neural network after the size subjected to the shrinking treatment are detected, the characteristics of the target points are determined, the characteristics of a plurality of original points included in the field corresponding to the target points in the original image are determined, and the coordinates of the original image are corrected according to the characteristics of the target points and the characteristics of the plurality of original points. And carrying out coordinate correction on the original picture by utilizing the characteristics of the target point and the characteristics of a plurality of original points, so that the coordinate error introduced by lossy scaling can be compensated. Therefore, the problem of low correction accuracy of the coordinates in the related art can be solved, and the effect of improving the correction accuracy can be achieved.
In an exemplary embodiment, determining the region characteristics of the plurality of original points included in the neighborhood corresponding to the target point in the original picture includes: determining a neighborhood corresponding to the target point in the original picture based on the reduction ratio; and determining a characteristic map of a plurality of the origin points included in the neighborhood. In this embodiment, the original picture shrinkage may be that the matrix information of the original neighborhood ratio x*ratioy is represented by a point on the corresponding target picture, so that the neighborhood corresponding to the target point in the original picture may be determined according to the shrinkage ratio, and the feature map of the neighborhood corresponding to the target point in the original picture may be calculated. The reduction ratio may be set manually, or may be calculated according to the resolution, size, etc. of the original picture and the resolution, size, etc. of the input picture required by the target neural network.
In an exemplary embodiment, before the coordinate correction is performed on the original picture based on the feature map of the target point and the feature maps of the plurality of original points, the method further includes: calculating a first original point corresponding to the target point in the original image based on the reduction scale; the coordinate correction of the original picture based on the feature map of the target point and the feature maps of the plurality of original points includes: determining a feature map of a second original point with the greatest similarity with the feature map of the target point from the feature maps of the plurality of original points; and correcting the coordinates of the first original point to the coordinate position of the second original point. In this embodiment, the coordinates of each point in the neighborhood of the ratio x*ratioy in the original picture may be determined according to the reduction ratio, and the neighborhood around x and y in the original picture corresponds to the coordinates of one point in the target picture. The ratio x*ratioy of each point is recorded into x*ratioy integral graphs, calculated point by point, and finally obtained into x*ratioy characteristic graph values of the ratio, feat o0,Feato1,Feato2....,Featog, wherein g=ratio x*ratioy. And comparing the similarity of the values Feat n and Feat o0,Feato1,Feato2....,Featog to obtain the most similar value Feat o?, and then translating the coordinates x and y of the original picture to the coordinates where Feat o? is located to finish coordinate correction. Wherein. The coordinates of each point in the neighborhood of ratio x*ratioy in the original picture may be
The feature map value can be calculated by the following method:
in the first method, the area where the small image object (corresponding to the object image) is located or a certain representative area of the area, the sum of all points and the sum nri0,sumnri1…sumnrin of the pixels of the area where the original image starts corresponding to each offset point are corrected, and the offset area represented by the value closest to the sum is Feat, which is the fastest.
In the second method, the target area or a certain representative area of the area in the small image, and the frame difference of each corresponding image represented by each offset point corresponding to the original image are subabs 0,subabs1…subabsn, and the position of the candidate with the minimum frame difference, namely the position to be corrected, is subabs, namely Feat.
Thirdly, solving the point (x, y) with the largest gradient in the small map, intercepting a certain area to calculate a gradient map Grad in the field near the point, (x, y) is mapped to a large map coordinate (x ori,yori), gradient maps Grad ori0,gradori1…gradorin of the x*ratioy large maps are taken, and Grad ori? with the smallest error is found, wherein the Grad is Feat.
In the fourth method, an integral graph intelg is made in a region where a target is located in a small graph or in a certain representative region of the region, integral graphs intel ori0,Integori1…Integorin of corresponding graphs are made in each corresponding offset point corresponding to the original graph, various features eg, haar and Hog can be calculated on the integral graph, and the nearest positions in the features are compared, namely the positions to be corrected, namely feat _haar= Feat or feat _hog= Feat.
In an exemplary embodiment, before detecting the coordinates of the target point included in the received target picture, the method further comprises: determining the reduction scale based on the resolution of the original picture and the resolution of the picture supported by the target neural network; and carrying out reduction processing on the original picture according to the reduction ratio so as to obtain the target picture. The reduction ratio may be determined according to the resolution of the original picture and the resolution of the picture supported by the target neural network, and after the reduction ratio is determined, the reduction processing is performed according to a plurality of original pictures in the reduction ratio.
The following describes the coordinate correction with reference to specific embodiments:
FIG. 3 is a flowchart of a coordinate correcting method according to an embodiment of the present invention, as shown in FIG. 3, the flowchart includes:
and 0, scaling the original picture. For example, the resolution of the original picture is w ori,hori, the input size wnet, h net of the scaled picture to the deep learning network (corresponding to the target neural network described above). The scale is ratio x,ratioy.
And step 1, carrying out enhancement processing on the zoomed picture to obtain a target picture.
And 2, inputting the target picture into a CNN network.
And 3, correcting the target picture by the post-processing module. The coordinates x no,yno,wno,hno of the target picture are detected,
And 4, mapping coordinates, namely mapping the target point in the target picture with coordinates of each point in the field corresponding to the target point in the original picture.
And 5, calculating a characteristic diagram of a target point in the target picture, feat n.
In step 6, the matrix information of the neighborhood ratio x*ratioy of the large picture (corresponding to the original picture) is represented by a point on the corresponding small picture (corresponding to the target picture), and the matrix information is reversely pushed to the coordinate point by coordinate point of the original picture according to the scaling method, each point of the neighborhood of the ratio x*ratioy is taken,
The neighborhood around x, y in the original map corresponds to a point in the network input map. The ratio x*ratioy of each point is recorded into x*ratioy feature maps, calculated point by point, and finally obtained are x*ratioy feature map values of the ratio Feat o0,Feato1,Feato2....,Featog,g＝ratiox*ratioy. The feature map value can be calculated by the following method:
In the first method, the area where the small image target is located or a certain representative area of the area, the sum nri0,sumnri1…sumnrin of all points and the area pixel sum which is started by each offset point corresponding to the original image, the offset area represented by the value closest to the sum is the corrected area, and the sum is Feat, which is the fastest.
In the second method, the target area or a certain representative area of the area in the small image, and the frame difference of each corresponding image represented by each offset point corresponding to the original image are subabs 0,subabs1…subabsn, and the position of the candidate with the minimum frame difference, namely the position to be corrected, is subabs, namely Feat.
Thirdly, solving the point (x, y) with the largest gradient in the small map, intercepting a certain area to calculate a gradient map Grad in the field near the point, (x, y) is mapped to large map coordinates (xori, yori), gradient maps Grad ori0,gradori1…gradorin of the ratio x*ratioy large maps are obtained, and Grad ori? with the smallest error is found, wherein the Grad is Feat.
In the fourth method, an integral graph intelg is made in a region where a target is located in a small graph or in a certain representative region of the region, integral graphs intel ori0,Integori1…Integorin of corresponding graphs are made in each corresponding offset point corresponding to the original graph, various features eg, haar and Hog can be calculated on the integral graph, and the nearest positions in the features are compared, namely the positions to be corrected, namely feat _haar= Feat or feat _hog= Feat.
In step 7, the values Feat n and Feat o0,Feato1,Feato2....,Featog are compared for similarity to obtain the most similar value Feat o?, and then the coordinates x, y of the artwork are translated to the coordinates where Feat o? is located.
It should be noted that, the steps 0-4 describe a flow of the deep learning detection network, wherein, the steps 0, 1 and 3 are optional steps, and when the requirement for coordinate correction is high, the steps 0, 1 and 3 can be selectively executed, so that the precision of coordinate correction is improved. When the requirement for coordinate correction is not high, the steps 0, 1 and 3 can be selectively executed, so that the execution speed is improved.
From the description of the above embodiments, it will be clear to a person skilled in the art that the method according to the above embodiments may be implemented by means of software plus the necessary general hardware platform, but of course also by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (e.g. ROM/RAM, magnetic disk, optical disk) comprising instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to perform the method according to the embodiments of the present invention.
The embodiment also provides a coordinate correcting device, which is used for realizing the embodiment and the preferred implementation, and is not described in detail. As used below, the term "module" may be a combination of software and/or hardware that implements a predetermined function. While the means described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and contemplated.
Fig. 4 is a block diagram showing a configuration of a coordinate correcting apparatus according to an embodiment of the present invention, as shown in fig. 4, the apparatus including:
The detection module 42 is configured to detect coordinates of a target point included in a target image, where the target image is an image with a size suitable for a target neural network, which is obtained by performing reduction processing on an original image according to a reduction ratio;
A first determining module 44, configured to determine a feature map of a target area of the target point based on coordinates of the target point;
a second determining module 46, configured to determine a region feature map of a plurality of original points included in a neighborhood corresponding to the target point in the original picture;
And a correction module 48, configured to perform coordinate correction on the original picture based on the feature map of the target point and the feature maps of the plurality of original points.
In one exemplary embodiment, the second determination module 46 includes: a first determining unit, configured to determine a neighborhood corresponding to the target point in the original picture based on the reduction ratio; and a second determining unit configured to determine feature maps of the plurality of origin points included in the neighborhood.
In an exemplary embodiment, the apparatus further comprises: a calculating module, configured to calculate, based on the reduction scale, a first origin corresponding to the target point in the original image before performing coordinate correction on the original image based on the feature image of the target point and the feature images of the plurality of origins; the correction module 48 includes: a third determining unit configured to determine, from among the feature maps of the plurality of original points, a feature map of a second original point having a maximum similarity with the feature map of the target point; and a correction unit configured to correct the coordinates of the first origin to the coordinate positions of the second origin.
In an exemplary embodiment, the apparatus further comprises: a third determining module, configured to determine the reduction scale based on a resolution of the original picture and a resolution of a picture supported by the target neural network; and the processing module is used for carrying out reduction processing on the original picture according to the reduction ratio so as to obtain the target picture.
It should be noted that each of the above modules may be implemented by software or hardware, and for the latter, it may be implemented by, but not limited to: the modules are all located in the same processor; or the above modules may be located in different processors in any combination.
Embodiments of the present invention also provide a computer readable storage medium having a computer program stored therein, wherein the computer program is arranged to perform the steps of any of the method embodiments described above when run.
In one exemplary embodiment, the computer readable storage medium may include, but is not limited to: a usb disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory RAM), a removable hard disk, a magnetic disk, or an optical disk, or other various media capable of storing a computer program.
An embodiment of the invention also provides an electronic device comprising a memory having stored therein a computer program and a processor arranged to run the computer program to perform the steps of any of the method embodiments described above.
In an exemplary embodiment, the electronic apparatus may further include a transmission device connected to the processor, and an input/output device connected to the processor.
Specific examples in this embodiment may refer to the examples described in the foregoing embodiments and the exemplary implementation, and this embodiment is not described herein.
It will be appreciated by those skilled in the art that the modules or steps of the invention described above may be implemented in a general purpose computing device, they may be concentrated on a single computing device, or distributed across a network of computing devices, they may be implemented in program code executable by computing devices, so that they may be stored in a storage device for execution by computing devices, and in some cases, the steps shown or described may be performed in a different order than that shown or described herein, or they may be separately fabricated into individual integrated circuit modules, or multiple modules or steps of them may be fabricated into a single integrated circuit module. Thus, the present invention is not limited to any specific combination of hardware and software.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the principle of the present invention should be included in the protection scope of the present invention.
Claims (8)
1. A coordinate correction method, comprising:
Detecting coordinates of a target point included in a target picture, wherein the target picture is a picture with a size suitable for a target neural network, which is obtained by carrying out reduction processing on an original picture according to a reduction ratio;
Determining a feature map of a target area of the target point based on the coordinates of the target point;
determining a regional feature map of a plurality of original points included in a neighborhood corresponding to the target point in the original picture;
carrying out coordinate correction on the original picture based on the characteristic diagram of the target point and the characteristic diagrams of a plurality of original points;
before the coordinate correction is performed on the original picture based on the feature map of the target point and the feature maps of the plurality of original points, the method further includes: calculating a first original point corresponding to the target point in the original picture based on the reduction ratio;
The coordinate correction of the original picture based on the feature map of the target point and the feature maps of the plurality of original points includes: determining a feature map of a second original point with the greatest similarity with the feature map of the target point from the feature maps of the plurality of original points; and correcting the coordinates of the first original point to the coordinate position of the second original point.
2. The method of claim 1, wherein determining a region feature map of a plurality of origin points included in a neighborhood corresponding to the target point in the original picture comprises:
determining a neighborhood corresponding to the target point in the original picture based on the reduction ratio;
and determining a characteristic map of a plurality of the origin points included in the neighborhood.
3. The method according to claim 1, wherein before detecting coordinates of a target point included in the received target picture, the method further comprises:
determining the reduction scale based on the resolution of the original picture and the resolution of the picture supported by the target neural network;
and carrying out reduction processing on the original picture according to the reduction ratio so as to obtain the target picture.
4. A coordinate correcting apparatus, comprising:
The detection module is used for detecting coordinates of a target point included in a target picture, wherein the target picture is a picture with a size suitable for a target neural network, which is obtained by carrying out reduction processing on an original picture according to a reduction ratio;
A first determining module, configured to determine a feature map of a target area of the target point based on coordinates of the target point;
A second determining module, configured to determine a region feature map of a plurality of original points included in a neighborhood corresponding to the target point in the original picture;
The correction module is used for carrying out coordinate correction on the original picture based on the characteristic diagram of the target point and the characteristic diagrams of a plurality of original points;
The apparatus further comprises: a calculating module, configured to calculate, based on the reduction scale, a first origin corresponding to the target point in the original picture before performing coordinate correction on the original picture based on the feature map of the target point and the feature maps of the plurality of origins;
The correction module includes: a third determining unit configured to determine, from among the feature maps of the plurality of original points, a feature map of a second original point having a maximum similarity with the feature map of the target point; and a correction unit configured to correct the coordinates of the first origin to the coordinate positions of the second origin.
5. The apparatus of claim 4, wherein the second determining module comprises:
a first determining unit, configured to determine a neighborhood corresponding to the target point in the original picture based on the reduction ratio;
and a second determining unit configured to determine feature maps of the plurality of origin points included in the neighborhood.
6. The apparatus of claim 4, wherein the apparatus further comprises:
A third determining module, configured to determine the reduction scale based on a resolution of the original picture and a resolution of a picture supported by the target neural network;
and the processing module is used for carrying out reduction processing on the original picture according to the reduction ratio so as to obtain the target picture.
7. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein a computer program, wherein the computer program is arranged to perform the method of any of the claims 1 to 3 when run.
8. An electronic device comprising a memory and a processor, characterized in that the memory has stored therein a computer program, the processor being arranged to run the computer program to perform the method of any of the claims 1 to 3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010653896.9A CN112164109B (en) | 2020-07-08 | 2020-07-08 | Coordinate correction method, coordinate correction device, storage medium and electronic device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010653896.9A CN112164109B (en) | 2020-07-08 | 2020-07-08 | Coordinate correction method, coordinate correction device, storage medium and electronic device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112164109A CN112164109A (en) | 2021-01-01 |
| CN112164109B true CN112164109B (en) | 2024-09-06 |
Family
ID=73860485
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010653896.9A Active CN112164109B (en) | 2020-07-08 | 2020-07-08 | Coordinate correction method, coordinate correction device, storage medium and electronic device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112164109B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1595954A (en) * | 2004-03-19 | 2005-03-16 | 华亚微电子（上海）有限公司 | A method of definition compensation during video image zoom |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106780336B (en) * | 2016-12-19 | 2020-04-03 | 广东威创视讯科技股份有限公司 | Image reduction method and device |
| CN108229490B (en) * | 2017-02-23 | 2021-01-05 | 北京市商汤科技开发有限公司 | Key point detection method, neural network training method, device and electronic equipment |
-
2020
- 2020-07-08 CN CN202010653896.9A patent/CN112164109B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1595954A (en) * | 2004-03-19 | 2005-03-16 | 华亚微电子（上海）有限公司 | A method of definition compensation during video image zoom |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112164109A (en) | 2021-01-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109886997B (en) | Identification frame determining method and device based on target detection and terminal equipment | |
| CN110517278B (en) | Image segmentation and training method and device of image segmentation network and computer equipment | |
| WO2020098250A1 (en) | Character recognition method, server, and computer readable storage medium | |
| CN109960742B (en) | Local information searching method and device | |
| CN109784181B (en) | Picture watermark identification method, device, equipment and computer readable storage medium | |
| CN109840477B (en) | Method and device for recognizing shielded face based on feature transformation | |
| CN111340077B (en) | Attention mechanism-based disparity map acquisition method and device | |
| CN110232326B (en) | Three-dimensional object recognition method, device and storage medium | |
| CN109448007B (en) | Image processing method, image processing apparatus, and storage medium | |
| CN111860489A (en) | Certificate image correction method, device, equipment and storage medium | |
| CN113221869B (en) | Medical invoice structured information extraction method, device equipment and storage medium | |
| CN112381092B (en) | Tracking method, device and computer readable storage medium | |
| CN113674220A (en) | Image difference detection method, detection device and storage medium | |
| CN112084874B (en) | Object detection method and device and terminal equipment | |
| CN112418243B (en) | Feature extraction method, device and electronic equipment | |
| CN111627037B (en) | Image region extraction method, device, equipment and storage medium | |
| CN110210314B (en) | Face detection method, device, computer equipment and storage medium | |
| CN111627036B (en) | Image area correction method, device, equipment and storage medium | |
| CN108734712B (en) | Background segmentation method and device and computer storage medium | |
| CN112164109B (en) | Coordinate correction method, coordinate correction device, storage medium and electronic device | |
| CN111768406B (en) | Cell image processing method, device, equipment and storage medium | |
| CN110516660B (en) | Method and device for reading book page content based on image features | |
| CN114693697A (en) | Image processing method, device, equipment and computer readable storage medium | |
| CN112215304A (en) | Gray level image matching method and device for geographic image splicing | |
| CN112669346B (en) | Pavement emergency determination method and device |
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 |