CN115346051A - Optical remote sensing image detection method and device - Google Patents

Optical remote sensing image detection method and device Download PDF

Info

Publication number
CN115346051A
CN115346051A CN202211003182.9A CN202211003182A CN115346051A CN 115346051 A CN115346051 A CN 115346051A CN 202211003182 A CN202211003182 A CN 202211003182A CN 115346051 A CN115346051 A CN 115346051A
Authority
CN
China
Prior art keywords
remote sensing
detection
image
optical remote
information
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.)
Pending
Application number
CN202211003182.9A
Other languages
Chinese (zh)
Inventor
周旗开
牛福
张伟
谢英江
高妍
王进
甄曙辉
杨瑞峰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Institute of Systems Engineering of PLA Academy of Military Sciences
Original Assignee
Institute of Systems Engineering of PLA Academy of Military Sciences
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Institute of Systems Engineering of PLA Academy of Military Sciences filed Critical Institute of Systems Engineering of PLA Academy of Military Sciences
Priority to CN202211003182.9A priority Critical patent/CN115346051A/en
Publication of CN115346051A publication Critical patent/CN115346051A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/07—Target detection

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a method and a device for detecting an optical remote sensing image, wherein the method comprises the following steps: acquiring optical remote sensing image information; the optical remote sensing image information comprises a plurality of optical remote sensing graphs; carrying out positioning identification processing on the optical remote sensing image information by using a preset detection remote sensing target model to obtain an output characteristic diagram information set; the output characteristic diagram information set comprises a plurality of output characteristic diagram information; the detection remote sensing target model comprises a plurality of double attention mechanism modules; carrying out post-processing on the output characteristic diagram information set to obtain a target image detection information set; the target image detection information set includes a number of target image detection information. Therefore, the method is beneficial to improving the characteristic extraction capability of the optical remote sensing image, weakening the attention degree to the complex background, solving the problem that the detection is easy to be mistaken and missed due to the complex image background in the detection of the optical remote sensing image, and improving the detection precision and the detection effect of the optical remote sensing image.

Description

Optical remote sensing image detection method and device
Technical Field
The invention relates to the technical field of image processing, in particular to a method and a device for detecting an optical remote sensing image.
Background
Because the optical remote sensing image is an image shot under an overlooking visual angle, the scene of the optical remote sensing image is usually complex, a remote sensing target is often fused with an image background, and the target characteristic is not obvious, the problem of detection errors and omission easily occurs in the detection process, and the detection precision and efficiency are influenced. Therefore, the method and the device for detecting the optical remote sensing image are provided to improve the feature extraction capability of the optical remote sensing image, weaken the attention to a complex background, solve the problem that the detection is easy to be mistaken and missed in the detection of the optical remote sensing image due to the complex image background, and improve the detection precision and the detection effect of the optical remote sensing image.
Disclosure of Invention
The technical problem to be solved by the present invention is to provide a method and an apparatus for detecting an optical remote sensing image, which can perform positioning, identification and post-processing on the optical remote sensing image by detecting a remote sensing target model to obtain target image detection information, thereby facilitating improvement of feature extraction capability of the optical remote sensing image, weakening attention to a complex background, solving the problem of easy occurrence of false and missed detection due to the complex image background in the optical remote sensing image detection, and improving detection accuracy and detection effect of the optical remote sensing image.
In order to solve the technical problem, a first aspect of the embodiments of the present invention discloses a method for detecting an optical remote sensing image, where the method includes:
acquiring optical remote sensing image information; the optical remote sensing image information comprises a plurality of optical remote sensing graphs;
positioning, identifying and processing the optical remote sensing image information by using a preset detection remote sensing target model to obtain an output characteristic diagram information set; the output characteristic diagram information set comprises a plurality of output characteristic diagram information; the detection remote sensing target model comprises a plurality of double attention mechanism modules; the dual attention mechanism module is used for screening and weighting the feature vectors corresponding to the detection target so as to enhance the feature extraction accuracy of the detection remote sensing target model;
post-processing the output characteristic diagram information set to obtain a target image detection information set; the target image detection information set comprises a plurality of target image detection information.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the model for detecting a remote sensing target includes an input model, a first detection model, a second detection model, and an output model;
the method comprises the following steps of utilizing a preset detection remote sensing target model to carry out positioning, identification and processing on the optical remote sensing image information to obtain an output characteristic diagram information set, and comprises the following steps:
for any optical remote sensing image, converting the optical remote sensing image by using the input model to obtain first intermediate image information corresponding to the optical remote sensing image;
performing feature extraction processing on the first intermediate image information by using the first detection model to obtain second intermediate image information corresponding to the optical remote sensing image; the second intermediate image information includes M first sub-image information; m is a positive integer greater than or equal to 3;
performing feature extraction processing on the second intermediate image information by using the second detection model to obtain third intermediate image information corresponding to the optical remote sensing image; the third intermediate image information includes N second sub-image information; n is a positive integer not greater than M;
and converting the third intermediate image information by using the output model to obtain output characteristic diagram information corresponding to the optical remote sensing image.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the first detection model includes a first backbone network model, a second backbone network model, and a third backbone network model;
the first sub-image information comprises first main image information, and/or second main image information, and/or third main image information;
the step of performing feature extraction processing on the first intermediate image information by using the first detection model to obtain second intermediate image information corresponding to the optical remote sensing image comprises the following steps:
performing feature extraction processing on the first intermediate image information by using the first trunk network model to obtain first trunk image information corresponding to the optical remote sensing image;
performing feature extraction processing on the first trunk image information by using the second trunk network model to obtain second trunk image information corresponding to the optical remote sensing image;
and performing feature extraction processing on the two trunk image information by using the third trunk network model to obtain third trunk image information corresponding to the optical remote sensing image.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the performing post-processing on the output feature map information set to obtain a target image detection information set includes:
carrying out detection frame decoding processing on the output characteristic diagram information set to obtain a detection frame information set; the detection frame information set comprises a plurality of pieces of detection frame information;
carrying out category discrimination processing on the detection frame information set to obtain an image category information set; the set of image category information includes a number of image category information.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the performing a category identification process on the detection frame information set to obtain an image category information set includes:
calculating and screening the detection frame information set by using a preset first confidence threshold value to obtain a target characteristic map information set; the target characteristic diagram information set comprises a plurality of target characteristic diagram information;
and carrying out category decoding processing on the target characteristic image information set to obtain an image category information set.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the calculating and screening the detection box information set by using a preset first confidence threshold to obtain a target feature map information set includes:
performing confidence calculation on the detection frame information set to obtain a first confidence value set; the first set of confidence values comprises a number of first confidence values;
for any one first confidence value, judging whether the first confidence value is not less than a preset first confidence threshold value to obtain a first confidence judgment result;
and when the first confidence judgment result is yes, determining that the output characteristic diagram information corresponding to the first confidence value is the target characteristic diagram information corresponding to the first confidence value.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the detecting a remote sensing target model is obtained by the following training steps:
acquiring an original image information set;
labeling and data enhancement processing are carried out on the original image information set to obtain a first training image information set; the first training image information set comprises a plurality of pieces of first training image information;
determining target training image information according to the first training image information set;
training the first training model by using the target training image information to obtain a second training model;
judging whether the model training parameter information corresponding to the second training model meets training termination conditions or not to obtain a termination judgment result;
when the termination judgment result is negative, updating the first training model by using the second training model, and triggering and executing the third training image information set to determine target training image information;
and when the termination judgment result is yes, determining the second training model as the remote sensing target detection model.
The second aspect of the embodiment of the invention discloses an optical remote sensing image detection device, which comprises:
the acquisition module is used for acquiring optical remote sensing image information; the optical remote sensing image information comprises a plurality of optical remote sensing graphs;
the first processing module is used for carrying out positioning identification processing on the optical remote sensing image information by utilizing a preset detection remote sensing target model to obtain an output characteristic diagram information set; the output characteristic diagram information set comprises a plurality of output characteristic diagram information; the detection remote sensing target model comprises a plurality of double attention mechanism modules; the dual attention mechanism module is used for screening and weighting the feature vectors corresponding to the detection target so as to enhance the feature extraction accuracy of the detection remote sensing target model;
the second processing module is used for carrying out post-processing on the output characteristic diagram information set to obtain a target image detection information set; the target image detection information set comprises a plurality of target image detection information.
As an optional implementation manner, in the second aspect of the embodiment of the present invention, the model for detecting a remote sensing target includes an input model, a first detection model, a second detection model, and an output model;
the first processing module carries out positioning identification processing on the optical remote sensing image information by using a preset detection remote sensing target model, and the specific mode of obtaining an output characteristic diagram information set is as follows:
for any optical remote sensing image, converting the optical remote sensing image by using the input model to obtain first intermediate image information corresponding to the optical remote sensing image;
performing feature extraction processing on the first intermediate image information by using the first detection model to obtain second intermediate image information corresponding to the optical remote sensing image; the second intermediate image information includes M first sub-image information; m is a positive integer greater than or equal to 3;
performing feature extraction processing on the second intermediate image information by using the second detection model to obtain third intermediate image information corresponding to the optical remote sensing image; the third intermediate image information includes N second sub-image information; n is a positive integer not greater than M;
and converting the third intermediate image information by using the output model to obtain output characteristic diagram information corresponding to the optical remote sensing image.
As an optional implementation manner, in the second aspect of the embodiment of the present invention, the first detection model includes a first backbone network model, a second backbone network model, and a third backbone network model;
the first sub-image information comprises first main image information, and/or second main image information, and/or third main image information;
the first processing module performs feature extraction processing on the first intermediate image information by using the first detection model, and the specific way of obtaining second intermediate image information corresponding to the optical remote sensing image is as follows:
performing feature extraction processing on the first intermediate image information by using the first trunk network model to obtain first trunk image information corresponding to the optical remote sensing image;
performing feature extraction processing on the first trunk image information by using the second trunk network model to obtain second trunk image information corresponding to the optical remote sensing image;
and performing feature extraction processing on the two trunk image information by using the third trunk network model to obtain third trunk image information corresponding to the optical remote sensing image.
As an optional implementation manner, in the second aspect of the embodiment of the present invention, the specific manner of obtaining the target image detection information set by performing post-processing on the output feature map information set by the second processing module is as follows:
carrying out detection frame decoding processing on the output characteristic diagram information set to obtain a detection frame information set; the detection frame information set comprises a plurality of pieces of detection frame information;
carrying out category discrimination processing on the detection frame information set to obtain an image category information set; the set of image category information includes a number of image category information.
As an optional implementation manner, in the second aspect of the embodiment of the present invention, the second processing module performs a category identification process on the detection frame information set, and a specific manner of obtaining the image category information set is as follows:
calculating and screening the detection frame information set by using a preset first confidence threshold value to obtain a target characteristic map information set; the target characteristic graph information set comprises a plurality of target characteristic graph information;
and carrying out category decoding processing on the target characteristic diagram information set to obtain an image category information set.
As an optional implementation manner, in a second aspect of the embodiment of the present invention, the calculating and screening, by the second processing module, the detection box information set by using a preset first confidence threshold to obtain a target feature map information set includes:
performing confidence calculation on the detection frame information set to obtain a first confidence value set; the first set of confidence values comprises a number of first confidence values;
for any one first confidence value, judging whether the first confidence value is not less than a preset first confidence threshold value to obtain a first confidence judgment result;
and when the first confidence judgment result is yes, determining that the output characteristic diagram information corresponding to the first confidence value is the target characteristic diagram information corresponding to the first confidence value.
As an optional implementation manner, in the second aspect of the embodiment of the present invention, the detecting remote sensing target model is obtained by a training module performing the following training steps:
acquiring an original image information set;
labeling and data enhancement processing are carried out on the original image information set to obtain a first training image information set; the first training image information set comprises a plurality of pieces of first training image information;
determining target training image information according to the first training image information set;
training the first training model by using the target training image information to obtain a second training model;
judging whether the model training parameter information corresponding to the second training model meets training termination conditions or not to obtain a termination judgment result;
when the termination judgment result is negative, updating the first training model by using the second training model, and triggering and executing the third training image information set to determine target training image information;
and when the termination judgment result is yes, determining the second training model as the remote sensing target detection model.
The third aspect of the present invention discloses another optical remote sensing image detection apparatus, comprising:
a memory storing executable program code;
a processor coupled with the memory;
the processor calls the executable program code stored in the memory to execute part or all of the steps in the optical remote sensing image detection method disclosed by the first aspect of the embodiment of the invention.
In a fourth aspect of the present invention, a computer storage medium is disclosed, where the computer storage medium stores computer instructions, and when the computer instructions are called, the computer instructions are used to perform some or all of the steps in the optical remote sensing image detection method disclosed in the first aspect of the embodiments of the present invention.
Compared with the prior art, the embodiment of the invention has the following beneficial effects:
in the embodiment of the invention, optical remote sensing image information is obtained; the optical remote sensing image information comprises a plurality of optical remote sensing graphs; positioning, identifying and processing the optical remote sensing image information by using a preset detection remote sensing target model to obtain an output characteristic diagram information set; the output characteristic diagram information set comprises a plurality of output characteristic diagram information; the detection remote sensing target model comprises a plurality of double attention mechanism modules; the dual attention mechanism module is used for enhancing the feature extraction accuracy rate of the remote sensing target model; carrying out post-processing on the output characteristic diagram information set to obtain a target image detection information set; the target image detection information set includes a number of target image detection information. Therefore, the method can perform positioning, identifying and post-processing on the optical remote sensing image by detecting the remote sensing target model to obtain target image detection information, is favorable for improving the characteristic extraction capability of the optical remote sensing image, weakens the attention degree to a complex background, solves the problem that the detection is easy to be mistaken and missed due to the complex image background in the optical remote sensing image detection, and improves the detection precision and the detection effect of the optical remote sensing image.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a schematic flow chart of a method for detecting an optical remote sensing image according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart of another optical remote sensing image detection method disclosed in the embodiment of the present invention;
FIG. 3 is a schematic structural diagram of an optical remote sensing image detection apparatus according to an embodiment of the present disclosure;
FIG. 4 is a schematic structural diagram of another optical remote sensing image detection apparatus disclosed in the embodiments of the present invention;
fig. 5 is a schematic structural diagram of another optical remote sensing image detection device disclosed in the embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The terms "first," "second," and the like in the description and claims of the present invention and in the above-described drawings are used for distinguishing between different objects and not for describing a particular order. Furthermore, the terms "include" and "have," as well as any variations thereof, are intended to cover a non-exclusive inclusion. For example, a process, method, apparatus, product, or apparatus that comprises a list of steps or elements is not limited to only those steps or elements recited, but may alternatively include other steps or elements not expressly listed or inherent to such process, method, product, or apparatus.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the invention. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.
The invention discloses a method and a device for detecting an optical remote sensing image, which can perform positioning, identifying and post-processing on the optical remote sensing image through a detection remote sensing target model to obtain target image detection information, are favorable for improving the characteristic extraction capability of the optical remote sensing image, weaken the attention to a complex background, solve the problem that the detection is easy to miss and miss due to the complex image background in the detection of the optical remote sensing image, and improve the detection precision and the detection effect of the optical remote sensing image. The following are detailed descriptions.
Example one
Referring to fig. 1, fig. 1 is a schematic flow chart of a method for detecting an optical remote sensing image according to an embodiment of the present invention. The optical remote sensing image detection method described in fig. 1 is applied to an image processing system, such as a local server or a cloud server for optical remote sensing image detection management, and the embodiment of the present invention is not limited thereto. As shown in fig. 1, the optical remote sensing image detection method may include the following operations:
101. and obtaining optical remote sensing image information.
In the embodiment of the invention, the optical remote sensing image information comprises a plurality of optical remote sensing graphs.
102. And carrying out positioning identification processing on the optical remote sensing image information by utilizing a preset detection remote sensing target model to obtain an output characteristic diagram information set.
In an embodiment of the present invention, the output feature map information set includes a plurality of output feature map information.
In the embodiment of the invention, the remote sensing target detection model comprises a plurality of double attention mechanism modules.
In the embodiment of the invention, the dual attention mechanism module is used for screening and weighting the feature vectors corresponding to the detection target so as to enhance the feature extraction accuracy of the remote sensing target model.
103. And carrying out post-processing on the output characteristic diagram information set to obtain a target image detection information set.
In an embodiment of the present invention, the target image detection information set includes a plurality of target image detection information.
Optionally, the dual attention mechanism module includes a channel attention mechanism module and/or a space attention mechanism module, which is not limited in the embodiment of the present invention.
Optionally, the dual attention mechanism module enhances the feature extraction accuracy of the remote sensing target model by enhancing useful features in the image and suppressing useless features.
Optionally, the channel attention mechanism module is used for enhancing the characteristic content of the detection target.
Optionally, the spatial attention mechanism module is used to enhance position detection of the detection target.
Optionally, when the dual attention mechanism module processes data information, the data information is processed through the channel attention mechanism module, and then processed through the spatial attention mechanism module, so as to highlight important characteristic information in the optical remote sensing image, weaken background characteristic information, and further optimize target detection accuracy.
Optionally, the initial learning rate of the remote sensing target model is set to 0.01.
Optionally, the learning rate attenuation weight of the remote sensing target model is 0.0005.
Optionally, when the optional detection remote sensing target model is used for performing positioning identification processing on the optical remote sensing image, the learning rate is intelligently adjusted by using the cosine annealing learning rate attenuation model, so that the detection precision and the detection effect of the optical remote sensing image are further improved.
Therefore, the optical remote sensing image detection method described by the embodiment of the invention can carry out positioning, identification and post-processing on the optical remote sensing image through the detection remote sensing target model to obtain target image detection information, is beneficial to improving the characteristic extraction capability of the optical remote sensing image, weakens the attention degree to a complex background, solves the problem that detection errors and detection omissions are easy to occur due to the complex image background in the optical remote sensing image detection, and improves the detection precision and the detection effect of the optical remote sensing image.
In an optional embodiment, the model for detecting the remote sensing target comprises an input model, a first detection model, a second detection model and an output model;
the method comprises the following steps of utilizing a preset detection remote sensing target model to carry out positioning, identification and processing on optical remote sensing image information to obtain an output characteristic diagram information set, and comprises the following steps:
for any optical remote sensing image, converting the optical remote sensing image by using an input model to obtain first intermediate image information corresponding to the optical remote sensing image;
performing feature extraction processing on the first intermediate image information by using a first detection model to obtain second intermediate image information corresponding to the optical remote sensing image; the second intermediate image information includes M first sub-image information; m is a positive integer greater than or equal to 3;
performing feature extraction processing on the second intermediate image information by using a second detection model to obtain third intermediate image information corresponding to the optical remote sensing image; the third intermediate image information includes N second sub-image information; n is a positive integer not greater than M;
and converting the third intermediate image information by using the output model to obtain output characteristic diagram information corresponding to the optical remote sensing image.
Optionally, the output feature map information includes an output feature map
Optionally, the output characteristic diagram includes a first output characteristic diagram, and/or a second output characteristic diagram, and/or a third output characteristic diagram, which is not limited in the embodiment of the present invention.
Optionally, the dimension corresponding to the output feature map is 15.
Optionally, the image dimension feature corresponding to the output feature map is K × P.
Optionally, K is an integer multiple of 10.
Optionally, P is a positive integer greater than or equal to 1.
Optionally, K × K represents the number of feature points of the output feature map.
Optionally, the P represents the number of the detection frames corresponding to each feature point.
Optionally, the number of feature points corresponding to the third output feature map, the number of feature points corresponding to the second output feature map, and the number of feature points corresponding to the first output feature map are sequentially increased by a geometric multiple.
Optionally, the second detection model includes 4 dual attention mechanism modules.
Optionally, the first output feature map is obtained by processing the second intermediate image information by 2 dual attention mechanism modules.
Optionally, the second output feature map is obtained by processing the second intermediate image information by 3 dual attention mechanism modules.
Optionally, the third output feature map is obtained by processing the second intermediate image information by 4 dual attention mechanism modules.
Optionally, the optical remote sensing image detection method is adopted to process the shot optical remote sensing image, so that the detection target can be accurately identified and positioned, and accurate data reference is provided for applications such as maritime accident personnel search and rescue, military information reconnaissance, traffic monitoring and the like.
Therefore, the optical remote sensing image detection method described in the embodiment of the invention can obtain the output characteristic map information by positioning, identifying and processing the optical remote sensing image by using a plurality of models, is beneficial to improving the characteristic extraction capability of the optical remote sensing image, weakening the attention degree to a complex background, solving the problem that the detection is easy to be mistaken and missed due to the complex image background in the optical remote sensing image detection, and improving the detection precision and the detection effect of the optical remote sensing image.
In another optional embodiment, the first detection model includes a first backbone network model, a second backbone network model, and a third backbone network model;
the first sub-image information comprises first main image information, and/or second main image information, and/or third main image information;
the method for extracting the features of the first intermediate image information by using the first detection model to obtain the second intermediate image information corresponding to the optical remote sensing image comprises the following steps:
performing feature extraction processing on the first intermediate image information by using a first trunk network model to obtain first trunk image information corresponding to the optical remote sensing image;
performing feature extraction processing on the first trunk image information by using a second trunk network model to obtain second trunk image information corresponding to the optical remote sensing image;
and performing feature extraction processing on the two trunk image information by using a third trunk network model to obtain third trunk image information corresponding to the optical remote sensing image.
Optionally, the first backbone network model is sequentially composed of 1 Focus structure module, 1 dual attention mechanism module, 1C 3 network structure module, 1 dual attention mechanism module, and 1C 3 network structure module.
Optionally, the second backbone network model is sequentially composed of 1 dual attention mechanism module and 1C 3 network structure module.
Optionally, the third backbone network model is sequentially composed of 1 dual attention mechanism module, 1 spatial pyramid pooling layer, and 1C 3 network structure module.
Optionally, the third main image information is input to the dual attention mechanism module for processing.
Therefore, the optical remote sensing image detection method described in the embodiment of the invention can utilize a plurality of models to perform feature extraction processing on the first intermediate image information to obtain the second intermediate image information, is beneficial to improving the feature extraction capability of the optical remote sensing image, weakens the attention degree to a complex background, solves the problem that the detection error and the detection omission are easy to occur due to the complex image background in the optical remote sensing image detection, and improves the detection precision and the detection effect of the optical remote sensing image.
Example two
Referring to fig. 2, fig. 2 is a schematic flow chart of another optical remote sensing image detection method according to an embodiment of the present invention. The optical remote sensing image detection method described in fig. 2 is applied to an image processing system, such as a local server or a cloud server for optical remote sensing image detection management, and the embodiment of the present invention is not limited thereto. As shown in fig. 2, the optical remote sensing image detection method may include the following operations:
201. and acquiring optical remote sensing image information.
202. And carrying out positioning identification processing on the optical remote sensing image information by utilizing a preset detection remote sensing target model to obtain an output characteristic diagram information set.
203. And carrying out detection frame decoding processing on the output characteristic diagram information set to obtain a detection frame information set.
In an embodiment of the present invention, the detection box information set includes a plurality of pieces of detection box information.
204. And carrying out category discrimination processing on the detection frame information set to obtain an image category information set.
In an embodiment of the present invention, the image category information set includes a plurality of image category information.
In the embodiment of the present invention, for specific technical details and technical noun explanations of step 201 to step 202, reference may be made to the detailed description of step 101 to step 102 in the first embodiment, and details are not repeated in the embodiment of the present invention.
In this optional embodiment, as an optional implementation, the specific manner of performing detection frame decoding processing on the output feature map information set to obtain the detection frame information set is as follows:
for any output characteristic diagram information, acquiring prior frame information;
and calculating the output characteristic diagram information and the prior frame information by using a detection frame decoding model to obtain detection frame information corresponding to the output characteristic diagram information.
Optionally, the detection frame information includes a first center point coordinate, and/or a second center point coordinate, and/or a detection frame width, and/or a detection frame height, which is not limited in the embodiment of the present invention.
Optionally, the specific form of the detection frame decoding model is as follows:
b x ＝σ(t x )+c x ；
b y ＝σ(t y )+c y ；
Figure BDA0003807584690000121
Figure BDA0003807584690000122
wherein, b x Is a first center point coordinate; b is a mixture of y Is a second center point coordinate; b w The width of the detection frame; b h Is the height of the detection frame; sigma (t) x ) And σ (t) y ) Respectively a first offset and a second offset based on the coordinate of a grid point at the upper left corner of the center point of the rectangular frame, wherein sigma is an activation function; p is a radical of w And p h The width and height of the prior frame respectively; t is t x 、t y 、t w And t h Respectively a first predicted offset, a second predicted offset, a third predicted offset and a fourth predicted offset; c. C x And c y The first corner grid coordinate and the second corner grid coordinate are respectively based on the upper left corner grid of the center point of the rectangular frame.
Therefore, the optical remote sensing image detection method described by the embodiment of the invention can carry out positioning identification processing, detection frame decoding processing and category discrimination processing on the optical remote sensing image through the detection remote sensing target model to obtain target image detection information, is favorable for improving the feature extraction capability of the optical remote sensing image, weakens the attention to a complex background, solves the problem that detection errors and detection omissions are easy to occur due to the complex image background in the optical remote sensing image detection, and improves the detection precision and the detection effect of the optical remote sensing image.
In an optional embodiment, the performing the category identification processing on the detection frame information set to obtain an image category information set includes:
calculating and screening the detection frame information set by using a preset first confidence threshold value to obtain a target characteristic diagram information set; the target characteristic diagram information set comprises a plurality of target characteristic diagram information;
and carrying out category decoding processing on the target characteristic image information set to obtain an image category information set.
In this optional embodiment, as an optional implementation, the performing of the category decoding processing on the target feature map information set to obtain the image category information set specifically includes:
for any target feature map information, determining first prediction frame information corresponding to the target feature map information;
calculating the confidence of the first prediction frame information to obtain second confidence value information corresponding to the target characteristic diagram information;
screening the second confidence value information to obtain a target confidence value;
performing IOU calculation processing on the second confidence value information by using a preset IOU model and the target confidence value to obtain IOU value information;
screening the IOU value information by using a preset second confidence threshold value to obtain target IOU value information; the target IOU value information includes X target IOU values;
judging whether the X is less than or equal to an IOU threshold value to obtain a threshold value judgment result;
when the threshold judgment result is negative, determining second prediction frame information according to the target IOU value information;
performing intersection updating processing on the first prediction frame information by using the second prediction frame information, and triggering and executing the calculation of the confidence coefficient of the first prediction frame information to obtain second confidence value information corresponding to the target feature map information;
and if the threshold value judgment result is yes, determining the type information corresponding to the target IOU value information as the image type information corresponding to the target feature map information.
Optionally, X is a positive integer greater than or equal to 1.
Therefore, the optical remote sensing image detection method described in the embodiment of the invention can obtain the image category information through calculation and screening processing of the detection frame information and category decoding processing, is more favorable for improving the feature extraction capability of the optical remote sensing image, weakening the attention degree to the complex background, solving the problem that the detection is easy to be mistaken and missed due to the complex image background in the optical remote sensing image detection, and improving the detection precision and the detection effect of the optical remote sensing image.
In another optional embodiment, the above calculating and screening the detection box information set by using a preset first confidence threshold to obtain a target feature map information set includes:
performing confidence calculation on the detection frame information set to obtain a first confidence value set; the first set of confidence values comprises a number of first confidence values;
for any first confidence value, judging whether the first confidence value is not less than a preset first confidence threshold value to obtain a first confidence judgment result;
and when the first confidence judgment result is yes, determining that the output characteristic diagram information corresponding to the first confidence value is the target characteristic diagram information corresponding to the first confidence value.
Optionally, the value range of the first confidence value is [0,1].
Therefore, the optical remote sensing image detection method described in the embodiment of the invention can calculate and judge the confidence of the detection frame information to obtain the target feature map information, is more favorable for improving the feature extraction capability of the optical remote sensing image, weakening the attention to a complex background, solving the problem that the detection is easy to be mistaken and missed due to the complex image background in the optical remote sensing image detection, and improving the detection precision and the detection effect of the optical remote sensing image.
In yet another alternative embodiment, the above-mentioned model for detecting a remote sensing target is obtained by the following training steps:
acquiring an original image information set;
performing labeling and data enhancement processing on an original image information set to obtain a first training image information set; the first training image information set comprises a plurality of pieces of first training image information;
determining target training image information according to the first training image information set;
training the first training model by using the target training image information to obtain a second training model;
judging whether the model training parameter information corresponding to the second training model meets training termination conditions or not to obtain a termination judgment result;
when the termination judgment result is negative, updating the first training model by using the second training model, and triggering execution to determine target training image information according to a third training image information set;
and when the termination judgment result is yes, determining the second training model as a remote sensing target detection model.
In this optional embodiment, the specific way of performing labeling and data enhancement processing on the original image information set to obtain the first training image information set is as follows:
performing labeling processing on the original image information set to obtain a second training image information set; the second training image information set comprises a plurality of second training image information;
performing data enhancement processing on the second training image information set to obtain a third training image information set; the third training image information set comprises a plurality of pieces of third training image information;
and classifying the third training image information set according to a preset training image proportion parameter to obtain a first training image information set.
Optionally, the data enhancement processing includes random scaling, and/or flipping, and/or brightness enhancement, and/or contrast enhancement, and/or color enhancement, which is not limited in the embodiment of the present invention.
Optionally, the first training image information set obtained by performing data enhancement processing on the second training image information set is beneficial to enhancing the generalization processing capability of the remote sensing target model, and further improves the detection precision of the optical remote sensing image.
In this optional embodiment, as an optional implementation manner, the above-mentioned determining whether the model training parameter information corresponding to the second training model meets the training termination condition includes:
calculating a loss value of model training parameter information corresponding to the second training model to obtain loss function information; the loss function information comprises a first loss function value and/or a second loss function value and/or the number of model iterations;
judging whether the iteration times of the model are equal to an iteration threshold value or not to obtain a first loss judgment result;
when the first loss judgment result is yes, determining that the termination judgment result is yes;
when the first loss judgment result is negative, judging whether the first loss function value is less than or equal to a first loss threshold value to obtain a second loss judgment result;
when the second loss judgment result is yes, determining that the termination judgment result is yes;
when the second loss judgment result is negative, judging whether the second loss function value is less than or equal to a second loss threshold value to obtain a third loss judgment result;
when the third loss judgment result is yes, determining that the termination judgment result is yes;
and when the third loss judgment result is negative, determining that the termination judgment result is negative.
Optionally, the model training parameter information includes loss function parameter information, and/or evaluation function parameter information, and/or training iteration times, which is not limited in the embodiment of the present invention.
In this optional embodiment, as an optional implementation manner, the above calculating the loss value of the model training parameter information corresponding to the second training model to obtain the loss function information specifically includes:
determining the training iteration times in the model training parameter information as model iteration times;
calculating loss function parameter information in the model training parameter information by using a preset first loss function model to obtain a first loss function value;
and calculating evaluation function parameter information in the model training parameter information by using a preset second loss function model to obtain a second loss function value.
Optionally, the specific form of the first loss function model is as follows:
Figure BDA0003807584690000161
Figure BDA0003807584690000162
Figure BDA0003807584690000163
Figure BDA0003807584690000164
wherein: l is total Is a first loss function value; l is 1 To locate the loss function value; l is 2 A confidence loss function value; l is 3 Is a category loss function value; lambda [ alpha ] 1 Is the localization loss function coefficient; lambda [ alpha ] 2 Is the confidence loss function coefficient; lambda 3 Is a class loss function coefficient; y is the number of samples; i is a first sample number; j is the second sample number; b and B i Respectively representing the information of the prediction frame and the information of the detection frame; b and b gt Respectively the central point of the prediction frame and the central point of the detection frame; rho is the Euclidean distance between the central point of the prediction frame and the central point of the detection frame; c is the diagonal distance of the minimum circumscribed rectangle of the prediction frame and the detection frame; α is first frame information; v is second frame information; o is a confidence coefficient with the value range of [0, 1%]；
Figure BDA0003807584690000165
Is the prediction confidence by Sigmoid function.
Optionally, the specific form of the second loss function model is:
Figure BDA0003807584690000166
Figure BDA0003807584690000171
wherein mAP is a second loss function value; m represents the number of categories of all samples; AP is an area enclosed by horizontal and vertical coordinates with recall rate and accuracy; r is recall, and P (R) is precision and recall curves.
Optionally, the second loss function value represents the accuracy of the detection result of the category, and the detection accuracy of detecting the remote sensing target model can be improved through the judgment and analysis of the second loss function value and the second loss threshold.
Therefore, the method for detecting the optical remote sensing image, which is described by the embodiment of the invention, can obtain the detection remote sensing target model by training the first training model, is more favorable for improving the feature extraction capability of the optical remote sensing image, weakening the attention degree to a complex background, solving the problem that the detection is easy to be mistaken and missed due to the complex image background in the detection of the optical remote sensing image, and improving the detection precision and the detection effect of the optical remote sensing image.
EXAMPLE III
Referring to fig. 3, fig. 3 is a schematic structural diagram of an optical remote sensing image detection apparatus according to an embodiment of the present invention. The apparatus described in fig. 3 can be applied to an image processing system, such as a local server or a cloud server for optical remote sensing image detection management, and the embodiment of the present invention is not limited thereto. As shown in fig. 3, the apparatus may include:
an obtaining module 301, configured to obtain optical remote sensing image information; the optical remote sensing image information comprises a plurality of optical remote sensing graphs;
the first processing module 302 is configured to perform positioning identification processing on optical remote sensing image information by using a preset detection remote sensing target model to obtain an output feature map information set; the output characteristic diagram information set comprises a plurality of output characteristic diagram information; the detection remote sensing target model comprises a plurality of double attention mechanism modules; the dual attention mechanism module is used for screening and weighting the feature vectors corresponding to the detection target so as to enhance the feature extraction accuracy of the remote sensing target model;
the second processing module 303 is configured to perform post-processing on the output feature map information set to obtain a target image detection information set; the target image detection information set includes a number of target image detection information.
Therefore, by implementing the optical remote sensing image detection device described in fig. 3, the target image detection information can be obtained by performing positioning, identification and post-processing on the optical remote sensing image through detecting the remote sensing target model, which is beneficial to improving the feature extraction capability of the optical remote sensing image, weakening the attention degree to a complex background, solving the problem that detection errors and detection omission easily occur due to the complex image background in the optical remote sensing image detection, and improving the detection precision and detection effect of the optical remote sensing image.
In another alternative embodiment, as shown in fig. 4, the model for detecting the remote sensing target comprises an input model, a first detection model, a second detection model and an output model;
the first processing module 302 performs positioning, identifying and processing on the optical remote sensing image information by using a preset remote sensing target detection model, and obtains an output feature map information set in a specific manner as follows:
for any optical remote sensing image, converting the optical remote sensing image by using an input model to obtain first intermediate image information corresponding to the optical remote sensing image;
performing feature extraction processing on the first intermediate image information by using a first detection model to obtain second intermediate image information corresponding to the optical remote sensing image; the second intermediate image information includes M pieces of first sub-image information; m is a positive integer greater than or equal to 3;
performing feature extraction processing on the second intermediate image information by using a second detection model to obtain third intermediate image information corresponding to the optical remote sensing image; the third intermediate image information includes N second sub-image information; n is a positive integer not greater than M;
and converting the third intermediate image information by using the output model to obtain output characteristic diagram information corresponding to the optical remote sensing image.
Therefore, by implementing the optical remote sensing image detection device described in fig. 4, the output feature map information can be obtained by performing positioning identification processing on the optical remote sensing image by using a plurality of models, which is beneficial to improving the feature extraction capability of the optical remote sensing image, weakening the attention to a complex background, solving the problem that detection errors and detection omissions are easy to occur due to the complex image background in the optical remote sensing image detection, and improving the detection precision and detection effect of the optical remote sensing image.
In yet another alternative embodiment, as shown in fig. 4, the first detection model includes a first backbone network model, a second backbone network model, and a third backbone network model;
the first sub-image information comprises first trunk image information, and/or second trunk image information, and/or third trunk image information;
the first processing module 302 performs feature extraction processing on the first intermediate image information by using the first detection model, and a specific manner of obtaining second intermediate image information corresponding to the optical remote sensing image is as follows:
performing feature extraction processing on the first intermediate image information by using a first trunk network model to obtain first trunk image information corresponding to the optical remote sensing image;
performing feature extraction processing on the first trunk image information by using a second trunk network model to obtain second trunk image information corresponding to the optical remote sensing image;
and performing feature extraction processing on the two trunk image information by using a third trunk network model to obtain third trunk image information corresponding to the optical remote sensing image.
Therefore, by implementing the optical remote sensing image detection device described in fig. 4, the multiple models can be used for performing feature extraction processing on the first intermediate image information to obtain the second intermediate image information, which is beneficial to improving the feature extraction capability of the optical remote sensing image, weakening the attention degree to the complex background, solving the problem that the detection error and the detection omission are easy to occur due to the complex image background in the optical remote sensing image detection, and improving the detection precision and the detection effect of the optical remote sensing image.
In yet another alternative embodiment, as shown in fig. 4, the specific way of obtaining the target image detection information set by post-processing the output feature map information set by the second processing module 303 is as follows:
carrying out detection frame decoding processing on the output characteristic diagram information set to obtain a detection frame information set; the detection frame information set comprises a plurality of pieces of detection frame information;
carrying out category discrimination processing on the detection frame information set to obtain an image category information set; the set of image category information includes a number of image category information.
Therefore, by implementing the optical remote sensing image detection device described in fig. 4, the target image detection information can be obtained by performing positioning identification processing, detection frame decoding processing and category discrimination processing on the optical remote sensing image through the detection remote sensing target model, which is beneficial to improving the feature extraction capability of the optical remote sensing image, weakening the attention to a complex background, solving the problem that detection errors and omission easily occur due to the complex image background in the optical remote sensing image detection, and improving the detection precision and detection effect of the optical remote sensing image.
In yet another alternative embodiment, as shown in fig. 4, the second processing module 303 performs category discrimination processing on the detection frame information set to obtain an image category information set in a specific manner:
calculating and screening the detection frame information set by using a preset first confidence threshold value to obtain a target characteristic map information set; the target characteristic diagram information set comprises a plurality of target characteristic diagram information;
and carrying out category decoding processing on the target characteristic diagram information set to obtain an image category information set.
Therefore, by implementing the optical remote sensing image detection device described in fig. 4, the image category information can be obtained through calculation and screening processing of the detection frame information and category decoding processing, which is more beneficial to improving the feature extraction capability of the optical remote sensing image, weakening the attention degree to the complex background, solving the problem that the detection error and omission easily occur due to the complex image background in the optical remote sensing image detection, and improving the detection precision and detection effect of the optical remote sensing image.
In yet another alternative embodiment, as shown in fig. 4, the second processing module 303 performs calculation and filtering processing on the detection box information set by using a preset first confidence threshold to obtain a target feature map information set, including:
performing confidence calculation on the detection frame information set to obtain a first confidence value set; the first set of confidence values comprises a number of first confidence values;
for any first confidence value, judging whether the first confidence value is not less than a preset first confidence threshold value to obtain a first confidence judgment result;
and when the first confidence judgment result is yes, determining that the output characteristic diagram information corresponding to the first confidence value is the target characteristic diagram information corresponding to the first confidence value.
Therefore, by implementing the optical remote sensing image detection device described in fig. 4, the confidence degree of the detection frame information can be calculated and judged to obtain the target feature map information, which is more beneficial to improving the feature extraction capability of the optical remote sensing image, weakening the attention degree to the complex background, solving the problem that the detection is easy to be missed and missed due to the complex image background in the optical remote sensing image detection, and improving the detection precision and the detection effect of the optical remote sensing image.
In yet another alternative embodiment, as shown in FIG. 4, detecting the remote sensing target model is performed by training module 304 through the following training steps:
acquiring an original image information set;
performing labeling and data enhancement processing on an original image information set to obtain a first training image information set; the first training image information set comprises a plurality of pieces of first training image information;
determining target training image information according to the first training image information set;
training the first training model by using the target training image information to obtain a second training model;
judging whether the model training parameter information corresponding to the second training model meets training termination conditions or not to obtain a termination judgment result;
when the termination judgment result is negative, updating the first training model by using the second training model, and triggering execution to determine target training image information according to a third training image information set;
and when the termination judgment result is yes, determining the second training model as a remote sensing target detection model.
Therefore, by implementing the optical remote sensing image detection device described in fig. 4, the detection remote sensing target model can be obtained through training the first training model, which is more beneficial to improving the feature extraction capability of the optical remote sensing image, weakening the attention degree to the complex background, solving the problem that the detection is easy to be missed and missed due to the complex image background in the optical remote sensing image detection, and improving the detection precision and the detection effect of the optical remote sensing image.
Example four
Referring to fig. 5, fig. 5 is a schematic structural diagram of another optical remote sensing image detection apparatus according to an embodiment of the present disclosure. The apparatus described in fig. 5 can be applied to an image processing system, such as a local server or a cloud server for optical remote sensing image detection management, and the embodiment of the present invention is not limited thereto. As shown in fig. 5, the apparatus may include:
a memory 401 storing executable program code;
a processor 402 coupled with the memory 401;
the processor 402 calls the executable program code stored in the memory 401 for executing the steps of the method for detecting an optical remote sensing image described in the first embodiment or the second embodiment.
EXAMPLE five
The embodiment of the invention discloses a computer-readable storage medium which stores a computer program for electronic data exchange, wherein the computer program enables a computer to execute the steps in the optical remote sensing image detection method described in the first embodiment or the second embodiment.
EXAMPLE six
The embodiment of the invention discloses a computer program product, which comprises a non-transitory computer readable storage medium storing a computer program, wherein the computer program is operable to make a computer execute the steps in the optical remote sensing image detection method described in the first embodiment or the second embodiment.
The above-described embodiments of the apparatus are only illustrative, and the modules described as separate parts may or may not be physically separate, and the parts displayed as modules may or may not be physical modules, may be located in one place, or may be distributed on a plurality of network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above detailed description of the embodiments, those skilled in the art will clearly understand that each embodiment may be implemented by software plus a necessary general hardware platform, and may also be implemented by hardware. With this understanding in mind, the above technical solutions may essentially or in part contribute to the prior art, be embodied in the form of a software product, which may be stored in a computer-readable storage medium, including a Read-Only Memory (ROM), a Random Access Memory (RAM), a Programmable Read-Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), a One-time Programmable Read-Only Memory (OTPROM), an electronically Erasable Programmable Read-Only Memory (EEPROM), an optical Disc-Read (CD-ROM) or other storage medium capable of storing data, a magnetic tape, or any other computer-readable medium capable of storing data.
Finally, it should be noted that: the method and the device for detecting an optical remote sensing image disclosed in the embodiments of the present invention are only preferred embodiments of the present invention, and are only used for illustrating the technical solutions of the present invention, not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those skilled in the art; the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims (10)

1. An optical remote sensing image detection method is characterized by comprising the following steps:
acquiring optical remote sensing image information; the optical remote sensing image information comprises a plurality of optical remote sensing graphs;
positioning, identifying and processing the optical remote sensing image information by using a preset detection remote sensing target model to obtain an output characteristic diagram information set; the output characteristic diagram information set comprises a plurality of output characteristic diagram information; the detection remote sensing target model comprises a plurality of double attention mechanism modules; the dual attention mechanism module is used for screening and weighting the feature vectors corresponding to the detection target so as to enhance the feature extraction accuracy of the detection remote sensing target model;
post-processing the output characteristic diagram information set to obtain a target image detection information set; the target image detection information set comprises a plurality of target image detection information.
2. The optical remote sensing image detection method according to claim 1, wherein the detection remote sensing target model includes an input model, a first detection model, a second detection model, and an output model;
the method comprises the following steps of utilizing a preset detection remote sensing target model to carry out positioning recognition processing on the optical remote sensing image information to obtain an output characteristic diagram information set, and comprises the following steps:
for any optical remote sensing image, converting the optical remote sensing image by using the input model to obtain first intermediate image information corresponding to the optical remote sensing image;
performing feature extraction processing on the first intermediate image information by using the first detection model to obtain second intermediate image information corresponding to the optical remote sensing image; the second intermediate image information includes M first sub-image information; m is a positive integer greater than or equal to 3;
performing feature extraction processing on the second intermediate image information by using the second detection model to obtain third intermediate image information corresponding to the optical remote sensing image; the third intermediate image information includes N second sub-image information; n is a positive integer not greater than M;
and converting the third intermediate image information by using the output model to obtain output characteristic diagram information corresponding to the optical remote sensing image.
3. The optical remote sensing image detection method according to claim 2, wherein the first detection model includes a first backbone network model, a second backbone network model, and a third backbone network model;
the first sub-image information comprises first main image information, and/or second main image information, and/or third main image information;
the step of performing feature extraction processing on the first intermediate image information by using the first detection model to obtain second intermediate image information corresponding to the optical remote sensing image comprises the following steps:
performing feature extraction processing on the first intermediate image information by using the first trunk network model to obtain first trunk image information corresponding to the optical remote sensing image;
performing feature extraction processing on the first trunk image information by using the second trunk network model to obtain second trunk image information corresponding to the optical remote sensing image;
and performing feature extraction processing on the two trunk image information by using the third trunk network model to obtain third trunk image information corresponding to the optical remote sensing image.
4. The optical remote sensing image detection method according to claim 1, wherein the post-processing the output feature map information set to obtain a target image detection information set comprises:
carrying out detection frame decoding processing on the output characteristic diagram information set to obtain a detection frame information set; the detection frame information set comprises a plurality of pieces of detection frame information;
carrying out category discrimination processing on the detection frame information set to obtain an image category information set; the set of image category information includes a number of image category information.
5. The optical remote sensing image detection method according to claim 4, wherein the performing a category discrimination process on the detection frame information set to obtain an image category information set includes:
calculating and screening the detection frame information set by using a preset first confidence threshold value to obtain a target characteristic diagram information set; the target characteristic graph information set comprises a plurality of target characteristic graph information;
and carrying out category decoding processing on the target characteristic image information set to obtain an image category information set.
6. The optical remote sensing image detection method according to claim 5, wherein the calculating and screening the detection frame information set by using a preset first confidence threshold to obtain a target feature map information set comprises:
performing confidence calculation on the detection frame information set to obtain a first confidence value set; the first set of confidence values comprises a number of first confidence values;
for any one first confidence value, judging whether the first confidence value is not less than a preset first confidence threshold value to obtain a first confidence judgment result;
and when the first confidence judgment result is yes, determining that the output characteristic diagram information corresponding to the first confidence value is the target characteristic diagram information corresponding to the first confidence value.
7. The method for detecting optical remote sensing images according to claim 1, wherein the model for detecting remote sensing targets is obtained through the following training steps:
acquiring an original image information set;
labeling and data enhancement processing are carried out on the original image information set to obtain a first training image information set; the first training image information set comprises a plurality of pieces of first training image information;
determining target training image information according to the first training image information set;
training the first training model by using the target training image information to obtain a second training model;
judging whether the model training parameter information corresponding to the second training model meets training termination conditions or not to obtain termination judgment results;
when the termination judgment result is negative, updating the first training model by using the second training model, and triggering and executing the third training image information set to determine target training image information;
and when the termination judgment result is yes, determining the second training model as the remote sensing target detection model.
8. An optical remote sensing image detection apparatus, characterized in that the apparatus comprises:
the acquisition module is used for acquiring optical remote sensing image information; the optical remote sensing image information comprises a plurality of optical remote sensing graphs;
the first processing module is used for carrying out positioning identification processing on the optical remote sensing image information by utilizing a preset detection remote sensing target model to obtain an output characteristic diagram information set; the output characteristic diagram information set comprises a plurality of output characteristic diagram information; the detection remote sensing target model comprises a plurality of double attention mechanism modules; the dual attention mechanism module is used for screening and weighting the feature vectors corresponding to the detection target so as to enhance the feature extraction accuracy of the detection remote sensing target model;
the second processing module is used for carrying out post-processing on the output characteristic diagram information set to obtain a target image detection information set; the target image detection information set comprises a plurality of target image detection information.
9. An optical remote sensing image detection apparatus, characterized in that the apparatus comprises:
a memory storing executable program code;
a processor coupled with the memory;
the processor calls the executable program code stored in the memory to execute the optical remote sensing image detection method according to any one of claims 1 to 7.
10. A computer storage medium characterized in that it stores computer instructions which, when invoked, are adapted to perform the method of optical remote sensing image detection according to any of claims 1-7.
CN202211003182.9A 2022-08-19 2022-08-19 Optical remote sensing image detection method and device Pending CN115346051A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211003182.9A CN115346051A (en) 2022-08-19 2022-08-19 Optical remote sensing image detection method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211003182.9A CN115346051A (en) 2022-08-19 2022-08-19 Optical remote sensing image detection method and device

Publications (1)

Publication Number Publication Date
CN115346051A true CN115346051A (en) 2022-11-15

Family

ID=83953310

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211003182.9A Pending CN115346051A (en) 2022-08-19 2022-08-19 Optical remote sensing image detection method and device

Country Status (1)

Country Link
CN (1) CN115346051A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116091781A (en) * 2022-11-18 2023-05-09 中国人民解放军军事科学院***工程研究院 Data processing method and device for image recognition

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116091781A (en) * 2022-11-18 2023-05-09 中国人民解放军军事科学院***工程研究院 Data processing method and device for image recognition

Similar Documents

Publication Publication Date Title
CN108960211B (en) Multi-target human body posture detection method and system
CN110796048B (en) Ship target real-time detection method based on deep neural network
CN111461213B (en) Training method of target detection model and target rapid detection method
CN110647802A (en) Remote sensing image ship target detection method based on deep learning
CN112001428A (en) Anchor frame-free target detection network training method based on feature matching optimization
CN112818871B (en) Target detection method of full fusion neural network based on half-packet convolution
CN110688883A (en) Vehicle and pedestrian detection method and device
CN111539456B (en) Target identification method and device
CN111274964B (en) Detection method for analyzing water surface pollutants based on visual saliency of unmanned aerial vehicle
CN114565842A (en) Unmanned aerial vehicle real-time target detection method and system based on Nvidia Jetson embedded hardware
CN113487600A (en) Characteristic enhancement scale self-adaptive sensing ship detection method
CN112949453A (en) Training method of smoke and fire detection model, smoke and fire detection method and smoke and fire detection equipment
CN115346051A (en) Optical remote sensing image detection method and device
CN115272691A (en) Training method, recognition method and equipment for steel bar binding state detection model
CN114565824A (en) Single-stage rotating ship detection method based on full convolution network
CN114612755A (en) Fire detection method, fire detection system, electronic equipment and readable storage medium
CN116229419B (en) Pedestrian detection method and device
CN114463624A (en) Method and device for detecting illegal buildings applied to city management supervision
CN110334703B (en) Ship detection and identification method in day and night image
CN116091781B (en) Data processing method and device for image recognition
CN111402185B (en) Image detection method and device
CN111797940A (en) Image identification method based on ocean search and rescue and related device
CN110956156A (en) Deep learning-based red light running detection system
CN113688810B (en) Target capturing method and system of edge device and related device
CN115713620A (en) Infrared small target detection method and device, computing equipment and storage medium

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