CN111563928A - Exception photo abnormity identification and reminding method and system - Google Patents
Exception photo abnormity identification and reminding method and system Download PDFInfo
- Publication number
- CN111563928A CN111563928A CN202010224094.6A CN202010224094A CN111563928A CN 111563928 A CN111563928 A CN 111563928A CN 202010224094 A CN202010224094 A CN 202010224094A CN 111563928 A CN111563928 A CN 111563928A
- Authority
- CN
- China
- Prior art keywords
- photo
- proof
- abnormal
- land
- target pattern
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a proof photo abnormity identification and reminding method and system, wherein the method comprises the following steps: s101: acquiring survey information sent by mobile survey equipment; s102: acquiring the distance between the proof photo and the target pattern spot and the shooting direction, judging whether the proof photo is abnormal according to the distance and the shooting direction, if so, executing S104, and if not, executing S103; s103: obtaining the land type information of the proof photo through a deep learning model, judging whether the proof photo of the same target pattern spot is abnormal or not according to the land type information, if so, executing S104, and if not, executing S105; s104: determining that the proof photo is abnormal; s105: and judging by combining the data in the original soil utilization status database through a rule engine to obtain the investigation result of the target pattern spot. The method and the device can automatically identify the abnormal photos and feed back corresponding reminding information, the identification is accurate, the repeated work is avoided, the investigation result is obtained through the rule engine, the speed is high, and the error is not easy to occur.
Description
Technical Field
The invention relates to the field of homeland survey, in particular to a method and a system for identifying and reminding an evidence-raising photo abnormity.
Background
The homeland survey is a basic work for researching land utilization conditions, in order to accurately find the utilization property of each land, a surveyor needs to carry out field survey, judges land types according to the land utilization conditions, planting conditions and the like of the map spots on the land, combines the current land utilization conditions, permanent basic farmlands and other data according to corresponding survey rules, and collects field photos reflecting the characteristics of the land types as evidences.
The technical scheme of the current homeland survey mainly integrates the work flow in a man-machine combination mode, and the specific operation and rule analysis and judgment are still mainly carried out manually. However, the homeland survey has considerable expertise and complexity, and requires the surveyor to have corresponding experience and ability to the land use feature recognition, the photo shooting method and the survey rule, and the prior art has the following three disadvantages:
(1) in the prior art, land utilization characteristics (land conditions, planting conditions and the like) are identified based on subjective cognition of investigators, and a large number of the investigators (each province is more than 1 ten thousand) participate in investigation, so that the subjective cognition is difficult to unify standards in a quantitative mode.
(2) The prior art can not control the reasonability of the collected picture and the shooting quality in real time, can not prompt typical errors such as improper execution of the shooting requirement, unreasonable shooting method and the like in time, can only carry out manual inspection and supplementary shooting afterwards, and leads to repeated work.
(3) The national soil survey rules are complex, and the method relates to analysis and comparison of the occupation ratio and the area of data such as the current land utilization situation, permanent basic farmland and the like, and mutual logical constraint. The existing technical scheme is executed by manually and mechanically contrasting rules, and is low in speed and easy to make mistakes.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides the evidence-taking photo abnormity identification and reminding method and system, which can automatically identify the abnormal evidence-taking photo and feed back corresponding reminding information when the evidence-taking photo is received, avoid repeated work, automatically identify the ground class corresponding to the evidence-taking photo through the deep learning model, do not depend on subjective cognition, realize more accurate identification, obtain the investigation result through the rule engine, and have high speed and low possibility of error.
In order to solve the above problems, the present invention adopts a technical solution as follows: a proof photo abnormity identification and reminding method comprises the following steps:
s101: acquiring survey information sent by mobile survey equipment, wherein the survey information comprises an evidence-providing photo, station site coordinates, an azimuth angle and target spot coordinates;
s102: acquiring the distance between the proof photo and the target pattern spot and the shooting direction, judging whether the proof photo is abnormal or not according to the distance and the shooting direction, if so, executing S104, and if not, executing S103;
s103: obtaining the land type information of the proof photo through a deep learning model, judging whether the proof photo of the same target pattern spot is abnormal or not according to the land type information, if so, executing S104, and if not, executing S105;
s104: determining that the proof photo is abnormal;
s105: and judging by combining the data in the original soil utilization status database through a rule engine to obtain the investigation result of the target pattern spot.
Further, the step of obtaining survey information sent by the mobile survey device further comprises:
training a sample set based on a deep learning platform to obtain a deep learning model, wherein the sample set comprises photos corresponding to all secondary land categories.
Further, each secondary land category in the sample set corresponds to at least one photo, and the photos corresponding to different secondary land categories are stored in folders at different points.
Further, the step of obtaining the distance between the proof photo and the target pattern spot and the shooting direction specifically includes: acquiring the distance between the proof-presenting photo and the target pattern spot according to the station point coordinate and the target pattern spot coordinate; and acquiring the shooting direction of the proof photo according to the station position coordinates and the azimuth angle.
Further, the proof photo comprises a long-range photo, a short-range photo and a feature photo.
Further, the step of judging whether the proof photo is abnormal according to the distance and the shooting direction specifically comprises:
judging whether the distance between the distant view picture and the target pattern spot exceeds a first preset value or not, and if so, determining that the distant view picture is abnormal;
and judging whether the distance between the close-range photo or the characteristic photo and the target pattern spot exceeds a second preset value, and if so, determining that the close-range photo or the characteristic photo is abnormal.
And taking a ray from the station site coordinate along the azimuth angle, judging whether the ray is intersected with the target pattern spot, and if not, determining that the proof photo is abnormal.
Further, the land type information comprises a land type and a confidence degree of the proof photo.
Further, the step of judging whether the proof photos of the same target pattern spot are abnormal according to the land type information specifically includes:
judging whether the close-range photo or the feature photo of the same target pattern spot comprises at least two kinds of land information and the confidence degree is lower than a preset value, if so, determining that the close-range photo or the feature photo is abnormal;
and judging whether the distant view photo, the close view photo and the feature photo are in the same primary land type or not and whether the secondary land types corresponding to the close view photo and the feature photo are consistent, if so, determining that the proof photo is abnormal.
Based on the same inventive concept, the application also provides an identification and reminding system for the abnormal testification photo, wherein the system comprises mobile investigation equipment and a background system;
the mobile survey equipment is used for shooting the proof-proving photo of the target pattern spot and sending the proof-proving photo to a background system;
the background system realizes the above abnormal identification and reminding method of the evidence-presenting photo according to the evidence-presenting photo.
Compared with the prior art, the invention has the beneficial effects that: can be when receiving the photo of proving, the unusual photo of proving of automatic identification to corresponding warning information is fed back, avoids work to relapse, through the ground class that degree of depth learning model automatic identification proves photo corresponds, does not rely on subjective cognition, and the discernment is more accurate, and obtains the investigation result through the rule engine, and is fast and be difficult to make mistakes.
Drawings
FIG. 1 is a flowchart of an embodiment of a proof photo anomaly identification and reminding method according to the present invention;
FIG. 2 is a block diagram of an embodiment of an exemplary system for identifying and alerting of anomalies in photographs;
fig. 3 is a flowchart of an embodiment of an identification and reminding method for an abnormality of an proof photo executed by a background system in the identification and reminding system for an abnormality of an proof photo according to the present invention.
Detailed Description
The present invention will be further described with reference to the accompanying drawings and the detailed description, and it should be noted that any combination of the embodiments or technical features described below can be used to form a new embodiment without conflict.
Referring to fig. 1, fig. 1 is a flowchart illustrating an abnormal identification and reminding method for a proof photo according to an embodiment of the present invention. The method for identifying and reminding the abnormal condition of the proof photo of the invention is described in detail with reference to the attached figure 1.
In this embodiment, a proof photo anomaly identification and reminding method includes:
s101: and acquiring survey information sent by the mobile survey equipment, wherein the survey information comprises a proof-drawing picture, station site coordinates, an azimuth angle and target spot coordinates.
In this embodiment, the mobile survey device includes a mobile phone, a tablet, a smart watch, and other devices capable of positioning in real time and sending the proof photo to the background system in a wireless transmission manner.
The background system for receiving the survey information can be a server, a computer, a cloud platform, a mobile phone and other equipment capable of processing the received information.
In this embodiment, the step of acquiring the survey information sent by the mobile survey device further includes: and training a sample set based on the deep learning platform to obtain a deep learning model, wherein the sample set comprises photos corresponding to all secondary land categories.
The sample set is the basis of the deep learning technology for identifying the evidence demonstrating photo features, and the accuracy of the sample set is directly imaged to achieve the effect of feature identification. The sample set for training is formed by screening and classifying and marking sample pictures in a manual mode, wherein the screening of the sample pictures takes the principles of accurate content, prominent main body and obvious characteristics as the principle.
In this embodiment, each secondary land category in the sample set corresponds to at least one photo, and the photos corresponding to different secondary land categories are stored in folders at different points.
In a specific embodiment, the homeland survey database standard defines 13 primary land classes corresponding to 57 secondary land classes, and 4 secondary land classes (1301-.
Table one, map code and map name comparison table
In this embodiment, the deep learning platform trains a deep learning model for the easy dl deep learning platform to obtain proof photos, and a system is built in combination with open source components such as postgis and leaflets.
In a specific embodiment, the deep learning model is an easy dl optimized image classification model.
In this embodiment, the mobile survey device receives an instruction input by a surveyor, collects survey information of the target pattern according to the instruction, and sends the survey information to the background system.
In the present embodiment, the types of proof photos include a long-range photo, a short-range photo, and a feature photo.
In a specific embodiment, the operation instruction input by the user to the mobile survey device is to select a pattern spot as a target pattern spot in the software interface, and the mobile survey device collects the proof photo according to the operation instruction. Usually, when a target pattern is investigated and evidence-raising photos are collected, the evidence-raising photos required to be shot comprise 1 long shot, 2 short shots and 2 characteristic photos, and the mobile investigation equipment collects the evidence-raising photos and records a standing point (longitude and latitude form) and an azimuth angle of each photo when the photo is shot through a GPS module. After the collection is completed, the file for storing the proof photo, the station position coordinate, the azimuth angle and the target pattern spot coordinate string when the proof photo is shot are submitted to a background system through a 4G network for analysis.
In other embodiments, the mobile survey equipment may also be an unmanned aerial vehicle, and after determining that the target pattern spot of the proof picture needs to be collected according to an operation instruction input by a user, the unmanned aerial vehicle moves to an area where the target pattern spot is located and shoots the proof picture of the target pattern spot.
S102: and acquiring the distance between the proof photo and the target pattern spot and the shooting direction, judging whether the proof photo is abnormal according to the distance and the shooting direction, if so, executing S104, and if not, executing S103.
In this embodiment, the step of obtaining the distance between the proof photo and the target spot and the shooting direction specifically includes: acquiring the distance between the proof photo and the target pattern spot according to the station point coordinate and the target pattern spot coordinate; and acquiring the shooting direction of the proof photo according to the coordinates and the azimuth of the station position point.
In this embodiment, the step of determining whether the proof photo is abnormal according to the distance and the shooting direction specifically includes: judging whether the distance between the distant view picture and the target pattern spot exceeds a first preset value or not, and if so, determining that the distant view picture is abnormal; and judging whether the distance between the close-range photo or the characteristic photo and the target image spot exceeds a second preset value or not, and if so, determining that the close-range photo or the characteristic photo is abnormal.
In a specific embodiment, the first preset value is 100 meters and the second preset value is 20 meters. For each proof photo, the coordinates of the station position at the time of shooting are taken as the coordinates of the proof photo. And calculating the shortest distance between the station point and the target spot graph according to the station point coordinate during shooting, if the distance between the distant view picture and the target spot is more than 100 meters or the distance between the distant view picture and the target spot is more than 20 meters, determining that the corresponding 1 distant view picture or the near view picture and the characteristic picture are abnormal, and feeding back error prompt information of 'shooting distance is not in accordance with requirements' through the mobile investigation equipment.
In this embodiment, the step of determining whether the proof photo is abnormal according to the distance and the shooting direction further includes: and (4) making a ray from the station site coordinate along the azimuth angle, judging whether the ray is intersected with the target pattern spot, and if not, determining that the proof photo is abnormal.
In a specific embodiment, for each photo, according to the coordinates and azimuth of a station point during shooting, a ray is made in the direction of the shooting azimuth by taking the station point as an end point, if the ray intersects with a target pattern spot, the direction of the photo is considered to meet the requirement, if the ray does not intersect with the target pattern spot, the abnormal condition of the proof photo is determined, and prompt information that the shooting direction does not meet the requirement is fed back through mobile investigation equipment.
In the embodiment, the evidence-proving photo sent by the mobile investigation equipment is analyzed and judged to meet the requirement through the background system, and the feedback is carried out through the mobile investigation equipment in real time, so that the evidence-proving photo does not need to be shot again afterwards, and the repeated work is avoided.
In other embodiments, the step of determining whether the proof photo meets the requirement may also be performed on the mobile survey device, and the user analyzes the taken proof photo through the mobile survey device, determines whether the proof photo meets the requirement, and feeds back corresponding information.
S103: and acquiring the land type information of the proof photo through the deep learning model, judging whether the proof photo of the same target pattern spot is abnormal according to the land type information, if so, executing S104, and if not, executing S105.
In the present embodiment, the land information includes land information of the proof photo and the confidence of the recognition type.
The feature content reflected by the proof photo is the basis of analysis work such as land type recognition, photo conformity inspection and the like, so that each proof photo is firstly recognized by adopting a trained easy DL model, and the land type information of each proof photo and the confidence coefficient of the proof photo and the land type information are recorded. Wherein, the identification result of the proof photo may have a plurality of values, as shown in the following table two:
| serial number | Type (B) | Confidence level |
| 1 | 0101 (Paddy field) | 90％ |
| 2 | 0102 (Water pouring land) | 30％ |
| 3 | 0404 (other grass land) | 10％ |
Table two, demonstrating possible place information and confidence of the photo
In this embodiment, the step of determining whether the proof photos of the same target pattern are abnormal according to the land type information specifically includes: judging whether the close-range photo or the feature photo of the same target pattern spot comprises at least two kinds of land information and the confidence degree is lower than a preset value, if so, determining that the close-range photo or the feature photo is abnormal; and judging whether the distant view picture, the close view picture and the feature picture belong to the same primary land class and whether the secondary land classes corresponding to the close view picture and the feature picture are consistent, and if not, determining that the proof-proving picture is abnormal.
In a specific embodiment, the preset value is 80%, for the close-range and feature photos of the same target spot, the land information identified based on the deep learning model should be relatively clear, and if the identification result is a plurality of land types and the accuracy corresponding to each land type is lower than 80%, there may be a reason that the features are not highlighted during shooting or the land type features are relatively comprehensive and difficult to distinguish. The method for processing the situation is to feed back information of 'the land types can not be accurately identified' and a plurality of land type options through the mobile survey equipment, and a surveyor manually selects the land types or retakes the photos.
In another specific embodiment, for the distant view photos of the same target spot, because the distance is long and the contents of the images are difficult to unify, the confidence of the recognition result is not accurate, and the distant view photos and the close view or the feature photos are in the same class of land in principle, otherwise, the mobile survey equipment feeds back the prompt of 'inconsistent photo features'; and (4) requiring the secondary ground classes to be consistent for the close-range or characteristic photos of the same target, and otherwise, feeding back a prompt of 'inconsistent photo characteristics'.
S104: and determining the proof photo exception.
In this embodiment, when the proof photo is judged to be not in accordance with the requirement or not to have the characteristic consistency, the proof photo is determined to be abnormal, and the mobile survey equipment feeds back related information to the user, so that the user can shoot the photo of the target pattern again.
S105: and judging by combining the data in the original soil utilization status database through a rule engine to obtain the investigation result of the target pattern spot.
The homeland survey is not a simple land utilization characteristic survey, and besides distinguishing the conditions and planting conditions of the land on the spot, the land on the spot needs to be analyzed and determined according to the land types of the original land utilization database. For example: for cultivated lands in the raw land utilization status database, if the ground features are other grasslands (0404), the land is still surveyed according to dry land and marked as "not cultivated", and the survey is not allowed to be other grasslands.
In the present embodiment, in order to improve the accuracy of the survey result, after the land information of the proof photo is acquired by the deep learning model, the raw soil utilization data, the land information, and information related to other homeland survey rules are combined with the rule engine, and the data are processed by the rule engine to acquire the survey result of the target pattern.
In a specific embodiment, a rule engine is applied to analyze and prejudge and provide result options by combining data such as a photo feature recognition result, a land utilization status situation, a permanent basic farmland proportion analysis result and the like, and sequencing is performed according to the provided accuracy when a plurality of options exist so as to assist field investigation.
In the field check work of homeland survey, the manner of performing land type interpretation, evidence-raised photo collection and logic rule analysis based on a personal experience method has large subjective factors, the standards are difficult to unify, the result cannot be evaluated quantitatively, and the carelessness, errors and even repeated work are easy to cause. The method is based on the evidence-raised photo feature recognition method, combines station sites and azimuth angles during collection of evidence-raised photos, applies rule conversion and evaluation according to the superposition analysis result of target pattern spots and the current utilization data of the original soil, realizes photo conformance inspection, provides auxiliary suggestions for survey results, and improves the accuracy of the survey results.
Has the advantages that: according to the evidence-taking photo abnormity identification and reminding method, when the evidence-taking photo is received, the abnormal evidence-taking photo can be automatically identified, corresponding reminding information is fed back, repeated work is avoided, the ground class corresponding to the evidence-taking photo is automatically identified through the deep learning model, subjective cognition is not relied on, identification is more accurate, and the investigation result is obtained through the rule engine, so that the speed is high, and mistakes are not easy to make.
Based on the same inventive concept, the invention further provides an identification and reminding system for abnormal testifying photos, please refer to fig. 2 and fig. 3, wherein fig. 2 is a structural diagram of an embodiment of the identification and reminding system for abnormal testifying photos according to the invention; fig. 3 is a flowchart of an embodiment of an identification and reminding method for an abnormality of an attestation photo executed by a background system in the identification and reminding system for an abnormality of an attestation photo according to the present invention, and the identification and reminding system for an abnormality of an attestation photo according to the present invention is specifically described with reference to fig. 2 and 3.
In this embodiment, the field survey system includes a mobile survey device and a background system;
the mobile investigation equipment is used for shooting the proof photo of the target pattern spot and sending the proof photo to the background system:
the background system realizes the following identification and reminding method for the abnormality of the proof photo according to the proof photo:
s201: acquiring survey information sent by mobile survey equipment, wherein the survey information comprises a proof photo, station site coordinates, an azimuth angle and target spot coordinates.
In this embodiment, the mobile survey device includes a mobile phone, a tablet, a smart watch, and other devices capable of positioning in real time and sending the proof photo to the background system in a wireless transmission manner.
The background system for receiving the survey information can be a server, a computer, a cloud platform, a mobile phone and other equipment capable of processing the received information.
In this embodiment, the step of acquiring the survey information sent by the mobile survey device further includes: and training a sample set based on the deep learning platform to obtain a deep learning model, wherein the sample set comprises photos corresponding to all secondary land categories.
The sample set is the basis of the deep learning technology for identifying the evidence demonstrating photo features, and the accuracy of the sample set is directly imaged to achieve the effect of feature identification. The sample set for training is formed by screening and classifying and marking sample pictures in a manual mode, wherein the screening of the sample pictures takes the principles of accurate content, prominent main body and obvious characteristics as the principle.
In this embodiment, each secondary land category in the sample set corresponds to at least one photo, and the photos corresponding to different secondary land categories are stored in folders at different points.
In a specific embodiment, 13 primary land categories are defined by the standard of the homeland survey database, which correspond to 57 secondary land categories, and 4 secondary land categories (1301- & 1304) are added according to the work requirement, that is, the sample set comprises 61 secondary land categories, the specific content is shown in the following table III, 200 sample photos are respectively prepared corresponding to each secondary land category, each secondary land category photo is stored in a folder, and the folders are named by secondary land category codes.
Table three, map code and map name comparison table
In this embodiment, the deep learning platform is an easy dl deep learning platform, a deep learning model of the proof photo is obtained through training, and a system is built by combining open source components such as postgis and leaflets.
In a specific embodiment, the deep learning model is an easy dl optimized image classification model.
In this embodiment, the mobile survey device receives an instruction input by a surveyor, collects survey information of the target pattern according to the instruction, and sends the survey information to the background system.
In the present embodiment, the types of proof photos include a long-range photo, a short-range photo, and a feature photo.
In a specific embodiment, the operation instruction input by the user to the mobile survey device is to select a pattern spot as a target pattern spot in the software interface, and the mobile survey device collects the proof photo according to the operation instruction. Usually, when a target pattern is investigated and evidence-raising photos are collected, the evidence-raising photos required to be shot comprise 1 long shot, 2 short shots and 2 characteristic photos, and the mobile investigation equipment collects the evidence-raising photos and records a standing point (longitude and latitude form) and an azimuth angle of each photo when the photo is shot through a GPS module. After the collection is completed, the file for storing the proof photo, the station position coordinate, the azimuth angle and the target pattern spot coordinate string when the proof photo is shot are submitted to a background system through a 4G network for analysis.
In other embodiments, the mobile survey equipment may also be an unmanned aerial vehicle, and after determining that the target pattern spot of the proof picture needs to be collected according to an operation instruction input by a user, the unmanned aerial vehicle moves to an area where the target pattern spot is located and shoots the proof picture of the target pattern spot.
S202: and acquiring the distance between the proof photo and the target pattern spot and the shooting direction, judging whether the proof photo is abnormal according to the distance and the shooting direction, if so, executing S204, and if not, executing S203.
In this embodiment, the step of obtaining the distance between the proof photo and the target spot and the shooting direction specifically includes: acquiring the distance between the proof photo and the target pattern spot according to the station point coordinate and the target pattern spot coordinate; and acquiring the shooting direction of the proof photo according to the coordinates and the azimuth of the station position point.
In this embodiment, the step of determining whether the proof photo is abnormal according to the distance and the shooting direction specifically includes: judging whether the distance between the distant view picture and the target pattern spot exceeds a first preset value or not, and if so, determining that the distant view picture is abnormal; and judging whether the distance between the close-range photo or the characteristic photo and the target image spot exceeds a second preset value or not, and if so, determining that the close-range photo or the characteristic photo is abnormal.
In a specific embodiment, the first preset value is 100 meters and the second preset value is 20 meters. For each proof photo, the coordinates of the station position at the time of shooting are taken as the coordinates of the proof photo. And calculating the shortest distance between the station point and the target spot graph according to the station point coordinate during shooting, if the distance between the distant view picture and the target spot is more than 100 meters or the distance between the distant view picture and the target spot is more than 20 meters, determining that the corresponding 1 distant view picture or the near view picture and the characteristic picture are abnormal, and feeding back error prompt information of 'shooting distance is not in accordance with requirements' through the mobile investigation equipment.
In this embodiment, the step of determining whether the proof photo is abnormal according to the distance and the shooting direction further includes: and (4) making a ray from the station site coordinate along the azimuth angle, judging whether the ray is intersected with the target pattern spot, and if not, determining that the proof photo is abnormal.
In a specific embodiment, for each photo, according to the coordinates and azimuth of a station point during shooting, a ray is made in the direction of the shooting azimuth by taking the station point as an end point, if the ray intersects with a target pattern spot, the direction of the photo is considered to meet the requirement, if the ray does not intersect with the target pattern spot, the abnormal condition of the proof photo is determined, and prompt information that the shooting direction does not meet the requirement is fed back through mobile investigation equipment.
In the embodiment, the evidence-proving photo sent by the mobile investigation equipment is analyzed and judged to meet the requirement through the background system, and the feedback is carried out through the mobile investigation equipment in real time, so that the evidence-proving photo does not need to be shot again afterwards, and the repeated work is avoided.
In other embodiments, the step of determining whether the proof photo meets the requirement may also be performed on the mobile survey device, and the user analyzes the taken proof photo through the mobile survey device, determines whether the proof photo meets the requirement, and feeds back corresponding information.
S203: and acquiring the land type information of the proof photo through a deep learning model, judging whether the proof photo of the same target pattern spot is abnormal or not according to the land type information, if so, executing S204, and if not, executing S205.
In the present embodiment, the land information includes land information of the proof photo and the confidence of the recognition type.
The feature content reflected by the proof photo is the basis of analysis work such as land type recognition, photo conformity inspection and the like, so that each proof photo is firstly recognized by adopting a trained easy DL model, and the land type information of each proof photo and the confidence coefficient of the proof photo and the land type information are recorded. Wherein, the identification result of the proof photo may have a plurality of values, as shown in the following table four:
| serial number | Type (B) | Confidence level |
| 1 | 0101 (Paddy field) | 90％ |
| 2 | 0102 (Water pouring land) | 30％ |
| 3 | 0404 (other grass land) | 10％ |
Fourth, prove the possible place information and confidence of the photo
In this embodiment, the step of determining whether the proof photos of the same target pattern are abnormal according to the land type information specifically includes: judging whether the close-range photo or the feature photo of the same target pattern spot comprises at least two kinds of land information and the confidence degree is lower than a preset value, if so, determining that the close-range photo or the feature photo is abnormal; and judging whether the distant view picture, the close view picture and the feature picture belong to the same primary land class and whether the secondary land classes corresponding to the close view picture and the feature picture are consistent, and if not, determining that the proof-proving picture is abnormal.
In a specific embodiment, the preset value is 80%, for the close-range and feature photos of the same target spot, the land information identified based on the deep learning model should be relatively clear, and if the identification result is a plurality of land types and the accuracy corresponding to each land type is lower than 80%, there may be a reason that the features are not highlighted during shooting or the land type features are relatively comprehensive and difficult to distinguish. The method for processing the situation is to feed back information of 'the land types can not be accurately identified' and a plurality of land type options through the mobile survey equipment, and a surveyor manually selects the land types or retakes the photos.
In another specific embodiment, for the distant view photos of the same target spot, because the distance is long and the contents of the images are difficult to unify, the confidence of the recognition result is not accurate, and the distant view photos and the close view or the feature photos are in the same class of land in principle, otherwise, the mobile survey equipment feeds back the prompt of 'inconsistent photo features'; and (4) requiring the secondary ground classes to be consistent for the close-range or characteristic photos of the same target, and otherwise, feeding back a prompt of 'inconsistent photo characteristics'.
S204: and determining the proof photo exception.
In this embodiment, when the proof photo is judged to be not in accordance with the requirement or not to have the characteristic consistency, the proof photo is determined to be abnormal, and the mobile survey equipment feeds back related information to the user, so that the user can shoot the photo of the target pattern again.
S205: and judging by combining the data in the original soil utilization status database through a rule engine to obtain the investigation result of the target pattern spot.
The homeland survey is not a simple land utilization characteristic survey, and besides distinguishing the conditions and planting conditions of the land on the spot, the land on the spot needs to be analyzed and determined according to the land types of the original land utilization database. For example: for cultivated land in the raw land utilization status database, if the local characteristics are grasslands, the investigation is still conducted according to dry land and marked as "uncultivated", and the investigation is not allowed to be conducted for other grasslands.
In the present embodiment, in order to improve the accuracy of the survey result, after the land information of the proof photo is acquired by the deep learning model, the raw soil utilization data, the land information, and information related to other homeland survey rules are combined with the rule engine, and the data are processed by the rule engine to acquire the survey result of the target pattern.
In a specific embodiment, a rule engine is applied to analyze and prejudge and provide result options by combining data such as a photo feature recognition result, a land utilization status situation, a permanent basic farmland proportion analysis result and the like, and sequencing is performed according to the provided accuracy when a plurality of options exist so as to assist field investigation.
In the field check work of homeland survey, the manner of performing land type interpretation, evidence-raised photo collection and logic rule analysis based on a personal experience method has large subjective factors, the standards are difficult to unify, the result cannot be evaluated quantitatively, and the carelessness, errors and even repeated work are easy to cause. The method is based on the evidence-raised photo feature recognition method, combines station sites and azimuth angles during collection of evidence-raised photos, applies rule conversion and evaluation according to the superposition analysis result of target pattern spots and the current utilization data of the original soil, realizes photo conformance inspection, provides auxiliary suggestions for survey results, and improves the accuracy of the survey results.
Has the advantages that: the evidence-taking photo abnormity identification and reminding system can automatically identify abnormal evidence-taking photos and feed back corresponding reminding information when the evidence-taking photos are received, so that repeated work is avoided, the ground types corresponding to the evidence-taking photos are automatically identified through the deep learning model, subjective cognition is not relied on, the identification is more accurate, and the investigation result is obtained through the rule engine, so that the speed is high, and mistakes are not easy to make.
The above embodiments are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited thereby, and any insubstantial changes and substitutions made by those skilled in the art based on the present invention are within the protection scope of the present invention.
Claims (9)
1. A proof photo abnormity identification and reminding method is characterized by comprising the following steps:
s101: acquiring survey information sent by mobile survey equipment, wherein the survey information comprises an evidence-providing photo, station site coordinates, an azimuth angle and target spot coordinates;
s102: acquiring the distance between the proof photo and the target pattern spot and the shooting direction, judging whether the proof photo is abnormal or not according to the distance and the shooting direction, if so, executing S104, and if not, executing S103;
s103: obtaining the land type information of the proof photo through a deep learning model, judging whether the proof photo of the same target pattern spot is abnormal or not according to the land type information, if so, executing S104, and if not, executing S105;
s104: determining that the proof photo is abnormal;
s105: and judging by combining the data in the original soil utilization status database through a rule engine to obtain the investigation result of the target pattern spot.
2. The method for identifying and reminding the abnormality of the proof photo according to claim 1, wherein the step of obtaining the survey information sent by the mobile survey equipment further comprises:
training a sample set based on a deep learning platform to obtain a deep learning model, wherein the sample set comprises photos corresponding to all secondary land categories.
3. The method for identifying and reminding an abnormality of a proof photo according to claim 2, wherein each secondary land category in the sample set corresponds to at least one photo, and the photos corresponding to different secondary land categories are stored in folders at different points.
4. The method according to claim 1, wherein the step of obtaining the distance between the proof photo and the target spot and the shooting direction specifically comprises:
acquiring the distance between the proof-presenting photo and the target pattern spot according to the station point coordinate and the target pattern spot coordinate;
and acquiring the shooting direction of the proof photo according to the station position coordinates and the azimuth angle.
5. The method for abnormality recognition and reminder of a proof photo of claim 1, wherein the proof photo includes a long shot photo, a short shot photo, and a feature photo.
6. The method according to claim 5, wherein the step of determining whether the proof photo is abnormal according to the distance and the shooting direction specifically comprises:
judging whether the distance between the distant view picture and the target pattern spot exceeds a first preset value or not, and if so, determining that the distant view picture is abnormal;
and judging whether the distance between the close-range photo or the characteristic photo and the target pattern spot exceeds a second preset value, and if so, determining that the close-range photo or the characteristic photo is abnormal.
And taking a ray from the station site coordinate along the azimuth angle, judging whether the ray is intersected with the target pattern spot, and if not, determining that the proof photo is abnormal.
7. The method for recognizing and reminding the abnormality of the proof photo according to the claim 5, characterized in that the land type information comprises the land type and the confidence level of the proof photo.
8. The method according to claim 7, wherein the step of determining whether the proof photo of the same target pattern is abnormal according to the land type information specifically comprises:
judging whether the close-range photo or the feature photo of the same target pattern spot comprises at least two kinds of land information and the confidence degree is lower than a preset value, if so, determining that the close-range photo or the feature photo is abnormal;
and judging whether the distant view photo, the close view photo and the feature photo belong to the same primary land type or not and whether the secondary land types corresponding to the close view photo and the feature photo are consistent or not, and if not, determining that the proof-proving photo is abnormal.
9. An evidence-raising photo abnormity identification and reminding system is characterized by comprising mobile investigation equipment and a background system;
the mobile survey equipment is used for shooting a proof-proving photo of the target pattern spot and sending the proof-proving photo to the background system;
the background system realizes the evidence photo abnormity identification and reminding method according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010224094.6A CN111563928B (en) | 2020-03-26 | 2020-03-26 | Exception photo abnormity identification and reminding method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010224094.6A CN111563928B (en) | 2020-03-26 | 2020-03-26 | Exception photo abnormity identification and reminding method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111563928A true CN111563928A (en) | 2020-08-21 |
| CN111563928B CN111563928B (en) | 2021-05-25 |
Family
ID=72071493
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010224094.6A Active CN111563928B (en) | 2020-03-26 | 2020-03-26 | Exception photo abnormity identification and reminding method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111563928B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113920144A (en) * | 2021-09-30 | 2022-01-11 | 广东省国土资源测绘院 | Real-scene photo ground vision field analysis method and system |
| CN114596444A (en) * | 2021-08-30 | 2022-06-07 | 广东省国土资源测绘院 | Proof photo auxiliary auditing method, electronic equipment and storage medium |
| CN114840620A (en) * | 2022-04-01 | 2022-08-02 | 广东省国土资源测绘院 | Proof material management method, intelligent terminal and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060126935A1 (en) * | 2003-07-24 | 2006-06-15 | Tunney William P | Method and system for identifying multiple questionnaire pages |
| CN107516282A (en) * | 2017-08-01 | 2017-12-26 | 青岛市房地产信息与交易资金监管中心 | Three look into the Land Change Survey information management system of unification |
| CN108363951A (en) * | 2018-01-11 | 2018-08-03 | 东南大学 | Remote sensing image the corresponding deep learning sample database of class identification automatic obtaining method |
| CN109063026A (en) * | 2018-07-16 | 2018-12-21 | 宁夏回族自治区国土资源地理信息中心（宁夏国土资源档案馆） | A kind of geographic information data acquisition method and system |
| CN109559267A (en) * | 2018-11-24 | 2019-04-02 | 杭州吉翱世联土地勘测规划设计有限公司 | Land investigation internet+the means of proof, device, equipment and storage medium |
-
2020
- 2020-03-26 CN CN202010224094.6A patent/CN111563928B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060126935A1 (en) * | 2003-07-24 | 2006-06-15 | Tunney William P | Method and system for identifying multiple questionnaire pages |
| CN107516282A (en) * | 2017-08-01 | 2017-12-26 | 青岛市房地产信息与交易资金监管中心 | Three look into the Land Change Survey information management system of unification |
| CN108363951A (en) * | 2018-01-11 | 2018-08-03 | 东南大学 | Remote sensing image the corresponding deep learning sample database of class identification automatic obtaining method |
| CN109063026A (en) * | 2018-07-16 | 2018-12-21 | 宁夏回族自治区国土资源地理信息中心（宁夏国土资源档案馆） | A kind of geographic information data acquisition method and system |
| CN109559267A (en) * | 2018-11-24 | 2019-04-02 | 杭州吉翱世联土地勘测规划设计有限公司 | Land investigation internet+the means of proof, device, equipment and storage medium |
Non-Patent Citations (2)
| Title |
|---|
| 陈景平 等: "浅析第三次土地调查总体质量控制", 《江西测绘》 * |
| 黄锦凤 等: "第三次国土调查数据库建设及成果质量控制研究", 《国土资源导刊》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114596444A (en) * | 2021-08-30 | 2022-06-07 | 广东省国土资源测绘院 | Proof photo auxiliary auditing method, electronic equipment and storage medium |
| CN113920144A (en) * | 2021-09-30 | 2022-01-11 | 广东省国土资源测绘院 | Real-scene photo ground vision field analysis method and system |
| CN114840620A (en) * | 2022-04-01 | 2022-08-02 | 广东省国土资源测绘院 | Proof material management method, intelligent terminal and storage medium |
| CN114840620B (en) * | 2022-04-01 | 2023-06-27 | 广东省国土资源测绘院 | Evidence-providing material management method, intelligent terminal and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111563928B (en) | 2021-05-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111563928B (en) | Exception photo abnormity identification and reminding method and system | |
| CN111597377B (en) | Deep learning technology-based field investigation method and system | |
| US11881006B2 (en) | Machine learning assistant for image analysis | |
| US20110240730A1 (en) | Soil Sample Tracking System and Method | |
| US10444122B2 (en) | Soil sample tracking system and method | |
| US20200304954A1 (en) | Positioning method, positioning apparatus, server, and computer-readable storage medium | |
| CN113360587B (en) | Land surveying and mapping equipment and method based on GIS technology | |
| US20170345017A1 (en) | Automation assisted elevation certificate production system | |
| US20240151851A1 (en) | Lidar data and structural modeling based elevation determination | |
| CN111260182A (en) | Method, device and storage medium for on-site checking and evidence obtaining of natural resource asset audit | |
| EP3293486A1 (en) | Method and system for improving spatial accuracy of map data | |
| CN113537293A (en) | Wheat lodging area identification method based on unmanned aerial vehicle and full convolution neural network | |
| US20190279226A1 (en) | Automation assisted elevation certificate production system | |
| CN112631333B (en) | Target tracking method and device of unmanned aerial vehicle and image processing chip | |
| CN108734393A (en) | Matching process, user equipment, storage medium and the device of information of real estate | |
| CN107657262A (en) | A kind of computer automatic sorting Accuracy Assessment | |
| CN108764365A (en) | A kind of device signboard detection method | |
| CN116266402A (en) | Automatic object labeling method and device, electronic equipment and storage medium | |
| CN112308740A (en) | House area intelligent right adjusting method and system | |
| CN110390286B (en) | Vehicle trajectory tracking test data manufacturing method and device, and prediction and result testing method | |
| Van Zijl et al. | In pursuit of a South African national soil database: potential and pitfalls of combining different soil data sets | |
| CN113297998B (en) | Method, device, equipment and storage medium for identifying territorial space planning problem | |
| US20240046626A1 (en) | Drone-based Vegetation Assessment | |
| US11941880B2 (en) | System and method for image-based crop identification | |
| KR102570846B1 (en) | Field compost detection method and system capable of pollutant management |
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 |