CN-121982582-A - Intelligent construction land monitoring system based on unmanned aerial vehicle images

CN121982582ACN 121982582 ACN121982582 ACN 121982582ACN-121982582-A

Abstract

The invention relates to the technical field of construction monitoring, in particular to an intelligent construction land monitoring system based on unmanned aerial vehicle aerial images, which comprises a data acquisition module, a construction scheme docking module, a construction model construction module and a monitoring platform, wherein the data acquisition module is used for acquiring image data of a construction area, the construction scheme docking module is used for importing construction scheme information, the construction model construction module is used for constructing a reference model, and the monitoring platform is used for dynamically monitoring the construction area through data acquisition of the data acquisition module, the construction scheme docking module and the construction model construction module. The invention solves the problems of low inspection efficiency, lag finding due to violations, complicated evidence obtaining process and other pain points in the traditional construction supervision mode, and comprehensively improves the instantaneity, accuracy and traceability of land supervision.

Inventors

SUN HUI
RONG YAN
ZHANG XINYUE
WEN SENYUAN
LIU JIAXING
Lan Qiuzu
Peng Yidiao
Qin Baiqing
LIN FENG
WEI HUA
Lan Songyao
MA WENHUI
LIANG QIANQIAN
LU YUGE
WU MENGLAN
LI YONGJIAN
LI YONGYOU
HOU KAIWEN
XIE WEIWEI
HAN YU
CHEN XIAOQIANG
LI SHIJIAN

Assignees

广西路桥工程集团有限公司

Dates

Publication Date: 20260505
Application Date: 20260106

Claims (10)

1. The intelligent construction land monitoring system based on the unmanned aerial vehicle image is characterized by comprising a data acquisition module (1), a construction scheme docking module (2), a construction model construction module (3) and a monitoring platform (4), The data acquisition module (1) is used for acquiring the unmanned aerial vehicle aerial image so as to obtain image data of a construction area; The construction scheme docking module (2) is used for importing construction scheme information so as to obtain construction area data, construction project data and construction progress data; The construction model construction module (3) is used for acquiring data of the data acquisition module (1) and the construction scheme docking module (2) so as to construct a reference model according to image data before project construction; The monitoring platform (4) is used for acquiring data of the data acquisition module (1), the construction scheme docking module (2) and the construction model construction module (3), the monitoring platform (4) regularly shoots the construction area through an unmanned plane in an aerial way to obtain construction image data, the monitoring platform (4) updates the reference model according to the construction image data to generate a construction model in the construction model construction module (3), and the monitoring platform (4) analyzes the construction model according to the construction area data, the construction project data and the construction progress data to dynamically monitor the construction area.
2. The intelligent monitoring system for construction sites based on unmanned aerial vehicle images according to claim 1, wherein the construction model construction module (3) comprises a data preprocessing sub-module (31), a labeling sub-module (32), a data fusion sub-module (33) and a visualization processing sub-module (34) The data preprocessing sub-module (31) is used for acquiring data of the data acquisition module (1) and the construction scheme docking module (2) so as to perform illumination equalization processing on original images and POS positioning data in the image data to enhance texture characteristics of key targets, and the data preprocessing sub-module (31) is used for positioning the image data and constructing a spatial index table through red line data in the construction area data so as to match the image data with the red line data; The marking sub-module (32) marks the typical temporary land type of the image data in the data preprocessing sub-module (31) through a marking model and marks the red line data; The data fusion sub-module (33) is used for acquiring data of the data preprocessing sub-module (31), the data fusion sub-module (33) converts the image data and the red line data into the same coordinates, and calculates the intersection area of the construction area and the red line data boundary; The visualization processing sub-module (34) is used for data acquisition of the data preprocessing sub-module (31), the labeling sub-module (32) and the data fusion sub-module (33) so as to form the reference model.
3. The intelligent monitoring system for construction land based on unmanned aerial vehicle images, which is characterized in that the labeling submodule (32) constructs a labeling data set based on historical construction images, pixel-level labeling is carried out on typical temporary land types such as a soil taking field, a spoil field and the like and red line areas, and the labeling submodule (32) trains a lightweight semantic segmentation model by combining migration learning and data enhancement technology so as to construct and obtain the labeling model, wherein the expression of the labeling model is as follows: formula (1) Wherein, the Is a labeling model; searching for parameter values that minimize the objective function for parameter optimization operations; is a model parameter, including all learnable weights and biases in the neural network; is a super-parameter weight used for balancing the contribution of each loss; Dividing a loss function for semantics; Dividing a loss function for semantics; learning a loss function for migration; Enhancing a loss function for the data; Enhancing a loss function for the data; formula (2) Wherein, the Is the total number of pixels; the method is characterized in that the method is a true label value, 0 is a background, 1 is a soil taking field, 2 is a spoil field, and 3 is a red line; Predicting probability for the model; For the foreground and background balance weight, the temporary area is usually smaller, and usually 0.75-0.9 is taken; For the Focal Loss parameter, 2 is typically taken to enhance the focus on difficult-to-classify pixels; Formula (3) Wherein, the As a teacher model, a large model pre-trained on remote sensing images, such as DeepLabV3+; is a student model, a lightweight model, such as MobileUnet; the parameters of the pre-training model are usually obtained by training on a LandDiscover K remote sensing dataset; is a model parameter, including all learnable weights and biases in the neural network; Formula (4) Wherein, the To enhance the number of samples; Is the first Enhancing the transformed image by using seed data; Corresponding to the transformed label to ensure the synchronous transformation with the image; Formula (5) Wherein, the Regularization is carried out on L1, so that sparsity of model parameters is promoted; for calculating the weight coefficient of the quantity, controlling the balance of the model size and the performance; To measure model inference complexity for the calculation amount.
4. The intelligent monitoring system for construction land based on unmanned aerial vehicle images according to claim 3, wherein the marking sub-module (32) is further used for special processing of red line marking: Formula (6) Wherein, the The number of red line samples; for the boundary and region balance coefficient, usually 0.6-0.8 is taken; as a boundary loss function, special attention is paid to the accuracy of the red line boundary; and the cross-over ratio balance loss improves the positioning precision of the red line area.
5. The intelligent monitoring system for construction sites based on unmanned aerial vehicle images according to claim 1, wherein the monitoring platform (4) comprises a patrol module (41), a model updating module (42) and an analysis module (43), The inspection module (41) performs aerial photography on the construction area at regular time through an unmanned plane, and the inspection module (41) obtains construction image data through the data acquisition module (1); The model updating module (42) updates the reference model according to the construction image data through the construction model constructing module (3) so as to obtain the construction model; The analysis module (43) is used for data acquisition of the model updating module (42) to compare the construction model with the reference model so as to mark a construction area which is in a red line range in the construction model.
6. The intelligent monitoring system for construction sites based on unmanned aerial vehicle images according to claim 5, wherein the analysis module (43) identifies red line range areas in the construction model and the reference model through a semantic segmentation model, the analysis module (43) compares the red line range areas of the construction model with the red line range areas in the reference model at a pixel level, and the analysis module (43) accurately locates newly added construction areas in the red line range through a change detection algorithm.
7. The intelligent monitoring system for construction land based on unmanned aerial vehicle images according to claim 6, wherein: the analysis module (43) identifies the red line range expression as: formula (7) Wherein, the A red line mask for the construction model; the model selection can adopt 3D U-Net or PointNet ++ to process three-dimensional point cloud or voxel data and output a high-precision segmentation mask; And the analysis module (43) performs pixel level comparison and newly added region calculation after three-dimensional registration and coordinate alignment, wherein the expression of the pixel level comparison is as follows: Formula (8) Wherein, the Mask for newly added construction area, and newly added area is extracted as ; The analysis module (43) calculates the degree of change between the construction model and the reference model by MAD or CVA: Formula (9) Wherein, the To indicate the intensity of the change, the larger the value is, the more remarkable the change is, by setting a threshold value Screening out significant change areas: 。
8. The intelligent monitoring system for construction sites based on unmanned aerial vehicle images according to claim 5, wherein the monitoring platform (4) further comprises a manual auditing model (44), the manual auditing model (44) is used for data acquisition of the analysis module (43) so as to obtain a high similarity change area which appears in infrared rays, the manual auditing model (44) pushes the high similarity change area to a terminal of a supervisor, and the supervisor assists judgment of red lines through data of the data acquisition module (1), the construction scheme docking module (2) and the construction model construction module (3).
9. The intelligent monitoring system for construction sites based on unmanned aerial vehicle images according to claim 5, wherein the monitoring platform (4) further comprises a construction site monitoring module (45), the construction site monitoring module (45) is used for acquiring data of the inspection module (41), and the construction site monitoring module (45) is used for identifying and classifying and counting mechanical equipment and personnel on the construction site according to the construction image data.
10. The intelligent monitoring system for construction sites based on unmanned aerial vehicle images according to claim 9, wherein the construction site monitoring module (45) comprises a mechanical equipment identification sub-module (451), a personnel identification sub-module (452) and a classification statistics sub-module (453), The mechanical equipment identification sub-module (451) analyzes the construction image data by adopting a target detection frame, the mechanical equipment identification sub-module (451) positions a mechanical position in a first stage, and identifies and distinguishes bulldozers and similar diggers through a local characteristic strengthening network in a second stage so as to obtain a mechanical equipment identification result, and the mechanical equipment identification sub-module (451) sends the mechanical equipment identification result to a coordinate position corresponding to the construction model; The personnel identification sub-module (452) is used for extracting a cap color gamut histogram in an HSV color space after carrying out target detection on the construction image data, setting a red cap and yellow cap threshold interval, and reducing the misjudgment rate through an illumination compensation algorithm so as to obtain a personnel identification result, and the personnel identification sub-module (452) is used for sending the personnel identification result to a coordinate position corresponding to the construction model; The classification statistics sub-module (453) is used for acquiring data of the mechanical equipment identification sub-module (451), the personnel identification sub-module (452) and the construction model construction module (3), the classification statistics sub-module (453) is used for comparing and analyzing the mechanical equipment identification result and the personnel identification result with the construction area data, the construction project data and the construction progress data, the classification statistics sub-module (453) is used for setting corresponding mechanical equipment threshold values and personnel threshold values in different construction areas, and when the data of the mechanical equipment identification result and the personnel identification result in the construction area are larger than the corresponding mechanical equipment threshold values and personnel threshold values, the classification statistics sub-module (453) is used for triggering alarming and pushing responsible persons, and the classification statistics sub-module (453) is used for generating reports comprising classification statistics tables and abnormal event screenshots.

Description

Intelligent construction land monitoring system based on unmanned aerial vehicle images Technical Field The invention relates to the technical field of construction monitoring, in particular to an intelligent construction land monitoring system based on unmanned aerial vehicle aerial images. Background Based on the improvement of requirements on ecological environment protection, the restriction of red line rigidity of farmland protection, innovative practice and policy response of local government, in recent years, a series of policy files for strict land management are brought out from the national level, and the approval of construction lands is promoted to be increasingly standard and strict, so that the approval of the lands is increasingly standard and strict. However, due to the limitation of complicated terrains, the mountain area high-speed projects are often required to perform operations such as excavation, slope releasing and the like in order to build an operation platform and pull through a longitudinal channel, and the phenomena of super-red lines and over-range land are objectively easy to cause. Subjectively, temporary lands such as a soil taking field, a spoil field, a concrete mixing station, a reinforcing steel bar processing field, a project site, a construction passageway and the like have the phenomena of little batch use, no batch use, and east and west batch use. The prior projects lack effective self-checking means, and the problems can not be found and rectified in time, so that the problems can be passively dealt with for law enforcement, and higher administrative penalties and even legal risks are faced. Moreover, when illegal land conditions occur, illegal projects can be called to stop and stop, the material and engineering quality of illegal buildings in the construction process are difficult to guarantee, and the life and property safety ‌ of people is seriously threatened. Disclosure of Invention In order to solve the problems, the invention provides an intelligent monitoring system for construction land based on unmanned aerial vehicle images, which solves the problems of low inspection efficiency, lag finding due to violations, complicated evidence obtaining process and other pain points in the traditional construction supervision mode and comprehensively improves the real-time performance, accuracy and traceability of land supervision. In order to achieve the above purpose, the technical scheme adopted by the invention is as follows: an intelligent monitoring system for construction land based on unmanned aerial vehicle images comprises a data acquisition module, a construction scheme docking module, a construction model construction module and a monitoring platform, The data acquisition module is used for acquiring an unmanned aerial vehicle aerial image so as to acquire image data of a construction area; the construction scheme docking module is used for importing construction scheme information so as to obtain construction area data, construction project data and construction progress data; The construction model construction module is used for acquiring data of the data acquisition module and the construction scheme docking module so as to construct a reference model according to image data before project construction; The monitoring platform is used for acquiring data of the data acquisition module, the construction scheme docking module and the construction model building module, the monitoring platform regularly shoots the construction area through an unmanned aerial vehicle in an aerial way to obtain construction image data, the monitoring platform updates the reference model according to the construction image data to generate a construction model in the construction model building module, and the monitoring platform analyzes the construction model according to the construction area data, the construction project data and the construction progress data to dynamically monitor the construction area. 2. The intelligent monitoring system for construction sites based on unmanned aerial vehicle images according to claim 1, wherein the construction model building module comprises a data preprocessing sub-module, a labeling sub-module, a data fusion sub-module and a visualization processing sub-module The data preprocessing sub-module is used for acquiring the data of the data acquisition module and the construction scheme docking module so as to perform illumination equalization processing on the original image and POS positioning data in the image data, so as to enhance the texture characteristics of the key target; the data preprocessing sub-module is used for positioning the image data and constructing a spatial index table through red line data in the construction area data so as to match the image data with the red line data; the marking sub-module marks the typical temporary land type of the image data in the data preprocessing sub-module through a marking model