A river three-dimensional surveying and mapping method and system based on a UAV
By acquiring tilted images of the river channel using drones, constructing a sequence of attitude changes, and correcting the attitude angles in segments, a highly consistent point cloud is generated. This solves the problems of error introduction and insufficient consistency of results in traditional river channel mapping methods, and achieves high-precision 3D mapping results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING IWHR BIC
- Filing Date
- 2026-05-28
- Publication Date
- 2026-07-14
Smart Images
Figure CN122384752A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing technology, and in particular to a method and system for three-dimensional river mapping based on unmanned aerial vehicles (UAVs). Background Technology
[0002] The field of data processing technology involves the acquisition, storage, computation, and representation of multi-source data, covering core aspects such as spatial data acquisition, data preprocessing, coordinate transformation, feature extraction, and 3D reconstruction. Its methods include acquiring raw data through sensing devices, using ground control points for spatial positioning, combining image matching and point cloud generation for structured data processing, and achieving unified representation and management of geographic information through a spatial coordinate system, thereby forming a spatial data model that can be used for analysis and application. Traditional 3D river mapping methods, specifically addressing the acquisition of river topography and boundary information, employ total station measurements to obtain discrete point coordinates, combine this with leveling instruments for elevation determination, utilize manually established control points for spatial positioning, and organize the data through manual recording and post-processing. Then, based on the measured cross-section data, river boundary lines are drawn and a simple 3D representation is created. Simultaneously, manned aerial imagery can be combined, and the river's extent can be extracted and 2D graphics drawn through manual interpretation.
[0003] Traditional river mapping relies on total stations and levels to acquire discrete points and manually set up control points for spatial positioning. The data acquisition process is in the form of discrete sampling, which makes it difficult to cover the continuous changes in the river channel. Furthermore, manual recording and post-processing methods are prone to human error and recording deviation. The intervals between measurement sections lead to gaps in spatial representation. For example, relying on only a limited number of cross-sections in a meandering river section will cause distortion of the boundary morphology. When using machine-generated images combined with manual interpretation to extract the range, it is greatly affected by subjective experience. Different operators have significantly different understandings of the shoreline and water surface boundaries, resulting in insufficient consistency of results. Two-dimensional drawing methods lack the ability to continuously represent three-dimensional structures and elevation changes, and it is difficult to accurately reflect structural details in complex areas such as bridges or beaches. At the same time, the data processing process lacks a unified attitude constraint and a systematic error allocation mechanism. Local measurement errors gradually accumulate in the overall representation, affecting the accuracy of the river channel model and the reliability of subsequent analysis. Summary of the Invention
[0004] To achieve the above objectives, the present invention adopts the following technical solution: a method for three-dimensional river channel mapping based on unmanned aerial vehicles (UAVs), comprising the following steps: S1: Acquire UAV tilt images within the river channel range, collect the roll angle, pitch angle, heading angle and exposure time corresponding to each image, sort the exposure times and adjust the image number synchronously, verify the continuity of time intervals between adjacent images, and obtain a continuous aerial survey image sequence. S2: Based on the continuous aerial survey image sequence, extract the attitude angle parameters of adjacent frames of the corresponding images of the bank slope area and water surface area, calculate the change amplitude of roll angle, pitch angle and heading angle of adjacent images, summarize them to form a set of attitude change amplitude values, and obtain the image attitude change amplitude sequence. S3: Extract the image change amplitude corresponding to the river bend and straight river section according to the image attitude change amplitude sequence, compare the change amplitude with the set attitude angle change threshold, perform continuous grouping and segmentation on the images that meet the threshold, and obtain stable aerial survey image segments. S4: Extract the image attitude angle records corresponding to the bridge area and beach area for the stable aerial survey image segment, select the initial image attitude angle of the segment as the reference value, accumulate the image attitude angle changes in the segment to form an attitude sequence, calculate the deviation of the attitude sequence from the original attitude angle and distribute it in the segment, update the image attitude angle parameters, and obtain unified image attitude parameters. S5: Based on the unified image pose parameters, update the exterior orientation elements of the river channel cross-section area and shoreline node positions, extract the pixel coordinates of the corresponding points, and calculate the spatial point coordinates by combining the camera focal length and principal point coordinates, generate point clouds, and perform surface construction and orthophoto projection processing to obtain the three-dimensional mapping results of the river channel.
[0005] As a further aspect of the present invention, the continuous aerial survey image sequence includes a temporal index, inter-frame connection markers, and breakpoint check markers; the image attitude change amplitude sequence includes single-frame change records, change trend curves, and amplitude statistical labels; the stable aerial survey image segment includes segment start and end boundaries, segment length levels, and segment stability markers; the unified image attitude parameters include a reference attitude value, cumulative compensation amount, and unified correction value; and the three-dimensional river mapping results include a three-dimensional point set model, a river surface model, and an orthophoto map.
[0006] As a further aspect of the present invention, the process of adjusting the exposure time and image number simultaneously includes rearranging the UAV tilted images in ascending order of exposure time and generating consecutive integer numbers in the rearranged order.
[0007] As a further aspect of the present invention, the process of calculating the change amplitude of the roll angle, pitch angle and heading angle of adjacent images includes obtaining the absolute difference of the roll angle, pitch angle and heading angle of adjacent images respectively, and forming the corresponding change amplitude according to the image sequence.
[0008] As a further aspect of the present invention, the specific steps of S1 are as follows: S101: Acquire UAV tilt images within the river channel range, collect roll angle, pitch angle, heading angle and exposure time information for each image, extract the exposure time field for the image number set and perform serialization processing, match the attitude angle parameters with the time series according to the number index to form a unified parameter mapping relationship, and generate an image attitude time association table. S102: Based on the image pose time association table, perform sequential arrangement operation according to the exposure time sequence, and map the arrangement result to the image number set to complete the number reorganization. Perform calculation on adjacent time intervals and make consistency judgment with the time interval threshold, retain records that meet the continuous condition, and obtain the time sequence verification image number sequence. S103: Call the time-series verification image number sequence, reconstruct the image arrangement structure according to the number order, synchronously match the corresponding attitude angle parameters to the image position, perform continuity determination on the time interval of adjacent images and establish sequential correlation to obtain a continuous aerial survey image sequence.
[0009] As a further aspect of the present invention, the specific steps of S2 are as follows: S201: Based on the continuous aerial survey image sequence, extract the attitude angle parameters of adjacent frames of the corresponding images of the bank slope area and the water surface area, perform adjacent relationship matching for the image frame number sequence, combine the roll angle, pitch angle and heading angle in pairs according to the frame pairs, and classify and label them according to the area identifier to form a pair data structure of attitude angles, and generate a set of attitude angle combinations of adjacent frames of the area. S202: Based on the set of attitude angle combinations of adjacent frames in the region, perform change amplitude calculation on each set of attitude angle parameters, perform difference calculation on the roll angle, pitch angle and yaw angle between adjacent frames respectively, and perform structured encoding processing on the angle change results to form a unified amplitude representation sequence, and obtain the attitude angle change amplitude dataset; S203: Call the attitude angle change amplitude dataset, perform sequential summary processing on each amplitude value according to the image frame order, integrate the roll angle, pitch angle and yaw angle change amplitudes according to the frame sequence, and construct a continuous arrangement structure to obtain the image attitude change amplitude sequence.
[0010] As a further aspect of the present invention, the specific steps of S3 are as follows: S301: Extract the image change amplitude corresponding to the river bend and straight river section according to the image attitude change amplitude sequence, perform segmented index matching on the amplitude sequence according to the river section identifier, classify and collect the corresponding amplitude of the bend and straight river section images, establish the segment correspondence relationship according to the image order, and generate the river section classification attitude amplitude set. S302: Based on the river section classification attitude amplitude set, the image change amplitude is compared with the set attitude angle change threshold. According to the relationship between the amplitude and the threshold, the condition judgment is performed. Images that meet the threshold condition are marked and a set of filtering results is formed. The threshold filtering image index set is obtained. S303: Call the threshold to filter the image index set, perform continuity determination according to the image number order, group and aggregate images that meet the continuity condition, and establish a segment division structure based on the group boundary to obtain stable aerial survey image segments.
[0011] As a further aspect of the present invention, the specific steps of S4 are as follows: S401: Extract the corresponding image attitude angle records of the bridge area and the beach area for the stable aerial survey image segment, establish a regional index relationship based on the image number, classify and collect the attitude angle parameters by region, arrange them in the segment order, identify and bind the attitude angle parameters of the segment's starting image, and generate a segment reference attitude angle set. S402: Based on the segment reference attitude angle set, perform cumulative change calculation on the image attitude angle parameters within the segment, calculate the difference between the image attitude angle and the reference attitude angle, and organize them sequentially according to the image order to form a continuous attitude change representation structure and obtain the cumulative attitude change sequence. S403: Call the accumulated sequence of attitude changes, calculate the deviation between the sequence data and the original attitude angle record, perform distribution processing within the segment based on the deviation result and update the image attitude angle parameters, complete the attitude angle parameter reconstruction, and obtain the unified image attitude parameters.
[0012] As a further aspect of the present invention, the specific steps of S5 are as follows: S501: Based on the unified image attitude parameters, update the exterior orientation elements of the images corresponding to the river cross-section area and shoreline node positions, establish an exterior orientation element index relationship according to the image number and area identifier, perform matching update on the attitude parameters and exterior orientation parameters, and classify and integrate them according to the area type to generate a set of area exterior orientation element parameters. S502: Extract the pixel coordinates of the corresponding points according to the set of external orientation element parameters of the region, perform correlation calculation with the camera focal length and principal point coordinates, complete the image coordinate to spatial coordinate conversion process, and organize them in a structured manner according to spatial position to obtain a set of spatial point coordinates. S503: Call the set of spatial point coordinates to generate point cloud data, perform surface construction processing on the point cloud data and perform orthophoto transformation, organize and express the processing results according to the river channel area, and obtain the three-dimensional mapping results of the river channel.
[0013] A UAV-based three-dimensional river mapping system includes: The aerial survey acquisition module acquires UAV tilted images within the river channel area, collects the roll angle, pitch angle, heading angle and exposure time information corresponding to each image, arranges the exposure time information in sequence and adjusts the image numbering order synchronously, and performs continuity verification on the time interval between adjacent images to obtain a continuous aerial survey image sequence. The attitude analysis module extracts the attitude angle parameters of adjacent frames of the corresponding images of the bank slope area and the water surface area based on the continuous aerial survey image sequence, calculates the change amplitude of the roll angle, pitch angle and heading angle of adjacent images, and summarizes them to form a set of attitude change amplitudes, thus obtaining the image attitude change amplitude sequence. The segment division module extracts the image change amplitude corresponding to the river bend segment and the straight river segment based on the image attitude change amplitude sequence, compares the change amplitude with the set attitude angle change threshold, performs continuous grouping processing on images that meet the threshold condition and forms segment division to obtain stable aerial survey image segments. The attitude correction module extracts the corresponding image attitude angle records of the bridge area and the beach area for the stable aerial survey image segment, selects the initial image attitude angle of the segment as the reference value, accumulates the changes in image attitude angle within the segment and forms an attitude sequence, calculates the deviation between the attitude sequence and the original attitude angle, distributes the deviation results within the segment and updates the image attitude angle parameters to obtain unified image attitude parameters. The 3D modeling module updates the exterior orientation elements of the river channel cross-section area and shoreline node positions based on the unified image pose parameters, extracts the pixel coordinates of the corresponding points, and calculates the spatial point coordinates by combining the camera focal length and principal point coordinates, generates point cloud data, and performs surface construction and orthophoto projection processing to obtain the 3D mapping results of the river channel.
[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, by acquiring tilted images of UAVs and simultaneously collecting roll angle, pitch angle, heading angle and exposure time, a change sequence is constructed. Based on the cumulative deviation of the initial attitude angle of the section, attitude uniformity and error smoothing within the section are achieved. The exterior orientation elements are updated, and spatial points are calculated by combining the pixel coordinates of corresponding points, camera focal length and principal point coordinates to generate a highly consistent point cloud and complete surface construction and orthophoto projection. This improves the continuity, geometric consistency and detail restoration accuracy of the three-dimensional representation of the river channel. Attached Figure Description
[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0016] Figure 1This is a schematic diagram of the steps of the present invention; Figure 2 This is a detailed schematic diagram of S1 of the present invention; Figure 3 This is a detailed schematic diagram of S2 of the present invention; Figure 4 This is a detailed schematic diagram of S3 of the present invention; Figure 5 This is a detailed schematic diagram of S4 of the present invention; Figure 6 This is a detailed schematic diagram of S5 of the present invention; Figure 7 This is a system module diagram of the present invention. Detailed Implementation
[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.
[0018] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.
[0019] Please see Figure 1 This invention provides a method for three-dimensional river channel mapping based on unmanned aerial vehicles (UAVs), comprising the following steps: S1: Acquire UAV tilt images within the river channel area, collect the roll angle, pitch angle, heading angle and exposure time information corresponding to each image, arrange the exposure time information in sequence and adjust the image numbering order synchronously, and perform continuity verification on the time interval between adjacent images to obtain a continuous aerial survey image sequence. S2: Extract attitude angle parameters of adjacent frames of images corresponding to the bank slope area and water surface area based on the continuous aerial survey image sequence, calculate the change amplitude of roll angle, pitch angle and heading angle of adjacent images, and summarize them to form an attitude change amplitude set to obtain the image attitude change amplitude sequence. S3: Extract the image change amplitude corresponding to the river bend and straight river section based on the image attitude change amplitude sequence, compare the change amplitude with the set attitude angle change threshold, continuously group the images that meet the threshold condition and form the segment division to obtain stable aerial survey image segments. S4: Extract the image attitude angle records corresponding to the bridge area and the beach area for the stable aerial survey image segment. Select the initial image attitude angle of the segment as the reference value, accumulate the changes in image attitude angle within the segment and form an attitude sequence. Calculate the deviation between the attitude sequence and the original attitude angle, distribute the deviation results within the segment and update the image attitude angle parameters to obtain unified image attitude parameters. S5: Based on the unified parameters of image pose, update the exterior orientation elements of the river channel cross-section area and shoreline node positions, extract the pixel coordinates of the corresponding points, and calculate the spatial point coordinates by combining the camera focal length and principal point coordinates. Generate point cloud data and perform surface construction and orthophoto projection processing to obtain the three-dimensional mapping results of the river channel.
[0020] The continuous aerial survey image sequence includes a time-series index, inter-frame connection markers, and breakpoint check marks; the image attitude change amplitude sequence includes single-frame change records, change trend curves, and amplitude statistical labels; the stable aerial survey image segment includes the segment start and end boundaries, segment length level, and segment stability markers; the unified image attitude parameters include the baseline attitude value, cumulative compensation amount, and unified correction value; the three-dimensional river mapping results include a three-dimensional point set model, a river surface model, and an orthophoto map.
[0021] Please see Figure 2 The specific steps of S1 are as follows: S101: Acquire UAV tilt images within the river channel range, collect roll angle, pitch angle, heading angle and exposure time information for each image, extract the exposure time field for the image number set and perform serialization processing, match the attitude angle parameters with the time series according to the number index to form a unified parameter mapping relationship, and generate an image attitude time association table. A drone equipped with a three-axis stabilized gimbal flies along a predetermined route, continuously acquiring images at an altitude of 120 meters and a speed of 8 meters per second. The image format is set to JPEG files with a resolution of 6000 x 4000 pixels. Simultaneously, the onboard inertial measurement unit records roll angle, pitch angle, yaw angle, and exposure time information with millisecond-level time recording accuracy. A traversal reading operation is performed on the image set, extracting the exposure time field corresponding to each image and serializing it according to chronological order. For example, selecting five consecutive images with exposure times of 10.125 seconds, 10.230 seconds, 10.335 seconds, 10.440 seconds, and 10.545 seconds, respectively, and forming an increasing time sequence through sorting operations. Subsequently, the corresponding attitude angle parameters are matched one-to-one with the time series according to the image number. For example, image number 1 corresponds to a roll angle of 2.3 degrees, a pitch angle of -1.2 degrees, and a yaw angle of 85.6 degrees, while image number 2 corresponds to a roll angle of 2.5 degrees, a pitch angle of -1.1 degrees, and a yaw angle of 86.1 degrees. A unique mapping relationship is established through the number index. To verify the consistency of the time records, a time error threshold of 0.02 seconds is set. By comparing the differences between adjacent time values, a valid sequence is determined when the time difference is within the range of 0.10 seconds to 0.12 seconds. The image number, attitude angle parameters, and time series are integrated to form an image attitude time association table, where the number of records is consistent with the total number of images. For example, if 200 images are acquired, 200 association records are generated.
[0022] S102: Based on the image pose time association table, perform sequential arrangement operations according to the exposure time sequence, and map the arrangement results to the image number set to complete the number reorganization. Perform calculations on adjacent time intervals and make consistency judgments with the time interval threshold, retain records that meet the continuous condition, and obtain the time sequence verification image number sequence. All exposure times are sorted in ascending order, and the sorted results are mapped back to the image number set to reorganize the numbers. For example, the original numbering order is 5, 2, 3, 1, 4, and after sorting, the corresponding exposure time order is adjusted to 1, 2, 3, 4, 5. Then, adjacent time intervals are calculated one by one, subtracting the previous image's exposure time from the subsequent image's exposure time to obtain the time difference. For example, the time differences are 0.105 seconds, 0.110 seconds, 0.108 seconds, and 0.107 seconds. A time interval threshold range is set from 0.08 seconds to 0.15 seconds. Each time difference is compared to determine if it falls within this range. When all time differences meet the condition, the record is considered continuous. If a time difference of 0.25 seconds is found, it is considered an abnormal interval and the record is removed. A filtering operation is then performed to retain all image numbers that meet the continuity condition, forming a time-series verification image number sequence. For example, after filtering, the number sequence is 1, 2, 3, 4. By comparing experimental data, under the same flight path conditions, when the time interval is stable at around 0.10 seconds, the sequence continuity rate reaches 98.5%, while it is only 87.2% when no screening is performed. This screening logic improves sequence stability by using the time difference to make a judgment.
[0023] S103: Call the time sequence verification image number sequence, reconstruct the image arrangement structure according to the number order, synchronize the corresponding attitude angle parameters to the image position, perform continuity judgment on the time interval of adjacent images and establish sequential correlation to obtain a continuous aerial survey image sequence. The image data is reorganized according to a new numbering order; for example, images numbered 1, 2, 3, and 4 are arranged chronologically. Then, the corresponding attitude angle parameters are synchronously matched to the rearranged image positions, and roll, pitch, and yaw angle parameters are filled into the corresponding images one by one using the numbering index. A temporal continuity determination operation is performed on adjacent images, comparing the adjacent time difference with a continuity threshold interval. For example, if the continuity threshold is set to 0.12 seconds, adjacent differences of 0.108 seconds and 0.110 seconds are considered continuous. Further, a sequential association relationship is established, forming a chain connection structure between each image and its preceding and following images; for example, image 1 is associated with image 2, and image 2 is associated with image 3. This association structure forms a continuous aerial survey image sequence, and the sequence length is recorded; for example, a continuous sequence length of 180 images, accounting for 90% of the original 200 images. By comparing different threshold settings, when the threshold is increased to 0.20 seconds, the continuous sequence length increases to 195 images, but subsequent stability decreases by about 6%, indicating that using 0.12 seconds as the determination interval is more reasonable.
[0024] Please see Figure 3 The specific steps of S2 are as follows: S201: Extract attitude angle parameters of adjacent frames of images corresponding to the bank slope area and water surface area based on continuous aerial survey image sequence, generate area identifiers of bank slope area and water surface area according to the spatial location of the river channel corresponding to the image, perform adjacent relationship matching for image frame number sequence, combine roll angle, pitch angle and heading angle in pairs according to frame pair, and classify and label according to area identifier to form attitude angle pair data structure, and generate a set of attitude angle combination sets of adjacent frames of the area; The imagery is divided into two categories based on geolabelling data: bank slope markers and water surface markers. For example, image numbers 1 to 90 are labeled as bank slope areas, and numbers 91 to 180 are labeled as water surface areas. Adjacency matching is performed on the image frame number sequence, grouping adjacent frames together, such as frame 1 with frame 2, and frame 2 with frame 3. Roll, pitch, and yaw angles are extracted from each group and then paired for combination; for example, a roll angle of 2.3 degrees from frame 1 and a roll angle of 2.5 degrees from frame 2 form a data pair. These data pairs are then categorized and labeled according to their region identifiers: bank slope areas are labeled as category 1, and water surface areas as category 2. This process creates a pairwise attitude angle data structure, resulting in 179 data pairs, which are then stored as a set of attitude angle combinations for adjacent frames within each region. Actual data analysis shows that there are 89 combinations for bank slope areas and 89 combinations for water surface areas.
[0025] S202: Based on the attitude angle combination set of adjacent frames in the region, perform change amplitude calculation on each group of attitude angle parameters, perform difference calculation on roll angle, pitch angle and yaw angle between adjacent frames respectively, and perform structured encoding processing on the angle change results to form a unified amplitude representation sequence to obtain the attitude angle change amplitude dataset; The change amplitude is calculated based on the combination set of attitude angles from adjacent frames in the region. For each set of data, a difference operation is performed, subtracting the angle of the previous frame from the angle of the subsequent frame to obtain the change amplitude. For example, a change in roll angle from 2.3 degrees to 2.5 degrees results in a change amplitude of 0.2 degrees. The same calculation is performed sequentially for pitch and yaw angles, for example, a change in pitch angle of 0.1 degrees and a change in yaw angle of 0.5 degrees. Each set of results is organized into a unified structure, for example, recorded as a change in roll of 0.2 degrees, a change in pitch of 0.1 degrees, and a change in yaw of 0.5 degrees. For uniform representation, the three types of angle changes are encoded as triplet data in sequence. Amplitude threshold ranges are set for abnormal changes: 0 to 1.5 degrees for roll angle, 0 to 1.0 degrees for pitch angle, and 0 to 3.0 degrees for yaw angle. Comparison is used to determine whether the change is within the range. If the yaw change reaches 4.5 degrees, it is marked as abnormal and recorded. Experimental statistics show that under normal aerial survey conditions, the proportion of changes with amplitudes below the above thresholds reaches 95.3%.
[0026] S203: Call the attitude angle change amplitude dataset, perform sequential summary processing on each amplitude value according to the image frame order, integrate the roll angle, pitch angle and yaw angle change amplitudes according to the frame sequence, and construct a continuous arrangement structure to obtain the image attitude change amplitude sequence. The image frames are processed sequentially, and the three types of change amplitudes corresponding to each frame are arranged in sequence to form a structure. For example, the first frame corresponds to changes of 0.2 degrees, 0.1 degrees, and 0.5 degrees, and the second frame corresponds to changes of 0.3 degrees, 0.1 degrees, and 0.4 degrees. All frame data are integrated in numerical order to form a continuous arrangement, resulting in a sequence of image attitude change amplitudes. To verify the stability of the sequence, every 10 frames are used as a statistical unit to calculate the average change value. For example, the average change for the first 10 frames is 0.25 degrees for roll, 0.12 degrees for pitch, and 0.48 degrees for heading, all within the normal range. Comparing data from different regions, the average change in the bank slope area is about 0.08 degrees higher than that in the water surface area. This result is obtained by using the change amplitude in the statistics, providing basic data for subsequent segmentation processing.
[0027] Please see Figure 4 The specific steps of S3 are as follows: S301: Extract the image change amplitude corresponding to the river bend and straight river sections based on the image attitude change amplitude sequence, perform segmented index matching on the amplitude sequence according to the river section identifier, classify and collect the corresponding amplitudes of the bend and straight river sections, establish the segment correspondence relationship according to the image order, and generate a river section classification attitude amplitude set. By accessing the path coordinate data in the route planning file, the turning angle changes between adjacent waypoints are calculated point by point. The turning angle formed by three consecutive waypoints is used as the criterion. For example, the turning angle between waypoints 1, 2, and 3 is 3.2 degrees, which is converted to a curvature value of 0.028. A curvature value greater than 0.02 is marked as a curved segment, and a curvature value less than or equal to 0.02 is marked as a straight segment. Subsequently, image numbers are segmented and indexed based on this identifier. For example, if the curvature values corresponding to numbers 30 to 80 are all greater than 0.02, they are classified as curved segments. The attitude change amplitude data of the corresponding images are extracted and categorized sequentially according to their numbers. For example, in a curved segment, the roll changes of five consecutive images are 0.32 degrees, 0.35 degrees, 0.37 degrees, 0.34 degrees, and 0.36 degrees; the pitch changes are 0.18 degrees, 0.20 degrees, 0.19 degrees, 0.17 degrees, and 0.18 degrees; and the heading changes are 0.82 degrees, 0.88 degrees, 0.91 degrees, 0.86 degrees, and 0.89 degrees. The same extraction operation is performed on straight segments. For example, the roll changes in images numbered 1 to 29 are concentrated in the range of 0.12 degrees to 0.18 degrees. By recording each data point, a classification set is formed, and the start and end relationships of segments are established according to the image numbering order. For example, the curved segment starts at number 30 and ends at number 80. To verify the accuracy of the segmentation, three sets of data from different river sections were selected for comparison. The average roll change of the curved section was 0.35 degrees, while that of the straight section was 0.15 degrees, with a difference of 0.20 degrees. This indicates that the classification results have stable distinguishing characteristics. The segmentation process achieves the accuracy of segment division by using curvature in the determination.
[0028] S302: Based on the river segment classification attitude amplitude set, compare the image change amplitude with the set attitude angle change threshold, perform condition judgment according to the relationship between amplitude and threshold, mark the images that meet the threshold conditions and form a set of filtering results, and obtain the threshold filtering image index set. The roll, pitch, and yaw angle changes of each image are compared with preset attitude angle change thresholds. The roll angle threshold is calculated by statistically analyzing 200 sets of historical aerial survey data, yielding an average of 0.28 degrees and a standard fluctuation range of 0.05 degrees. The average plus one times the fluctuation range is taken as the upper limit threshold, i.e., 0.33 degrees. The pitch angle threshold is set to 0.20 degrees, based on an average of 0.15 degrees plus a fluctuation of 0.05 degrees. The yaw angle threshold is set to 1.00 degrees, based on an average of 0.80 degrees plus a fluctuation of 0.20 degrees. For each image, these three values are compared. If the roll, pitch, and yaw changes are all less than their respective thresholds, the image is marked as meeting the criteria. For example, if an image has a roll of 0.30 degrees, a pitch of 0.18 degrees, and a yaw of 0.95 degrees, all within the thresholds, it is marked as valid. If an image has a roll of 0.36 degrees, exceeding 0.33 degrees, it is directly determined as invalid. All eligible image numbers were compiled into a selection set. For example, in the curved section, there were 51 images, of which 34 met the criteria, resulting in a selection rate of 66.7%. Through experimental comparison of different threshold combinations, when the roll threshold was increased to 0.40 degrees, the number of effective images increased to 41, but the subsequent stability decreased by about 12%. This indicates that the current threshold setting of 0.33 degrees has a better balance. This comparison process uses multiple sets of data to calculate and achieve a reasonable selection of the threshold.
[0029] S303: Call the threshold to filter the image index set, perform continuity determination according to the image number order, group and aggregate images that meet the continuity condition, and establish a segment division structure based on the group boundary to obtain stable aerial survey image segments. The selected image numbers are sorted in ascending order, and continuity is determined for each image. Images with a difference of 1 between adjacent numbers are considered a continuous sequence; for example, images 45 and 46 are considered continuous due to a difference of 1, while images 46 and 48 are considered interrupted due to a difference of 2. Continuous numbers are then aggregated, grouping consecutive number segments into the same group; for example, images 45 to 60 are grouped together. The start and end numbers of each group are recorded; for example, the first group is 45 to 60, and the second group is 62 to 70. Each group is considered a segment boundary, and a segment division structure is established based on these boundaries. The length of each segment is further calculated; for example, segment 1 has 16 images, segment 2 has 9 images, and segment 3 has 14 images. A minimum effective segment length of 10 images is set, and segments with a length of 10 or more images are retained through comparison; for example, segment 2 is discarded. Stable aerial survey image segments are obtained; for example, four segments are retained. By comparing experimental data, the average attitude fluctuation was 0.42 degrees when no grouping was performed, and it decreased to 0.21 degrees after grouping, a decrease of 50%. This continuity determination improves the stability of the section by using the difference in the numbering.
[0030] Please see Figure 5 The specific steps of S4 are as follows: S401: Extract the attitude angle records of the bridge area and the beach area corresponding to the stable aerial survey image segment, establish the regional index relationship according to the image number, classify and collect the attitude angle parameters by region, arrange them in the segment order, identify and bind the attitude angle parameters of the segment's starting image, and generate the segment reference attitude angle set. By reading the river's geographic vector data, the bridge location was defined as an area 30 meters before and after the bridge's centerline, and the beach area as a range extending 50 meters inward from the water's edge. The geographic coordinates of the images were used to determine their respective areas; for example, images numbered 60 to 75 fell within the bridge area, and images numbered 76 to 95 fell within the beach area. A regional index relationship was established based on the image numbers, and the corresponding attitude angle parameters were aggregated into the corresponding regional sets. The image data were arranged in segment order; for example, in segment 1, the bridge area images were arranged according to numbers 60 to 75. The attitude angles of the starting image of the segment were extracted as reference values; for example, image number 60 had a roll of 2.1 degrees, a pitch of -1.0 degrees, and a heading of 84.8 degrees. This set of data was identified as the segment's reference attitude angles and bound to the segment number to form a reference dataset.
[0031] S402: Based on the segment reference attitude angle set, perform cumulative change calculation on the image attitude angle parameters within the segment, calculate the difference between the image attitude angle and the reference attitude angle, and organize them sequentially according to the image order to form a continuous attitude change representation structure and obtain the cumulative attitude change sequence. The current image attitude angle is compared with the reference attitude angle item by item. For example, the difference between the roll of image number 65 (2.5 degrees) and the reference value (2.1 degrees) is 0.4 degrees; the difference between the pitch (-0.8 degrees) and the reference value (-1.0 degrees) is 0.2 degrees; and the difference between the heading (85.6 degrees) and the reference value (84.8 degrees) is 0.8 degrees. This calculation is repeated for all images within the segment and arranged in numerical order to form a continuous change sequence. Statistical analysis is performed on the sequence, controlling the cumulative change range within the roll (0 to 0.8 degrees), pitch (0 to 0.5 degrees), and heading (0 to 1.2 degrees) ranges. When the change of an image exceeds this range, it is marked as abnormal; for example, if the heading change of an image reaches 1.5 degrees, it is recorded as abnormal. Statistical analysis of 20 images shows that the normal change rate is 90%. This cumulative calculation quantifies the change trend by using the reference attitude in the difference calculation.
[0032] S403: Call the attitude change cumulative sequence, calculate the deviation between the sequence data and the original attitude angle record, distribute the data within the segment according to the deviation result and update the image attitude angle parameters, complete the attitude angle parameter reconstruction, and obtain the unified image attitude parameters. The deviation is obtained by subtracting the original change value from the cumulative change value. For example, if the cumulative roll change of an image is 0.4 degrees and the original record is 0.55 degrees, the deviation is 0.15 degrees. The original attitude angles are then corrected based on the deviation value, proportionally allocating the deviation to the attitude angle parameters. For example, the roll angle is adjusted from 2.5 degrees to 2.35 degrees. The same operation is performed on all images within the segment to update the attitude angle parameters. The change in error before and after the update is statistically analyzed. The original standard deviation of the attitude angles was 0.36 degrees, which decreased to 0.18 degrees after the update, a reduction of 50%. This deviation calculation achieves unified attitude processing by using the cumulative change and the original data in the calculation.
[0033] Please see Figure 6 The specific steps of S5 are as follows: S501: Based on the unified image attitude parameters, update the exterior orientation elements of the river cross section area and shoreline node positions. Establish an exterior orientation element index relationship according to the image number and area identifier. Perform matching update on the attitude parameters and exterior orientation parameters. Classify and integrate according to area type to generate a set of area exterior orientation element parameters. First, the original exterior orientation element records of the images are read. This data comes from the exterior orientation parameter file calculated after the UAV flight. The data format is one record per image, containing spatial coordinates and attitude angle information. The attitude uniform parameters are matched against the exterior orientation element records one by one according to the image number. For example, if the original heading angle of image number 70 is 85.9 degrees, and after parameter uniformization it becomes 85.3 degrees, a replacement operation is performed to update 85.9 degrees to 85.3 degrees, while retaining its spatial coordinates. Spatial matching is performed based on the image geographic coordinates and river cross-section location data to divide the images into cross-section regions or shoreline node regions. For example, through coordinate comparison, if the image center point is less than 10 meters from the cross-section line, it is marked as a cross-section region; if it is greater than 10 meters and within a 50-meter shoreline buffer zone, it is marked as a shoreline node region. An index relationship is established for each classified image, and the updated exterior orientation elements are aggregated in numerical order. For example, the cross-section region contains 120 images, and the shoreline node region contains 80 images. A consistency check is performed on the images within each region, ensuring that the difference in updated heading angles is within 0.2 degrees. For example, for consecutive images numbered 70 and 71, their updated heading angles are 85.3 degrees and 85.4 degrees respectively, with a difference of 0.1 degrees, thus they are considered continuous and consistent. Through this update and classification integration operation, a set of exterior orientation element parameters for the region is generated, providing a unified data foundation for subsequent spatial coordinate calculations.
[0034] S502: Extract the pixel coordinates of corresponding points based on the set of external orientation element parameters of the region, correlate the pixel coordinates with the camera focal length and principal point coordinates to complete the transformation of image coordinates to spatial coordinates, and organize them in a structured manner according to spatial location to obtain a set of spatial point coordinates. By combining feature point matching and manual verification, the pixel positions of the same ground feature in multiple images are obtained. First, automatic matching is performed using image grayscale features, filtering feature points with a grayscale gradient greater than 30 as candidate points. Then, neighbor comparison is used to filter point pairs with a matching error of less than 2 pixels. For example, the pixel coordinates of a certain control point in three images are 1205, 2310, 1212, 2302, and 1198, 2320. These pixel coordinates are then associated with camera parameters (camera focal length 35 mm, principal point coordinates 3000, 2000). By calculating the pixel offset, the lateral and longitudinal offsets are converted into spatial direction ratios. Then, spatial projection calculation is performed using the spatial position coordinates from the exterior orientation elements to obtain the three-dimensional coordinates of the control point in space. For example, the calculated spatial lateral coordinates are 125.6 meters, longitudinal coordinates are 210.9 meters, and elevation is 32.6 meters. The above process is repeated for multiple points with the same name, and the results are organized according to point number to form a set of spatial point coordinates. To verify the calculation accuracy, the same control point was calculated three times, with the spatial coordinate deviation controlled within 0.05 meters. Further analysis of data from 50 control points showed an average spatial error of 0.18 meters, meeting the accuracy requirements for river channel mapping. This processing method combines pixel coordinates with exterior orientation parameters to achieve the conversion from image coordinates to spatial coordinates, providing foundational data for subsequent point cloud generation. Table 1. Data Table for Calculating Spatial Point Coordinates
[0035] Table 1 shows the calculation results of some spatial point coordinates, and the spatial distribution of each point is continuous.
[0036] S503: Call the spatial point coordinate set to generate point cloud data, perform surface construction processing on the point cloud data and perform orthophoto transformation, organize and express the processing results according to the river area, and obtain the three-dimensional mapping results of the river. All spatial points are written into the point cloud data structure point by point according to their three-dimensional coordinates. First, the spatial points are sorted by number, and then written one by one in coordinate order. For example, spatial points numbered 1 to 50000 are stored sequentially to form a point cloud set. Next, point cloud density analysis is performed, and the number of points per square meter is counted as the density value. For example, if the density of a certain area is 150 points per square meter, a density threshold of 120 points per square meter is set. Areas below the threshold are filled by interpolation, increasing the density to at least the threshold. The density is further increased based on the point cloud construction and interpolation. Next, surface construction is performed. Connections are established between adjacent points with a spatial distance of less than 1.0 meter. Three adjacent points are combined to form triangular faces, and all points are traversed and connected to construct a complete triangular mesh structure. For example, 8000 triangular faces are generated in a certain local area. The constructed triangular mesh is then subjected to continuity verification, controlling the elevation difference between faces to within 0.5 meters. Faces exceeding this range are discarded or reconstructed. Subsequently, an orthophoto transformation was performed, projecting the 3D point cloud vertically onto a horizontal plane to generate 2D image data. This data was then organized into blocks according to river region; for example, the bridge area was output as one data block, and the beach area as another. By comparing experimental data, the processed point cloud density increased from the original 130 points per square meter to 190 points per square meter, and the spatial elevation error decreased from 0.45 meters to 0.28 meters. This result was achieved through the combined effect of spatial point coordinates and the point cloud construction process. Through point supplementation processing and subsequent point cloud construction and interpolation encryption processing, the point cloud density was further increased to 190 points per square meter.
[0037] Please see Figure 7 A UAV-based three-dimensional river mapping system includes: The aerial survey acquisition module acquires UAV tilted images within the river channel area, collects the roll angle, pitch angle, heading angle and exposure time information corresponding to each image, arranges the exposure time information in sequence and adjusts the image numbering order synchronously, and performs continuity verification on the time interval between adjacent images to obtain a continuous aerial survey image sequence. The attitude analysis module extracts the attitude angle parameters of adjacent frames of corresponding images of the bank slope area and water surface area based on the continuous aerial survey image sequence, calculates the change amplitude of roll angle, pitch angle and heading angle of adjacent images, and summarizes them to form a set of attitude change amplitude values, thus obtaining the image attitude change amplitude sequence. The segment division module extracts the image change amplitude corresponding to the river bend and straight river section based on the image attitude change amplitude sequence, compares the change amplitude with the set attitude angle change threshold, and continuously groups the images that meet the threshold condition to form segment division, thus obtaining stable aerial survey image segments. The attitude correction module extracts the corresponding image attitude angle records of the bridge area and the beach area for stable aerial survey image segments. It selects the initial image attitude angle of the segment as a reference value, accumulates the changes in image attitude angles within the segment and forms an attitude sequence. It calculates the deviation between the attitude sequence and the original attitude angle, distributes the deviation results within the segment and updates the image attitude angle parameters to obtain unified image attitude parameters. The 3D modeling module updates the exterior orientation elements of the river channel cross-section area and shoreline node positions based on the unified image pose parameters, extracts the pixel coordinates of the corresponding points, and calculates the spatial point coordinates by combining the camera focal length and principal point coordinates, generates point cloud data, and performs surface construction and orthophoto projection processing to obtain the 3D mapping results of the river channel.
[0038] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of protection of the described technical solutions.
Claims
1. A method for three-dimensional river channel mapping based on unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: S1: Acquire UAV tilt images within the river channel range, collect the roll angle, pitch angle, heading angle and exposure time corresponding to each image, sort the exposure times and adjust the image number synchronously, verify the continuity of time intervals between adjacent images, and obtain a continuous aerial survey image sequence. S2: Based on the attitude angle parameters of adjacent frames in the continuous aerial survey image sequence, calculate the change amplitude of roll angle, pitch angle and heading angle of adjacent images, summarize them to form a set of attitude change amplitudes, and obtain the image attitude change amplitude sequence. S3: Based on the corresponding image change amplitude in the image attitude change amplitude sequence, compare the change amplitude with a set attitude angle change threshold, continuously group and divide the images that meet the threshold into segments to obtain stable aerial survey image segments. S4: Extract the corresponding image attitude angle records for each segment of the stable aerial survey image segment, select the initial image attitude angle of the segment as a reference value, accumulate and calculate the changes in image attitude angles within the segment to form an attitude sequence, calculate the deviation between the attitude sequence and the original attitude angle and distribute it within the segment, update the image attitude angle parameters, and obtain unified image attitude parameters. S5: Based on the unified image attitude parameters, update the exterior orientation elements of the image within the stable aerial survey image segment, extract the pixel coordinates of the corresponding points, and calculate the spatial point coordinates by combining the camera focal length and principal point coordinates, generate a point cloud, and perform surface construction and orthophoto projection processing to obtain the three-dimensional mapping results of the river channel.
2. The method for three-dimensional river channel mapping based on unmanned aerial vehicles according to claim 1, characterized in that: The continuous aerial survey image sequence includes a time-series index, inter-frame connection markers, and breakpoint check marks; the image attitude change amplitude sequence includes single-frame change records, change trend curves, and amplitude statistical labels; the stable aerial survey image segment includes segment start and end boundaries, segment length levels, and segment stability markers; the unified image attitude parameters include reference attitude values, cumulative compensation amounts, and unified correction values; the three-dimensional river channel mapping results include a three-dimensional point set model, a river channel surface model, and an orthophoto map.
3. The method for three-dimensional river channel mapping based on unmanned aerial vehicles according to claim 1, characterized in that: The process of adjusting the exposure time and image number simultaneously includes rearranging the UAV tilted images in ascending order of exposure time and generating consecutive integer numbers in the rearranged order.
4. The method for three-dimensional river channel mapping based on unmanned aerial vehicles according to claim 1, characterized in that: The process of calculating the change amplitude of roll angle, pitch angle and heading angle of adjacent images includes calculating the absolute difference of roll angle, pitch angle and heading angle of adjacent images respectively, and forming the corresponding change amplitude according to the order of the image sequence.
5. The method for three-dimensional river channel mapping based on unmanned aerial vehicles according to claim 1, characterized in that, The specific steps of S1 are as follows: S101: Acquire UAV tilt images within the river channel range, collect roll angle, pitch angle, heading angle and exposure time information for each image, extract the exposure time field for the image number set and perform serialization processing, match the attitude angle parameters with the time series according to the number index to form a unified parameter mapping relationship, and generate an image attitude time association table. S102: Based on the image pose time association table, perform sequential arrangement operation according to the exposure time sequence, and map the arrangement result to the image number set to complete the number reorganization. Perform calculation on adjacent time intervals and make consistency judgment with the time interval threshold, retain records that meet the continuous condition, and obtain the time sequence verification image number sequence. S103: Call the time-series verification image number sequence, reconstruct the image arrangement structure according to the number order, synchronously match the corresponding attitude angle parameters to the image position, perform continuity determination on the time interval of adjacent images and establish sequential correlation to obtain a continuous aerial survey image sequence.
6. The method for three-dimensional river channel mapping based on unmanned aerial vehicles according to claim 1, characterized in that, The specific steps of S2 are as follows: S201: Based on the continuous aerial survey image sequence, extract the attitude angle parameters of adjacent frames of the corresponding images of the bank slope area and the water surface area, generate area identifiers of the bank slope area and the water surface area according to the spatial location of the river channel corresponding to the images, perform adjacent relationship matching for the image frame number sequence, combine the roll angle, pitch angle and heading angle in pairs according to frame pairs, and classify and label them according to the area identifiers to form a pairwise attitude angle data structure, and generate a set of attitude angle combinations of adjacent frames of the region; S202: Based on the set of attitude angle combinations of adjacent frames in the region, perform change amplitude calculation on each set of attitude angle parameters, perform difference calculation on the roll angle, pitch angle and yaw angle between adjacent frames respectively, and perform structured encoding processing on the angle change results to form a unified amplitude representation sequence, and obtain the attitude angle change amplitude dataset; S203: Call the attitude angle change amplitude dataset, perform sequential summarization processing on the amplitude values according to the image frame order, integrate the roll angle, pitch angle and yaw angle change amplitudes according to the frame sequence, and construct a continuous arrangement structure to obtain the image attitude change amplitude sequence.
7. The method for three-dimensional river channel mapping based on unmanned aerial vehicles according to claim 1, characterized in that, The specific steps for S3 are as follows: S301: Extract the image change amplitude corresponding to the river bend and straight river section according to the image attitude change amplitude sequence, perform segmented index matching on the amplitude sequence according to the river section identifier, classify and collect the corresponding amplitude of the bend and straight river section images, establish the segment correspondence relationship according to the image order, and generate the river section classification attitude amplitude set. S302: Based on the river section classification attitude amplitude set, the image change amplitude is compared with the set attitude angle change threshold. According to the relationship between the amplitude and the threshold, the condition judgment is performed. Images that meet the threshold condition are marked and a set of filtering results is formed. The threshold filtering image index set is obtained. S303: Call the threshold to filter the image index set, perform continuity determination according to the image number order, group and aggregate images that meet the continuity condition, and establish a segment division structure based on the group boundary to obtain stable aerial survey image segments.
8. The method for three-dimensional river channel mapping based on unmanned aerial vehicles according to claim 1, characterized in that, The specific steps of S4 are as follows: S401: Extract the corresponding image attitude angle records of the bridge area and the beach area for the stable aerial survey image segment, establish a regional index relationship based on the image number, classify and collect the attitude angle parameters by region, arrange them in the segment order, identify and bind the attitude angle parameters of the segment's starting image, and generate a segment reference attitude angle set. S402: Based on the segment reference attitude angle set, perform cumulative change calculation on the image attitude angle parameters within the segment, calculate the difference between the image attitude angle and the reference attitude angle, and organize them sequentially according to the image order to form a continuous attitude change representation structure and obtain the cumulative attitude change sequence. S403: Call the accumulated sequence of attitude changes, calculate the deviation between the sequence data and the original attitude angle record, perform distribution processing within the segment based on the deviation result and update the image attitude angle parameters, complete the attitude angle parameter reconstruction, and obtain the unified image attitude parameters.
9. The method for three-dimensional river channel mapping based on unmanned aerial vehicles according to claim 1, characterized in that, The specific steps of S5 are as follows: S501: Based on the unified image attitude parameters, update the exterior orientation elements of the images corresponding to the river cross-section area and shoreline node positions, establish an exterior orientation element index relationship according to the image number and area identifier, perform matching update on the attitude parameters and exterior orientation parameters, and classify and integrate them according to the area type to generate a set of area exterior orientation element parameters. S502: Extract the pixel coordinates of the corresponding points according to the set of external orientation element parameters of the region, perform correlation calculation with the camera focal length and principal point coordinates, complete the image coordinate to spatial coordinate conversion process, and organize them in a structured manner according to spatial position to obtain a set of spatial point coordinates. S503: Call the set of spatial point coordinates to generate point cloud data, perform surface construction processing on the point cloud data and perform orthophoto transformation, organize and express the processing results according to the river channel area, and obtain the three-dimensional mapping results of the river channel.
10. A three-dimensional river mapping system based on unmanned aerial vehicles (UAVs), characterized in that, The system is used to implement the UAV-based three-dimensional river mapping method according to any one of claims 1-9, and the system comprises: The aerial survey acquisition module acquires UAV tilted images within the river channel area, collects the roll angle, pitch angle, heading angle and exposure time information corresponding to each image, arranges the exposure time information in sequence and adjusts the image numbering order synchronously, and performs continuity verification on the time interval between adjacent images to obtain a continuous aerial survey image sequence. The attitude analysis module extracts the attitude angle parameters of adjacent frames of the corresponding images of the bank slope area and the water surface area based on the continuous aerial survey image sequence, calculates the change amplitude of the roll angle, pitch angle and heading angle of adjacent images, and summarizes them to form a set of attitude change amplitudes, thus obtaining the image attitude change amplitude sequence. The segment division module extracts the image change amplitude corresponding to the river bend segment and the straight river segment based on the image attitude change amplitude sequence, compares the change amplitude with the set attitude angle change threshold, performs continuous grouping processing on images that meet the threshold condition and forms segment division to obtain stable aerial survey image segments. The attitude correction module extracts the corresponding image attitude angle records of the bridge area and the beach area for the stable aerial survey image segment, selects the initial image attitude angle of the segment as the reference value, accumulates the changes in image attitude angle within the segment and forms an attitude sequence, calculates the deviation between the attitude sequence and the original attitude angle, distributes the deviation results within the segment and updates the image attitude angle parameters to obtain unified image attitude parameters. The 3D modeling module updates the exterior orientation elements of the river channel cross-section area and shoreline node positions based on the unified image pose parameters, extracts the pixel coordinates of the corresponding points, and calculates the spatial point coordinates by combining the camera focal length and principal point coordinates, generates point cloud data, and performs surface construction and orthophoto projection processing to obtain the 3D mapping results of the river channel.