CN107560593B - Special unmanned aerial vehicle image air-three free network construction method based on minimum spanning tree - Google Patents
Special unmanned aerial vehicle image air-three free network construction method based on minimum spanning tree Download PDFInfo
- Publication number
- CN107560593B CN107560593B CN201710747621.XA CN201710747621A CN107560593B CN 107560593 B CN107560593 B CN 107560593B CN 201710747621 A CN201710747621 A CN 201710747621A CN 107560593 B CN107560593 B CN 107560593B
- Authority
- CN
- China
- Prior art keywords
- sub
- network
- image
- point
- points
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
- Navigation (AREA)
Abstract
The method for constructing the special unmanned aerial vehicle image air-three free network based on the minimum spanning tree solves the problems of relative orientation, model connection failure and the like in the conventional automatic aerial triangulation aiming at special unmanned aerial vehicle images such as falling water, cloud shielding, overlapping leaks and the like, automatically constructs the air-three free network in the whole area under the condition of not providing POS auxiliary data, obtains initial values of external orientation elements and object space point coordinates of all images in a measurement area, performs light beam method area network adjustment, and integrally solves the spatial coordinates of the external orientation elements and all object space points of each image in the whole area. If a small number of known control point coordinates are provided, the entire area can be brought into the known control point ground coordinate system, achieving absolute orientation. The processing level and the processing capacity of the images of the special unmanned aerial vehicle are greatly improved.
Description
Technical Field
The invention relates to a construction method of an unmanned aerial vehicle image air-three free network, in particular to a construction method of a special unmanned aerial vehicle image air-three free network based on a minimum spanning tree, and belongs to the technical field of unmanned aerial vehicle image processing methods.
Background
the low-altitude remote sensing system is a miniaturized and specialized remote sensing monitoring system with high maneuverability and low cost. The unmanned aerial vehicle is used as a flight platform, the high-resolution remote sensing equipment is used as an airborne sensor, the high-resolution remote sensing data is obtained as an application target, and the ground rapid real-time investigation and monitoring capability is achieved. As an emerging remote sensing technology, the advantages are mainly embodied in the following aspects: firstly, an airport is not needed (under the general condition, an airspace is not needed), and the lifting is flexible and convenient to implement various emergency tasks; secondly, the equipment is convenient to carry and transfer and has strong maneuverability; thirdly, the cost of platform construction, maintenance and operation is extremely low; fourthly, the influence of weather is small, the air flight is flexible, and the air flight device is suitable for rapid operation in small and medium areas; fifthly, low-altitude photography is performed, and the image spatial resolution is high; and sixthly, the image with high overlapping degree can be obtained, and the reliability and the precision of subsequent data processing are favorably improved.
The unmanned aerial vehicle low-altitude remote sensing system is rapidly developed and applied in recent years due to unique advantages, and particularly plays a great role in aspects such as forestry investigation, national soil resource investigation, geological disaster monitoring, large-scale topographic mapping and map updating, urban three-dimensional modeling and the like.
When unmanned aerial vehicle carries out the operation at several hundred meters low latitude, because its self weight is lighter, the air current influence is big, makes its flight attitude in the air very unstable, leads to the image of acquireing to have great rotatory partial and overlap degree scheduling problem inadequately to arouse the aerial photograph leak, influence the post processing of unmanned aerial vehicle data, these problems bring following difficulty for the automatic sky three processing of unmanned aerial vehicle remote sensing image: the method has the advantages that firstly, the adjacent images have large rotating deflection angles and large scale difference, the success rate and the reliability of gray scale correlation are reduced, and a series of problems of subsequent automatic relative orientation and the like are influenced; secondly, the flight path is bent, and the course overlapping degree and the side overlapping degree are irregular, so that difficulty is brought to extraction of the connection point; thirdly, the traditional method for constructing the navigation band fails due to the aerial photography loophole, and the full-automatic aerial triangulation of the survey area cannot be carried out.
Aerial triangulation, a key technique in photogrammetry, is the basis for a subsequent series of photogrammetry products and applications, such as creating DTMs, digital ortho images (DOM), stereography, and the like. For a general unmanned aerial vehicle image (relative to a 'special' image mentioned in the invention), according to a conventional photogrammetry air-to-air three-dimensional operation process, the land coordinates of external orientation elements and encryption points of all images in a measurement area can be obtained by carrying out block adjustment, and then the required products (such as DEM, DOM and the like) can be obtained by carrying out subsequent processing.
however, the prior art can not realize full-automatic, high-efficiency and high-precision aerial triangulation. For example, for the drowning image, a manual processing method is adopted to add connection points at standard points or measure a certain number of connection points along the edge of the water area, and then relative orientation is performed to help solve the problem of interruption of the air and the water. However, the most traditional manual point adding method seriously hinders the improvement of production efficiency due to the defects of non-intelligence, low efficiency, labor consumption and the like. Moreover, for large-area drowning images, the relative orientation and the failure of model connection cannot be solved by a manual dotting method sometimes. At this point, the flight path can only be divided into segments to clear the drowning area. However, the conventional space-three area division is manually divided and designed according to actual landforms, that is, the human eyes browse each image and then judge which images may not be automatically matched, relatively oriented or connected by models. The traditional method of dividing the regions by subjective judgment of human eyes is not high in accuracy rate, and misjudgment or omission is easy to occur. The division and modification of sub-measurement areas of the measurement areas need to be repeatedly carried out, so that the complexity and the workload of the operation are greatly increased, and the production efficiency is also seriously influenced.
In the prior art, when a GPS/IMU positioning and orientation technology is used for carrying out island aerial triangulation, a high-precision POS system is carried on unmanned aerial vehicle equipment to directly measure external orientation elements of a shooting instant drowning image, and forward intersection is carried out to obtain space coordinates of island and land image object points. The method for directly positioning geography by using the POS system can realize no ground control in a measuring area, improves the operation efficiency, and is the most obvious advantage of the direct orientation method. However, since the integration method is not tight enough, it is difficult to meet the accuracy requirement of large-scale topographic map mapping, and strict system parameter calibration is usually required, thereby increasing the complexity of the operation. At present, because the volume of unmanned aerial vehicle system is generally all less, and load is lighter, load POS system and carry out low latitude photography and have certain degree of difficulty in actual flight operation. Even some unmanned aerial vehicle low-altitude remote sensing systems cannot carry a GPS/IMU system, so that exterior orientation elements of images in difficult areas (such as drowning areas, desert areas and the like) cannot be directly obtained, which is a problem often encountered in practical application.
disclosure of Invention
Aiming at the defects of the prior art, the method for constructing the three-dimensional free network of the special unmanned aerial vehicle image based on the minimum spanning tree automatically constructs the three-dimensional free network of the aerial zone and the three-dimensional interruption caused by relative orientation or model connection failure in the conventional aerial triangulation operation process of a special unmanned aerial vehicle image (such as an image with large area of water falling, large area of forest coverage or cloud shielding, a photographic leak and the like), automatically constructs the three-dimensional free network of the whole area under the condition of not providing POS auxiliary data, obtains the initial values of the external orientation elements and the object space point coordinates of all images in the measuring area, performs beam method area network adjustment, and integrally solves the spatial coordinates of the external orientation elements and the object space points of each image in the whole area. If a small number of known control point coordinates are provided, the entire area can be brought into the known control point ground coordinate system, achieving absolute orientation. The processing level and the processing capacity of the images of the special unmanned aerial vehicle are greatly improved.
In order to achieve the technical effects, the technical scheme adopted by the invention is as follows:
The method for constructing the three-free-network-in-air special unmanned aerial vehicle image based on the minimum spanning tree is used for carrying out aerial triangulation processing on the special unmanned aerial vehicle image and comprises the following three steps:
Firstly, establishing a sub-flight-zone model, establishing a single three-dimensional model for each sub-flight-zone according to a continuous method in a relatively oriented mode, and then performing model connection to establish a free network of the sub-flight-zone;
establishing a sub-flight zone model according to each divided sub-flight zone, and establishing a corresponding sub-flight zone model, wherein the method comprises the following steps: firstly, measuring coordinates of image points and correcting system errors; secondly, establishing a single three-dimensional model by adopting a continuous method for relative orientation; thirdly, model connection is carried out, and a sub-navigation band free network is established;
Secondly, constructing a regional free network, firstly calculating the connection strength between adjacent sub-navigation bands and establishing a corresponding sub-navigation band relation graph, then selecting a connection edge of the regional network based on the principle of a minimum spanning tree and searching for an optimal initial point, and finally combining coordinate systems of all the sub-navigation bands in the region through absolute orientation;
The method comprises the following steps of constructing a regional free network to connect all sub-navigation bands in a survey region into a whole through absolute orientation: firstly, defining the connection relationship between an upper sub-navigation band and a lower sub-navigation band, namely defining a weight; secondly, establishing a relation graph, namely an adjacency matrix, of the whole sub-flight zone according to the defined connection relation; thirdly, selecting a connecting edge of the area network based on the principle of a minimum spanning tree, finding an optimal starting sub-navigation band as a starting point of area growth, taking a coordinate system of the starting sub-navigation band as a reference coordinate system, and bringing coordinates of other sub-navigation bands into a coordinate system unified with the starting sub-navigation band through absolute orientation to complete the construction of the air-to-air free network of the whole measuring area;
Thirdly, adjusting the block by using a beam method, and integrally solving strict solutions of external orientation elements and encrypted point coordinates of all images in a measurement area by using a collinear condition of three points including an image point, an object point and a photographing center, wherein the strict solutions are used for detecting a network construction result;
The adjustment of the area net by the beam method comprises the following steps: determining approximate values of external orientation elements and coordinates of undetermined points of each image, carrying out relative orientation and model connection on each stereo pair to construct a free flight band network, and carrying out absolute orientation on a flight path by utilizing a control point in the flight band and a public point between adjacent flight paths to obtain the ground coordinates of the external orientation elements and the encrypted points of each image as the approximate values of unknown numbers; secondly, from the coordinates of the control points and the image points of the undetermined points on each image, listing an error equation according to the collinear condition equation of each photographing light; thirdly, establishing a modified equation by a point-by-point method, and solving external orientation elements of each image according to a cyclic block solving method; and fourthly, obtaining the ground coordinates of the undetermined point according to the space forward intersection, and taking the average value of the common points of the adjacent images as a final result.
The method for constructing the special unmanned aerial vehicle image air-three free network based on the minimum spanning tree further comprises the following specific steps of:
Step 1, measuring coordinates of image points and correcting system errors; measuring the plane coordinates of the encryption point image selected in advance for each image pair, and correcting the system error of the encryption point image;
2, establishing a single three-dimensional model by adopting relative orientation of a continuous method; calculating the relative orientation element of the right photo relative to the left photo by taking the left photo of each image pair in the sub-navigation band as a reference, calculating the relative orientation element of the image pair, and then calculating the coordinates of the model points in the auxiliary coordinate systems of the respective image spaces according to a front intersection method to establish a single three-dimensional model;
step 3, model connection is carried out, and a sub-navigation band free network is established; the premise of connecting the single models into the sub-flight-zone model is that different scales of each model are classified into a unified scale, the scales of the latter model are classified into the scales of the former model from left to right in sequence under the condition that elevations of three connecting points in overlapping areas of adjacent image pairs are equal, the sub-flight-zone model taking the scale of the first model as a reference is established, and model point coordinates and station shooting point coordinates of all stereo image pairs are incorporated into a unified coordinate system of the whole flight-zone.
The method for constructing the special unmanned aerial vehicle image air-to-air three-free network based on the minimum spanning tree is further characterized in that in the area free network constructed in the second step, the upper and lower sub-flight band relation definition considers the sub-flight bands as the vertexes of the network in the graph theory, if the adjacent upper and lower sub-flight bands have the connection relation, the corresponding two vertexes are connected with each other by edges, and the size of the connection relation between the adjacent sub-flight bands is determined by the following factors:
The method comprises the following steps of firstly, the number of common points is defined as the number A of the common points in the overlapping range between adjacent sub-navigation bands, the number of the common points is normalized, and A is expressed as: a is s/N, where s represents the number of common points between adjacent sub-bands, and N represents the total number of common points or the maximum number of inter-band points;
Second, common point distribution, which refers to the image point distribution B of common points in the overlapping range between adjacent sub-strips, and respectively calculates the optimal distribution d for each object space point between stripsminoptimum distribution dminBased on the image point with the same name closest to the center of the image, then the optimal distribution of all object points is averaged to be used as a measure for the distribution size of the common point between the two sub-bands, so B is expressed as: b ═ Σ dmin) H is the number of object space points between the air belts, and B is normalized in the same way;
Thirdly, the overlapping degree refers to the number C of the same-name image points of the common points in the overlapping range between the adjacent sub-navigation bands, and the average overlapping degree between the navigation bands is taken as the measurement standard of C, so that C can be expressed as: c ═ Σ ki) Q, where kiRepresenting the number of redundant observations of the ith object space point, q is the number of object space points, and similarly, C is normalized to [0, 1%]To (c) to (d);
the magnitude of the relationship between sub-bands, i.e. the weight w on the edge in the network, is expressed as:
w＝μ1A+μ2B+μ3c (wherein,. mu.)1+μ2+μ3＝1)
Wherein, mu1、μ2、μ3The proportional coefficients are A, B, C respectively, and represent the weights of the number of common points, the distribution of the common points and the overlapping degree in the relation measurement between the two sub-strips in sequence, and the sum of the three is 1.
the method for constructing the three-dimensional free network of the special unmanned aerial vehicle image based on the minimum spanning tree is further characterized in that in the second step of constructing the regional free network, a relational graph of the whole sub-navigation band is established through an adjacent matrix, for a non-directional network formed by n vertexes, the size of the adjacent matrix is n multiplied by n, elements in the matrix are represented by Edge [ i ] [ j ], i and j are vertexes, W is a weight, and the adjacent matrix of the non-directional network can be defined as follows:
in the adjacent matrix of the undirected mesh, if 0< Edge [ i ] [ j ] <infinity, it means that there is an undirected Edge between vertex i and vertex j, the weight is Edge [ i ] [ j ], the greater the weight is, the greater the connection strength between the two vertices is, and if Edge [ i ] [ j ] is infinity, it means that there is no Edge connection between vertex i and vertex j.
The method for constructing the special unmanned aerial vehicle image air-three free network based on the minimum spanning tree is further characterized in that in the construction of the regional free network in the second step, all sub-navigation zones in a serial connection region are selected according to the connection mode with the maximum connection strength of the whole regional network, so that the weight sum of all the zones is maximum, the maximum spanning tree is converted into a counterexample of the minimum spanning tree, the connection edge of the regional network is obtained based on the principle of the minimum spanning tree, and three conditions are met: firstly, only the existing edges in the network can be used for constructing the spanning tree; secondly, only n-1 edges can be used for connecting n vertexes in the network, wherein n is the number of the vertexes; third, no loop can be generated.
The method for constructing the special unmanned aerial vehicle image air-three free network based on the minimum spanning tree is characterized in that the maximum spanning tree is converted into the minimum spanning tree, the minimum spanning tree is negative according to the weight, the Prim algorithm is adopted for the minimum spanning tree, the vertex is taken as the leading factor, and from the initial vertex, other vertexes are sequentially added into the spanning tree by selecting the currently available minimum weight edge.
The method for constructing the special unmanned aerial vehicle image air-three free network based on the minimum spanning tree is further characterized in that in the constructed regional free network in the second step, the selection of the optimal starting point adopts graded weight reduction to inhibit the growth of the network to the depth direction, and specifically comprises the following steps: when a certain vertex is selected as a starting point of region growth, the starting point is firstly taken as a level 1 vertex, other vertexes directly connected with the level 1 vertex are taken as level 2 vertexes, other vertexes directly connected with the level 2 vertex are taken as level 3 vertexes, the series of all vertexes in the network is determined by analogy, and the hierarchical weight reduction is to perform weight reduction treatment on a weight w (i, i +1) on the edge between the level i vertex and the level i +1 vertex according to the following formula:
w′(i，i+1)＝λi-1*w(i，i+1) (i＝1，2，3…)
wherein w is the weight, λi-1is weight reducing coefficient, lambda is constant, i-1 is the weight reducing series of the edge, W' (i, i +1) is the result of the weight reducing processing on the edge between the ith level vertex and the (i +1) th level vertex, the edge directly connected with the 1 level vertex is not subjected to the weight reducing processing, each sub-navigation band is selected as the starting point in turn, and the weight sum W of the connecting edges in the region is respectively calculated by adopting the strategy of graded weight reducing1，W2，…，WnComparison W1，W2，…，WnThe maximum weight and the corresponding sub-navigation band are selected as the starting point of the region growing, namely the optimal starting point to be searched.
The method for constructing the special unmanned aerial vehicle image air-three free network based on the minimum spanning tree further comprises the following specific steps of in the construction of the regional free network in the second step, constructing the regional free network in an absolute orientation mode: firstly, solving seven absolute orientation elements according to a common point between adjacent sub-navigation bands, solving three offsets and a scale for eliminating rough differences of the seven absolute orientation elements, then solving three rotation angles, taking an object point coordinate system corresponding to an optimal starting point as a reference coordinate system of the whole area network, and sequentially bringing the camera stations and model point coordinates of other sub-navigation bands into a coordinate system unified with the starting sub-navigation band through an absolute orientation formula according to a generated area network connection mode to finish the unification of the whole area coordinate system.
Compared with the prior art, the invention has the advantages that:
1. The method for constructing the special unmanned aerial vehicle image air-three free network based on the minimum spanning tree solves the problems of relative orientation, model connection failure and the like in conventional automatic aerial triangulation aiming at special unmanned aerial vehicle images such as falling water, cloud shielding, overlapping leaks and the like, solves the problem of automatic construction of the image air-three free network in difficult areas by applying relevant theories and algorithms of graph theory under the condition of not providing POS auxiliary data, and realizes the method by programming, so that the processing level and the processing capacity of the special unmanned aerial vehicle images are greatly improved.
2. The method for constructing the special unmanned aerial vehicle image air-three free network based on the minimum spanning tree introduces the concept of a sub-flight zone aiming at the condition that relative orientation or model connection failure occurs in a survey area. After the free network of each sub-navigation band is built, the construction work of the regional free network can be carried out. Firstly, defining the relation between sub-navigation bands (namely weight definition), converting the whole area network (formed by the sub-navigation bands) into a non-directional connected graph with the weight in graph theory (namely a non-directional network), and establishing the relation graph between the sub-navigation bands by means of an adjacent matrix. And then generating a connection mode of the area network based on the principle of a minimum spanning tree, and selecting an optimal starting point by adopting a strategy of grading and weight reduction. And finally, carrying out absolute orientation to unify the coordinate system of the whole area, namely, finishing the construction of the area free network. The problem that relative orientation or model connection failure occurs in a measurement area is effectively solved. If a small number of known control point coordinates are provided, the entire area can be brought into the known control point ground coordinate system, achieving absolute orientation.
3. the method for constructing the special unmanned aerial vehicle image air-three free network based on the minimum spanning tree solves the difficult problem in air triangulation, solves the very key technical problem in unmanned aerial vehicle photogrammetry, and lays a good foundation for a series of subsequent photogrammetry products and applications. The method is beneficial to the low-altitude unmanned remote sensing technology as an important means for acquiring emerging space data, and is widely applied in many fields. The method is beneficial to high resolution of the acquired image, fast data acquisition and short mapping period, and plays a greater advantage in large-scale topographic map mapping and map updating in medium and small ranges.
Drawings
Fig. 1 is a schematic view of a sub-flight band of the method for constructing an aerial three-free network of images of a special unmanned aerial vehicle provided by the invention.
Fig. 2 is a schematic view of a directed network of the method for constructing a special unmanned aerial vehicle image air-three free network provided by the invention.
FIG. 3 is a schematic diagram of a minimum spanning tree for a undirected mesh architecture provided by the present invention.
Fig. 4 is a schematic view of a connection mode of a sub-flight band of the method for constructing the image air-to-air three-free network of the special unmanned aerial vehicle provided by the invention.
Fig. 5 is a schematic view of absolute orientation of a sub-flight band of the method for constructing an aerial three-free-network of images of a special unmanned aerial vehicle provided by the invention.
FIG. 6 is a schematic diagram of a user zooming operation of the gesture interaction method provided by the present invention.
fig. 7 is a flowchart of an overall technical route of the method for constructing the image air-three free network of the special unmanned aerial vehicle provided by the invention.
Detailed Description
the technical scheme of the method for constructing the special unmanned aerial vehicle image air-three free network based on the minimum spanning tree provided by the invention is further described below with reference to the accompanying drawings, so that a person skilled in the art can better understand the method and can implement the method.
Referring to fig. 1 to 7, when the method for constructing the aerial three free networks based on the minimum spanning tree is used for aerial triangulation of a type of special unmanned aerial vehicle images (such as images falling into water in a large area, blocking clouds and the like) according to a conventional photogrammetry processing flow, the automatic creation of a flight band and the aerial three interruption are often caused by relative orientation and model connection failure, and further subsequent products (such as DEM, orthographic images and the like) cannot be manufactured. If the flight path is divided into a plurality of sections to skip over difficult images such as falling into water, the whole survey area can be regarded as being composed of a plurality of broken flight bands (introducing the concept of 'sub flight bands'). And (3) after each sub-flight zone free network is constructed according to relative orientation and model connection, establishing an air-to-three free network of the whole measuring area, namely connecting all the sub-flight zones, and then performing area network beam adjustment. Carry out aerial triangulation to special unmanned aerial vehicle image and handle, include three steps:
firstly, establishing a sub-flight-zone model, establishing a single three-dimensional model for each sub-flight-zone according to a continuous method in a relatively oriented mode, and then performing model connection to establish a free network of the sub-flight-zone;
Establishing a sub-flight zone model according to each divided sub-flight zone, and establishing a corresponding sub-flight zone model, wherein the method comprises the following steps: firstly, measuring coordinates of image points and correcting system errors; secondly, establishing a single three-dimensional model by adopting a continuous method for relative orientation; thirdly, model connection is carried out, and a sub-navigation band free network is established;
secondly, constructing a regional free network, firstly calculating the connection strength between adjacent sub-navigation bands and establishing a corresponding sub-navigation band relation graph, then selecting a connection edge of the regional network based on the principle of a minimum spanning tree and searching for an optimal initial point, and finally combining coordinate systems of all the sub-navigation bands in the region through absolute orientation;
The method comprises the following steps of constructing a regional free network to connect all sub-navigation bands in a survey region into a whole through absolute orientation: firstly, defining the connection relationship between an upper sub-navigation band and a lower sub-navigation band, namely defining a weight; secondly, establishing a relation graph, namely an adjacency matrix, of the whole sub-flight zone according to the defined connection relation; thirdly, selecting a connecting edge of the area network based on the principle of a minimum spanning tree, finding an optimal starting sub-navigation band as a starting point of area growth, taking a coordinate system of the starting sub-navigation band as a reference coordinate system, and bringing coordinates of other sub-navigation bands into a coordinate system unified with the starting sub-navigation band through absolute orientation to complete the construction of the air-to-air free network of the whole measuring area;
thirdly, adjusting the block by using a beam method, and integrally solving strict solutions of external orientation elements and encrypted point coordinates of all images in a measurement area by using a collinear condition of three points including an image point, an object point and a photographing center, wherein the strict solutions are used for detecting a network construction result;
The adjustment of the area net by the beam method comprises the following steps: determining approximate values of external orientation elements and coordinates of undetermined points of each image, carrying out relative orientation and model connection on each stereo pair to construct a free flight band network, and carrying out absolute orientation on a flight path by utilizing a control point in the flight band and a public point between adjacent flight paths to obtain the ground coordinates of the external orientation elements and the encrypted points of each image as the approximate values of unknown numbers; secondly, from the coordinates of the control points and the image points of the undetermined points on each image, listing an error equation according to the collinear condition equation of each photographing light; thirdly, establishing a modified equation by a point-by-point method, and solving external orientation elements of each image according to a cyclic block solving method; and fourthly, obtaining the ground coordinates of the undetermined point according to the space forward intersection, and taking the average value of the common points of the adjacent images as a final result.
firstly, establishing a sub-navigation band model
The flight band is an image sequence which is obtained by performing aerial photography along a certain direction and is overlapped with each other in front and back. The sub-flight band is a plurality of sectional flight bands formed by breaking the original flight band due to falling water, cloud shielding, photographic loopholes and the like on the basis of one flight band obtained by conventional aerial photography, the sectional flight bands are called as sub-flight bands, and each sub-flight band is specified to at least comprise two (more than two) images.
after dividing each sub-flight zone in the area according to the definition, a corresponding sub-flight zone model (namely a sub-flight zone free network) needs to be established, and the method mainly comprises the following steps:
1. And measuring the coordinates of the image points and correcting the system error. The coordinates of the image plane of the encryption point selected in advance for each image pair are measured, and systematic error correction is carried out on the coordinates, and the process is generally automatically completed by photogrammetry software.
2. and establishing a single three-dimensional model by adopting a continuous method for relative orientation. And calculating the relative orientation element of the right photo relative to the left photo by taking the left photo of each image pair in the sub-navigation band as a reference. After the relative orientation elements of the image pairs are calculated, the coordinates of the model points in the auxiliary coordinate systems of the respective image spaces are calculated according to a front intersection method, and a single three-dimensional model is established.
3. And connecting the models and establishing a free network of the sub-navigation band. The premise of connecting the single models into the sub-navigation band model is to put different scales of the models into a uniform scale. Generally, under the condition that the elevations of three connecting points in the overlapping area of adjacent pairs are equal, a scale of a next model is sequentially classified into a scale of a previous model from left to right, and a sub-flight-band model taking the scale of a first model as a reference is established. And incorporating the model point coordinates and the camera station point coordinates of all the stereopairs into a coordinate system unified by a full flight band.
Thus, the building of the sub-navigation band model (namely, the sub-navigation band free network) is completed, and the next step is to build the regional free network.
secondly, constructing a regional free net
The problem of regional growth of the sub-flight zones is involved in building a regional free network, namely, all the sub-flight zones in a survey area are connected into a whole by adopting a certain connection mode through absolute orientation. Firstly, defining the connection relation (namely weight definition) between an upper sub-navigation band and a lower sub-navigation band, then establishing a relation graph (namely an adjacent matrix) of the whole sub-navigation band according to the well-defined connection relation, then selecting the optimal connection edge of the regional network based on the principle of a minimum spanning tree, finding an optimal starting sub-navigation band as the starting point of regional growth, taking the coordinate system of the starting sub-navigation band as a reference coordinate system, and incorporating the coordinates of other sub-navigation bands into the coordinate system unified with the starting sub-navigation band through absolute orientation to complete the construction work of the air-three free network of the whole measuring area.
1. inter-sub-flight-band relationship definition
Regarding the sub-bands as the vertices of the network in the graph theory, if there is a connection relationship between the adjacent upper and lower sub-bands (because the left and right sub-bands are disconnected, the relationship between them is not considered, only the upper and lower sub-bands are considered), then the corresponding two vertices are connected by an edge, and then how to define the connection relationship (i.e. the weight value on the edge), i.e. to determine the relationship evaluation criteria between the sub-bands.
According to the principle of aerial photography, the side-to-side overlapping must exist between the adjacent upper and lower sub-bands, and a certain number of common points between the bands are considered to be distributed in the overlapping range. Therefore, to assess the connection (or strength) between sub-bands, it is necessary to associate a common point between bands. In addition to the number of common points, the distribution of the image points of the common points should be taken into account. In addition, the degree of overlap of the inter-flight band points (i.e., the number of redundant observations) is a reliability index, and should be reflected in the definition of this relationship with a large weight.
Therefore, the magnitude of the connection relationship between adjacent sub-bands determined by the present invention is mainly related to the following factors:
(1) number of common points: refers to the number of common points (denoted by the letter a) in the overlapping range between adjacent sub-strips. For ease of computation and comparison, the number of common points needs to be normalized, so a can be expressed as: and A is s/N, wherein s represents the number of common points between adjacent sub-bands, and N represents the total number of common points or the maximum number of inter-band points.
(2) Common point distribution: refers to the distribution of image points (denoted by the letter B) of common points in the overlapping range between adjacent sub-swaths. For each object space point between the flight paths, respectively calculating an optimal distribution (based on the image point with the same name closest to the center of the image, using dmin＝min(di) Expressed), then averaging the optimal distribution of all object points as a measure of the size of the distribution of the common points between the two sub-bands, so B can be expressed as: b ═ Σ dmin) And h is the number of object space points between the flight zones. Similarly, B needs to be normalized.
(3) Overlapping degree: also known as unwanted observations, refers to the number of like-name image points (denoted by the letter C) of common points in the overlapping range between adjacent sub-swaths, and the average overlap between swaths is taken as a measure of C, so C can be expressed as: c ═ Σ ki) Q, where kiRepresenting the number of redundant observations of the ith object space point, and q is the object space pointAnd (4) the number. Similarly, C needs to be normalized to [0, 1%]In the meantime.
therefore, considering the above factors together, the magnitude of the relationship between sub-bands (i.e. the weight on the edge in the network) can be expressed as:
w＝μ1A+μ2B+μ3C (wherein,. mu.)1+μ2+μ3＝1)
Wherein, mu1、μ2、μ3The proportional coefficients are A, B, C respectively, and represent the weights of the number of common points, the distribution of the common points and the overlapping degree in the relation measurement between the two sub-strips in sequence, and the sum of the three is 1. Larger values indicate more importance of the corresponding factor. In the experiment, the magnitude should be reasonably allocated according to the actual situation, and the coefficient value is determined empirically (in the experiment of the present invention, let μ1＝0.4，μ2＝0.1，μ3＝0.5)。
2. sub-flight zone relation graph establishment
After the connection relationship between the upper and lower sub-navigation bands in the area is defined, the weight value on the edge in the network is determined, so that the whole area network (formed by the sub-navigation bands) can be represented as a non-directional connected graph (namely, a non-directional network, represented by G) with the weight in the graph theory. In the figure, a vertex (indicated by V) represents a navigation band in the area, two vertices are connected by an edge (indicated by E) to indicate that two corresponding navigation bands are overlapped (or connected), and the connection strength is represented by a weight W on the edge. The relationship between sub-bands is established as follows.
the storage mode of the graph can be represented by an adjacency matrix, so that the relation graph between the sub-bands can be established by the adjacency matrix. For a undirected mesh consisting of n vertices with a neighboring matrix of size n × n, where the elements in the matrix are represented, the neighboring matrix of the undirected mesh can be defined as:
In the adjacent matrix of the undirected mesh, if 0< Edge [ i ] [ j ] <infinity, it means that there is an undirected Edge between vertex i and vertex j, and the weight is Edge [ i ] [ j ], and the greater the weight, the greater the connection strength between two vertices. If Edge [ i ] [ j ] is ∞, it means that there is no Edge connection between vertex i and vertex j. Sometimes, for processing convenience, the weight value when not connected can also be defined as 0, and is consistent with the element on the diagonal (as is the case in the present invention).
As shown in the schematic diagram of the sub-bands in fig. 1, the whole area is divided into 11 sub-bands, the red part in the diagram represents the sub-bands, and the image without red mark represents the broken image (the part is ignored for the moment). According to the common point and image point information of the overlapping area between the adjacent sub-navigation bands, a corresponding undirected network (as shown in fig. 2, 11 vertexes and 13 edges are total) is established for the graph 1, and the weight information of the edge is obtained, so that an adjacent matrix can be constructed to store the vertexes, the edges and the weight information of the undirected network. Therefore, a sub-navigation band relation graph of the whole measuring area is established.
3. Area network connection edge selection
After the whole regional network (formed by the sub-navigation bands) is converted into a non-directional network and a relation graph between the sub-navigation bands is established, the connecting edges of the regional network are selected, namely, an optimal connecting mode is selected to connect all the sub-navigation bands in the region in series, and no loop can be generated.
There may be more than one way to connect all vertices in a graph, since there may be loop conditions in a connected graph. The weight on the edge of the network reflects the connection strength between the two sub-navigation bands, and the connection strength is higher if the weight is higher. Therefore, to select an "optimal" connection method to connect all vertices in the network, the sum of the weights of the edges should be maximized, so that the connection strength of the entire area network is maximized and the connection is the most secure.
The above problem can be understood as a counter example of the minimum spanning tree problem (requiring minimum sum of weights) in graph theory, that is, a maximum spanning tree is required to maximize the sum of weights of each edge, and three conditions are satisfied: firstly, only the existing edges in the network can be used for constructing the spanning tree; secondly, only n-1 edges can be used for connecting n vertexes in the network; third, no loop can be generated. The most studied problem in graph theory is the minimum spanning tree, and it is easy to convert the maximum spanning tree into the minimum spanning tree (for example, the weight is negative), so that the problem can be solved easily. Therefore, the invention is based on the principle of minimum spanning tree to find the connection edge of the area network.
Fig. 3 is a minimal spanning tree (obtained according to Prim algorithm of the minimal spanning tree) constructed in fig. 2, in which 10 edges in the original network are used to connect 11 vertices, no loop is generated, and the sum of weights of the edges is maximized. Fig. 4 is a schematic view of a corresponding connection mode of the sub-navigation belt. Thus, the connecting edge of the whole area network is selected.
The basic idea of Prim algorithm is that vertices are dominant: starting from the initial vertex, other vertexes are added to the spanning tree in sequence by selecting the currently available minimum weight edge.
And (3) dividing the vertex set V into two subsets T and T' in the Prim algorithm by setting the connected undirected network as G (V, E). T: a current spanning tree vertex set; t': the set of vertices that do not belong to the current spanning tree. It is clear that there are: t ═ V ═ U ═ T'.
(1) The Prim algorithm comprises the following specific processes:
1) Selecting a starting vertex u from a connected undirected network G0it is first added to the set T; then select and u0Associated edge (u) with minimum weight0V), add vertex v to vertex set T.
2) in each step, the side (u, v) with the minimum weight is selected from the sides of one vertex (set as u) in T and the other vertex (set as v) in T', and the vertex v is added into the set T. This continues until all vertices in the network are added to the spanning tree vertex set T.
(2) Algorithm legend
Initially, the set T is empty, the starting vertex 1 is added to the set T, and then each vertex is added to the set T as follows:
1) Only 1 vertex in the set T is present, namely vertex 1, one vertex is in the edge of T, the other vertex is in the edge of T', the edge with the minimum weight is (1,6), the weight is 10, and the vertex 6 is added into the set T through the edge;
2) There are 2 vertexes in the set T, i.e. vertexes 1 and 6, one vertex is in the edge of T, the edge with the smallest weight is (6,5), the weight is 25, and vertex 5 is added to the set T through this edge;
3) There are 3 vertexes in the set T, i.e. vertexes 1,6, 5, one vertex is in T, the other vertex is in T' side, the side with the smallest weight is (5,4), its weight is 22, through this side vertex 4 is added to the set T;
4) There are 4 vertexes in the set T, namely, vertexes 1,6, 5,4, one vertex is in T, the other vertex is in T' edge, the edge with the minimum weight is (4,3), the weight is 12, and the vertex 3 is added to the set T through the edge;
5) there are 5 vertexes in the set T, i.e. vertexes 1,6, 5,4, 3, one vertex is in T, the other vertex is in T' side, the side with the minimum weight is (3,2), the weight is 16, and vertex 2 is added to the set T through this side;
6) Now there are 6 vertices in the set T, i.e. vertices 1,6, 5,4, 3,2, one vertex in T and the other vertex in T', the edge with the smallest weight is (2,7) and the weight is 14, and the vertex 7 is added to the set T through this edge.
All the vertexes are added into the set T, the minimum spanning tree is constructed, and the final constructed minimum spanning tree has the weight sum of 99.
4. optimal starting point selection
After the effective connecting edge of the whole area network is selected, all the sub-navigation bands are connected in series. The networking strategy of the invention is to start from a certain sub-navigation band and carry out clustering growth of the region according to the selected connection mode, thereby connecting all the sub-navigation bands. Considering that absolute orientation between adjacent sub-bands has the effect of error propagation, the network-building effect of selecting different sub-bands as starting points (or parent nodes) will naturally be different. That is, the present invention also relates to the selection of an optimal starting point.
In order to reduce error propagation paths as much as possible in consideration of error propagation factors, the invention adopts a hierarchical weight reduction strategy to inhibit the increase of the network in the depth direction. The specific method comprises the following steps: when a certain vertex is selected as a starting point of region growing, the starting point is firstly taken as a level 1 vertex, other vertexes directly connected with the level 1 vertex are taken as level 2 vertexes, other vertexes directly connected with the level 2 vertex are taken as level 3 vertexes, and the like, so that the series of all vertexes in the network is determined. The "hierarchical weight reduction" means that the weight w (i, i +1) on the edge between the ith-level vertex and the (i +1) th-level vertex is reduced according to the following formula:
w′(i，i+1)＝λi-1*w(i，i+1)(i＝1，2，3…)
Wherein w is the weight, λi-1Is a weight-reducing coefficient, λ is a constant (in this embodiment, λ is 0.8), i-1 is the weight-reducing series of the edge, W' (i, i +1) is the result of weight-reducing processing on the edge between the ith-level vertex and the (i +1) th-level vertex, the edge directly connected with the 1-level vertex is not weight-reducing processed, each sub-navigation band is sequentially selected as a starting point, and the weight sum W of the connected edges in the region is respectively calculated by adopting a strategy of hierarchical weight-reducing1，W2，…，WnComparison W1，W2，…，WnThe maximum weight and the corresponding sub-navigation band are selected as the starting point of the region growing, namely the optimal starting point to be searched.
the best starting point is found for the maximum spanning tree in fig. 3 according to the above idea, and the finally found best starting point is the vertex 4, which means that the starting sub-band for region growing should be selected as the sub-band 4.
5. Absolute orientation building of area free nets
after the optimal starting point of the area network is selected, clustering of the areas is required to be carried out, and the area free network is constructed. Because the coordinate systems of object space points of the sub-navigation belts are inconsistent at present, and the coordinate systems of the object space points of the whole area are unified when a free network of the whole area is constructed, the absolute orientation is used. The specific method comprises the following steps: firstly, solving seven absolute orientation elements (namely seven parameters) by adopting a two-step solution method according to common points between adjacent sub-navigation bands, firstly solving three offsets and a scale, eliminating rough differences, and then solving three rotation angles, then taking an object side point coordinate system corresponding to the optimal starting point as a reference coordinate system (or called as a reference coordinate system) of the whole area network, and sequentially incorporating the camera stations and model point coordinates of other sub-navigation bands into a coordinate system unified with the starting sub-navigation band according to a generated area network connection mode (figure 4) through the following absolute orientation formula, thereby completing the unification of the whole area coordinate system.
(Absolute orientation formula)
where λ is the scale of the model, R is a rotation matrix consisting of three rotation angles (Φ, Ω, K), and (Δ X, Δ Y, Δ Z) are translation coefficients. (X, Y, Z) is the object point (or filming point) coordinates to be converted, and is the object point (or filming point) coordinates in the reference coordinate system.
The schematic diagram of absolute orientation of the sub-flight band shown in fig. 5 shows the direction of absolute orientation as indicated by the arrow, and the numbers next to the arrow indicate the order of absolute orientation. And taking the sub-navigation band 4 as a starting point and the coordinate system thereof as a reference coordinate system, and sequentially (according to the direction and the sequence of the front end in the figure 5) transferring the coordinate systems of other sub-navigation bands to the coordinate system consistent with the sub-navigation band 4, so that all the sub-navigation band coordinate systems of the whole area are unified, namely the air-to-three free net of the whole area is constructed.
Third, adjustment of area network by light beam method
After the three free nets of the whole measuring area are constructed, adjustment work of the whole area can be carried out, and a beam method area net adjustment method is adopted. Based on the principle that three points of an image point, an object space point and a photographing center are collinear, a single light beam is used as a basic unit of adjustment, a collinear condition equation is used as a mathematical model of the adjustment, coordinates of the image point are used as an observed value, an error equation is listed, and strict solutions of external orientation elements and encrypted point coordinates of all images in a measuring area are integrally solved. Wherein, the initial value of the unknown number is obtained by establishing the sub-flight band free net and the building area free net in the front. If the ground coordinates of the known control point are provided, the whole area can be included in the ground coordinate system of the known control point.
the adjustment of the area network by the beam method is the strictest method, the theory and the algorithm of the method are relatively mature and comprehensive, and a plurality of adjustment software with powerful functions, such as PATB, BINGO and the like, are developed internationally. The invention uses the beam method area network adjustment method as follows:
The light beam method area network aerial triangulation is provided based on the principle that three points of an image point, a photographing center and an object space point are collinear, and the basic idea is that a beam of light formed by each image is used as a basic unit of adjustment, and a collinear condition equation is used as a basic equation of adjustment. The rays of the common point between the models are optimally intersected through the rotation and translation of each light beam in the air, and the whole area is included in a known control point ground coordinate system. Therefore, a uniform error equation in the whole area is established, and six exterior orientation elements of each photo in the whole area and the ground coordinates of all points to be solved are integrally solved.
The basic flow of beam method area network aerial triangulation is shown in fig. 6, and the main contents thereof include:
(1) Determining the approximate value of the exterior orientation element and the coordinate of the undetermined point of each image (the specific method is that each stereo pair is relatively oriented and model-connected to construct a free flight band network, and the control points in the flight band and the public points between adjacent flight bands are used for absolutely orienting the flight bands to obtain the ground coordinates of the exterior orientation element and the encryption point of each image as the approximate value of an unknown number);
(2) Starting from the coordinates of the image points of the control point and the undetermined point on each image, listing an error equation according to a collinear condition equation of each photographing light;
(3) Establishing a modified equation point by point, and solving one type of unknowns according to a cyclic block solving method, wherein the outer orientation element of each image is usually solved firstly;
(4) and (4) obtaining the ground coordinates of the undetermined points according to the space forward intersection, and taking the average value of the common points of the adjacent images as a final result.
Four, integral technical route
In summary, the overall technical route flow of the present invention can be described as fig. 7, and the method for automatically constructing an air-to-three free network for a special unmanned aerial vehicle image mainly includes the following steps:
(1) Establishing a sub-navigation band model: and (3) relatively orienting each sub-navigation band according to a continuous method to establish a single three-dimensional model, and then performing model connection to establish a free network of the sub-navigation band.
(2) Constructing a regional free network: firstly, calculating the connection strength between adjacent sub-strips and establishing a corresponding sub-strip relation graph, then selecting effective connection edges of an area network based on the principle of a minimum spanning tree and searching for an optimal starting point, and finally combining coordinate systems of all the sub-strips in the area through absolute orientation.
(3) Adjustment of the area network by a beam method: and (3) integrally solving a strict solution of the external orientation elements and the encrypted point coordinates of all the images in the measurement area by utilizing the collinear condition of the three points of the image point, the object space point and the photographing center, and taking the strict solution as an inspection of a network construction result.
The invention provides a method for constructing a special unmanned aerial vehicle image air-three free network based on a minimum spanning tree, which is characterized in that a navigation band failure and an air-three interruption are automatically created due to relative orientation or model connection failure in the conventional air triangulation operation process of a type of special unmanned aerial vehicle images (such as images with large-area water fall, large-area forest coverage or cloud shielding and camera holes), the air-three free network of the whole area is automatically constructed under the condition of not providing POS auxiliary data, the external orientation elements and the initial values of object space point coordinates of all images in a measured area are obtained, the adjustment of the area network by a light beam method is carried out, and the external orientation elements and the space coordinates of all object space points of each image in the whole area are integrally solved. The processing level and the processing capacity of the images of the special unmanned aerial vehicle are greatly improved.
the foregoing illustrates and describes the principles, general features, and advantages of the present invention. Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (8)
1. The method for constructing the three-free-network-in-air special unmanned aerial vehicle image based on the minimum spanning tree is characterized in that aerial triangulation processing is carried out on the special unmanned aerial vehicle image, and comprises the following three steps:
Firstly, establishing a sub-flight-zone model, establishing a single three-dimensional model for each sub-flight-zone according to a continuous method in a relatively oriented mode, and then performing model connection to establish a free network of the sub-flight-zone;
Establishing a sub-flight zone model according to each divided sub-flight zone, and establishing a corresponding sub-flight zone model, wherein the method comprises the following steps: firstly, measuring coordinates of image points and correcting system errors; secondly, establishing a single three-dimensional model by adopting a continuous method for relative orientation; thirdly, model connection is carried out, and a sub-navigation band free network is established;
Secondly, constructing a regional free network, firstly calculating the connection strength between adjacent sub-navigation bands and establishing a corresponding sub-navigation band relation graph, then selecting a connection edge of the regional network based on the principle of a minimum spanning tree and searching for an optimal initial point, and finally combining coordinate systems of all the sub-navigation bands in the region through absolute orientation;
The method comprises the following steps of constructing a regional free network to connect all sub-navigation bands in a survey region into a whole through absolute orientation: firstly, defining the connection relationship between an upper sub-navigation band and a lower sub-navigation band, namely defining a weight; secondly, establishing a relation graph, namely an adjacency matrix, of the whole sub-flight zone according to the defined connection relation; thirdly, selecting a connecting edge of the area network based on the principle of a minimum spanning tree, finding an optimal starting sub-navigation band as a starting point of area growth, taking a coordinate system of the starting sub-navigation band as a reference coordinate system, and bringing coordinates of other sub-navigation bands into a coordinate system unified with the starting sub-navigation band through absolute orientation to complete the construction of the air-to-air free network of the whole measuring area;
thirdly, adjusting the block by using a beam method, and integrally solving strict solutions of external orientation elements and encrypted point coordinates of all images in a measurement area by using a collinear condition of three points including an image point, an object point and a photographing center, wherein the strict solutions are used for detecting a network construction result;
The adjustment of the area net by the beam method comprises the following steps: determining approximate values of external orientation elements and coordinates of undetermined points of each image, carrying out relative orientation and model connection on each stereo pair to construct a free flight band network, and carrying out absolute orientation on a flight path by utilizing a control point in the flight band and a public point between adjacent flight paths to obtain the ground coordinates of the external orientation elements and the encrypted points of each image as the approximate values of unknown numbers; secondly, from the coordinates of the control points and the image points of the undetermined points on each image, listing an error equation according to the collinear condition equation of each photographing light; thirdly, establishing a modified equation by a point-by-point method, and solving external orientation elements of each image according to a cyclic block solving method; and fourthly, obtaining the ground coordinates of the undetermined point according to the space forward intersection, and taking the average value of the common points of the adjacent images as a final result.
2. The method for constructing the aerial three free network of the special unmanned aerial vehicle image based on the minimum spanning tree as claimed in claim 1, wherein the specific method for establishing the sub-navigation band model is as follows:
Step 1, measuring coordinates of image points and correcting system errors; measuring the plane coordinates of the encryption point image selected in advance for each image pair, and correcting the system error of the encryption point image;
2, establishing a single three-dimensional model by adopting relative orientation of a continuous method; calculating the relative orientation element of the right photo relative to the left photo by taking the left photo of each image pair in the sub-navigation band as a reference, calculating the relative orientation element of the image pair, and then calculating the coordinates of the model points in the auxiliary coordinate systems of the respective image spaces according to a front intersection method to establish a single three-dimensional model;
step 3, model connection is carried out, and a sub-navigation band free network is established; the premise of connecting the single models into the sub-flight-zone model is that different scales of each model are classified into a unified scale, the scales of the latter model are classified into the scales of the former model from left to right in sequence under the condition that elevations of three connecting points in overlapping areas of adjacent image pairs are equal, the sub-flight-zone model taking the scale of the first model as a reference is established, and model point coordinates and station shooting point coordinates of all stereo image pairs are incorporated into a unified coordinate system of the whole flight-zone.
3. the method for constructing the aerial three free network of the special unmanned aerial vehicle image based on the minimum spanning tree as claimed in claim 1, wherein in the area free network constructed in the second step, the relationship between the upper and lower sub-bands defines that the sub-bands are regarded as vertexes of the network in the graph theory, if the connection relationship exists between the adjacent upper and lower sub-bands, edges are given to the two corresponding vertexes to be connected, and the size of the connection relationship between the adjacent sub-bands is determined by the following factors:
the method comprises the following steps of firstly, the number of common points is defined as the number A of the common points in the overlapping range between adjacent sub-navigation bands, the number of the common points is normalized, and A is expressed as: a is s/N, where s represents the number of common points between adjacent sub-bands, and N represents the total number of common points or the maximum number of inter-band points;
Second, common point distribution, which refers to the image point distribution B of common points in the overlapping range between adjacent sub-strips, and respectively calculates the optimal distribution d for each object space point between stripsminoptimum distribution dminBased on the image point with the same name closest to the center of the image, then the optimal distribution of all object points is averaged to be used as a measure for the distribution size of the common point between the two sub-bands, so B is expressed as: b ═ Σ dmin) H is the number of object space points between the air belts, and B is normalized in the same way;
Thirdly, the overlapping degree refers to the number C of the same-name image points of the common points in the overlapping range between the adjacent sub-navigation bands, and the average overlapping degree between the navigation bands is taken as the measurement standard of C, so that C can be expressed as: c ═ Σ ki) Q, where kirepresenting the number of redundant observations of the ith object space point, q is the number of object space points, and similarly, C is normalized to [0, 1%]To (c) to (d);
the magnitude of the relationship between sub-bands, i.e. the weight w on the edge in the network, is expressed as:
w＝μ1A+μ2B+μ3C (wherein,. mu.)1+μ2+μ3＝1)
wherein, mu1、μ2、μ3The proportional coefficients are A, B, C respectively, and represent the weights of the number of common points, the distribution of the common points and the overlapping degree in the relation measurement between the two sub-strips in sequence, and the sum of the three is 1.
4. The method for constructing the image space-three free network of the special unmanned aerial vehicle based on the minimum spanning tree as claimed in claim 1, wherein in the area free network constructed in the second step, the relationship diagram of the whole sub-navigation band is constructed by an adjacency matrix, for a undirected network composed of n vertexes, the size of the adjacency matrix is n × n, elements in the matrix are represented by Edge [ i ] [ j ], i, j are vertexes, W is a weight, and the adjacency matrix of the undirected network can be defined as:
In the adjacent matrix of the undirected mesh, if 0< Edge [ i ] [ j ] <infinity, it means that there is an undirected Edge between vertex i and vertex j, the weight is Edge [ i ] [ j ], the greater the weight is, the greater the connection strength between the two vertices is, and if Edge [ i ] [ j ] is infinity, it means that there is no Edge connection between vertex i and vertex j.
5. The method for constructing the aerial three free network of the special unmanned aerial vehicle image based on the minimum spanning tree as claimed in claim 1, wherein in the constructing of the regional free network in the second step, all the sub-navigation bands in the region are connected in series according to the connection mode with the maximum connection strength of the whole regional network, so that the weight sum of all the bands is maximized, the maximum spanning tree is converted into the counterexample of the minimum spanning tree, the connection edge of the regional network is obtained based on the principle of the minimum spanning tree, and three conditions are satisfied: firstly, only the existing edges in the network can be used for constructing the spanning tree; secondly, only n-1 edges can be used for connecting n vertexes in the network, wherein n is the number of the vertexes; third, no loop can be generated.
6. The method for constructing the aerial three free network of the special unmanned aerial vehicle image based on the minimum spanning tree as claimed in claim 5, wherein the maximum spanning tree is converted into the minimum spanning tree, and the minimum spanning tree is negative according to the weight, the Prim algorithm is adopted for the minimum spanning tree, the vertex is taken as the leading factor, and from the initial vertex, other vertices are added into the spanning tree in sequence by selecting the currently available minimum weight edge.
7. the method for constructing the aerial three free network based on the minimum spanning tree image of the special unmanned aerial vehicle as claimed in claim 1, wherein in the area free network constructed in the second step, the selection of the optimal starting point adopts a hierarchical weight reduction to inhibit the growth of the network in the depth direction, and specifically comprises the following steps: when a certain vertex is selected as a starting point of region growth, the starting point is firstly taken as a level 1 vertex, other vertexes directly connected with the level 1 vertex are taken as level 2 vertexes, other vertexes directly connected with the level 2 vertex are taken as level 3 vertexes, the series of all vertexes in the network is determined by analogy, and the hierarchical weight reduction is to carry out weight reduction treatment on a weight w (i, i +1) on the edge between the level i vertex and the level i +1 vertex according to the following formula:
W’(i，i+1)＝λi-1*w(i，i+1)(i＝1，2，3…D)
Wherein w is the weight, λi-1Is weight reducing coefficient, lambda is constant, i-1 is the weight reducing series of the edge, W' (i, i +1) is the result of the weight reducing processing on the edge between the ith level vertex and the (i +1) th level vertex, the edge directly connected with the 1 level vertex is not subjected to the weight reducing processing, each sub-navigation band is selected as the starting point in turn, and the weight sum W of the connecting edges in the region is respectively calculated by adopting the strategy of graded weight reducing1，W2，…，Wncomparison W1，W2，…，Wnthe maximum weight and the corresponding sub-navigation band are selected as the starting point of the region growing, namely the optimal starting point to be searched.
8. the minimum spanning tree-based special unmanned aerial vehicle image air-to-air free network construction method according to claim 1, wherein in the second step of constructing the regional free network, a specific method for constructing the regional free network in an absolute orientation manner is as follows: firstly, solving seven absolute orientation elements according to a common point between adjacent sub-navigation bands, solving three offsets and a scale for eliminating rough differences of the seven absolute orientation elements, then solving three rotation angles, taking an object point coordinate system corresponding to an optimal starting point as a reference coordinate system of the whole area network, and sequentially bringing the camera stations and model point coordinates of other sub-navigation bands into a coordinate system unified with the starting sub-navigation band through an absolute orientation formula according to a generated area network connection mode to finish the unification of the whole area coordinate system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710747621.XA CN107560593B (en) | 2017-08-28 | 2017-08-28 | Special unmanned aerial vehicle image air-three free network construction method based on minimum spanning tree |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710747621.XA CN107560593B (en) | 2017-08-28 | 2017-08-28 | Special unmanned aerial vehicle image air-three free network construction method based on minimum spanning tree |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107560593A CN107560593A (en) | 2018-01-09 |
| CN107560593B true CN107560593B (en) | 2019-12-17 |
Family
ID=60977180
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710747621.XA Active CN107560593B (en) | 2017-08-28 | 2017-08-28 | Special unmanned aerial vehicle image air-three free network construction method based on minimum spanning tree |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107560593B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108761271A (en) * | 2018-03-30 | 2018-11-06 | 广州中科云图智能科技有限公司 | A kind of power grid screen of trees detection method and system |
| CN108600026B (en) * | 2018-05-07 | 2021-08-27 | 重庆邮电大学 | Neighbor relation storage method and device |
| CN109887026B (en) * | 2019-02-20 | 2021-07-16 | 深圳市未来感知科技有限公司 | Multi-view positioning tracking method, device and equipment and computer readable storage medium |
| CN110312085A (en) * | 2019-06-06 | 2019-10-08 | 武汉易科空间信息技术股份有限公司 | Image interfusion method and system based on multiple unmanned plane technologies |
| CN110853142A (en) * | 2019-11-20 | 2020-02-28 | 中国民航科学技术研究院 | Airport clearance three-dimensional model construction method and device based on unmanned aerial vehicle shooting |
| CN111244822B (en) * | 2020-03-20 | 2021-06-01 | 广东电网有限责任公司 | Fixed-wing unmanned aerial vehicle line patrol method, system and device in complex geographic environment |
| CN112269202A (en) * | 2020-10-15 | 2021-01-26 | 武汉大学 | Motion carrier assisted space reference transmission system and method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101226057A (en) * | 2008-02-01 | 2008-07-23 | 武汉朗视软件有限公司 | Digital close range photogrammetry method |
| CN102096816A (en) * | 2011-01-28 | 2011-06-15 | 武汉大学 | Multi-scale multi-level image segmentation method based on minimum spanning tree |
| CN102506824A (en) * | 2011-10-14 | 2012-06-20 | 航天恒星科技有限公司 | Method for generating digital orthophoto map (DOM) by urban low altitude unmanned aerial vehicle |
| CN105469061A (en) * | 2015-08-04 | 2016-04-06 | 电子科技大学中山学院 | Topographic feature line extraction method and device |
| CN106127782A (en) * | 2016-06-30 | 2016-11-16 | 北京奇艺世纪科技有限公司 | A kind of image partition method and system |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI413020B (en) * | 2008-12-31 | 2013-10-21 | Ind Tech Res Inst | Method and system for searching global minimum |
-
2017
- 2017-08-28 CN CN201710747621.XA patent/CN107560593B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101226057A (en) * | 2008-02-01 | 2008-07-23 | 武汉朗视软件有限公司 | Digital close range photogrammetry method |
| CN102096816A (en) * | 2011-01-28 | 2011-06-15 | 武汉大学 | Multi-scale multi-level image segmentation method based on minimum spanning tree |
| CN102506824A (en) * | 2011-10-14 | 2012-06-20 | 航天恒星科技有限公司 | Method for generating digital orthophoto map (DOM) by urban low altitude unmanned aerial vehicle |
| CN105469061A (en) * | 2015-08-04 | 2016-04-06 | 电子科技大学中山学院 | Topographic feature line extraction method and device |
| CN106127782A (en) * | 2016-06-30 | 2016-11-16 | 北京奇艺世纪科技有限公司 | A kind of image partition method and system |
Non-Patent Citations (1)
| Title |
|---|
| 《无人机遥感影像的城市绿地信息提取》;杨柳 等;《测绘科学》;20170228;第42卷(第2期);59-64 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107560593A (en) | 2018-01-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107560593B (en) | Special unmanned aerial vehicle image air-three free network construction method based on minimum spanning tree | |
| CN102506824B (en) | Method for generating digital orthophoto map (DOM) by urban low altitude unmanned aerial vehicle | |
| US7944547B2 (en) | Method and system of generating 3D images with airborne oblique/vertical imagery, GPS/IMU data, and LIDAR elevation data | |
| CN111597666B (en) | Method for applying BIM to transformer substation construction process | |
| US20230024326A1 (en) | Using maps comprising covariances in multi-resolution voxels | |
| Lo Brutto et al. | UAV platforms for cultural heritage survey: first results | |
| CN105783878A (en) | Small unmanned aerial vehicle remote sensing-based slope deformation detection and calculation method | |
| CN107504957A (en) | The method that three-dimensional terrain model structure is quickly carried out using unmanned plane multi-visual angle filming | |
| CN108168521A (en) | One kind realizes landscape three-dimensional visualization method based on unmanned plane | |
| Yu et al. | Modeling of landslide topography based on micro-unmanned aerial vehicle photography and structure-from-motion | |
| Rijsdijk et al. | Unmanned aerial systems in the process of juridical verification of cadastral border | |
| CN104360362B (en) | Method and system for positioning observed object via aircraft | |
| CN101008676A (en) | Method for measuring forest by unmanned aerial vehicle aerial photography remote sensing | |
| CN106940181B (en) | Unmanned aerial vehicle image control distribution network construction and aerial vehicle selectable range matching method | |
| Tournadre et al. | UAV linear photogrammetry | |
| CN103942828A (en) | Culture-heritage three-dimensional-scene generation system and method | |
| US11288861B2 (en) | Maps comprising covariances in multi-resolution voxels | |
| KR20120041819A (en) | Method for generating 3-d high resolution ndvi urban model | |
| CN112862966B (en) | Method, device, equipment and storage medium for constructing surface three-dimensional model | |
| JP2014106118A (en) | Digital surface model creation method, and digital surface model creation device | |
| CN110046563A (en) | A kind of transmission line of electricity measuring height of section modification method based on unmanned plane point cloud | |
| Zhou et al. | Application of UAV oblique photography in real scene 3d modeling | |
| JP6146731B2 (en) | Coordinate correction apparatus, coordinate correction program, and coordinate correction method | |
| CN114359489A (en) | Method, device and equipment for making real-scene image in pipeline construction period and storage medium | |
| TW201938991A (en) | Route planning method for aerial photography utilizing multi-axial unmanned aerial vehicle regulating the aerial photography height according to topography |
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 | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20211116 Address after: 214000 room 7-810, Aokai City Plaza, 1777 Zhonghui Avenue, Huishan District, Wuxi City, Jiangsu Province (Urban Railway Huishan station area) Patentee after: Wuxi Zike Technology Co., Ltd Address before: 448000 building c2-5, No. 201, Peigong Avenue, Duodao District, Jingmen City, Hubei Province (people's Wanfu entrepreneurship Park) Patentee before: Jingmen Chengyuan Electronic Technology Co., Ltd |