CN114972440B - Chained tracking method for ES database pattern spot objects for homeland investigation - Google Patents
Chained tracking method for ES database pattern spot objects for homeland investigation Download PDFInfo
- Publication number
- CN114972440B CN114972440B CN202210705310.8A CN202210705310A CN114972440B CN 114972440 B CN114972440 B CN 114972440B CN 202210705310 A CN202210705310 A CN 202210705310A CN 114972440 B CN114972440 B CN 114972440B
- Authority
- CN
- China
- Prior art keywords
- image
- node
- tracking
- remote sensing
- database
- 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
- 238000000034 method Methods 0.000 title claims abstract description 77
- 238000011835 investigation Methods 0.000 title claims abstract description 39
- 230000011218 segmentation Effects 0.000 claims abstract description 31
- 238000005516 engineering process Methods 0.000 claims abstract description 12
- 238000007781 pre-processing Methods 0.000 claims abstract description 10
- 239000013598 vector Substances 0.000 claims abstract description 10
- 238000001514 detection method Methods 0.000 claims description 49
- 238000004422 calculation algorithm Methods 0.000 claims description 35
- 238000011156 evaluation Methods 0.000 claims description 22
- 239000003550 marker Substances 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 15
- 238000010586 diagram Methods 0.000 claims description 13
- 238000012937 correction Methods 0.000 claims description 10
- 238000000605 extraction Methods 0.000 claims description 8
- 238000003709 image segmentation Methods 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 8
- 238000005457 optimization Methods 0.000 claims description 7
- 230000005855 radiation Effects 0.000 claims description 7
- 238000004458 analytical method Methods 0.000 claims description 6
- 230000009467 reduction Effects 0.000 claims description 6
- 238000012163 sequencing technique Methods 0.000 claims description 6
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 4
- 238000007476 Maximum Likelihood Methods 0.000 claims description 3
- 230000002457 bidirectional effect Effects 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 claims description 3
- 230000008030 elimination Effects 0.000 claims description 3
- 238000003379 elimination reaction Methods 0.000 claims description 3
- 238000003032 molecular docking Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000007789 sealing Methods 0.000 claims description 3
- 238000012795 verification Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 5
- 239000000463 material Substances 0.000 abstract description 4
- 239000002689 soil Substances 0.000 abstract description 4
- 230000007704 transition Effects 0.000 abstract description 4
- 230000003245 working effect Effects 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 10
- 238000004590 computer program Methods 0.000 description 7
- 230000003993 interaction Effects 0.000 description 6
- 238000012935 Averaging Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000002310 reflectometry Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/27—Replication, distribution or synchronisation of data between databases or within a distributed database system; Distributed database system architectures therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/29—Geographical information databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
- G06N5/041—Abduction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30241—Trajectory
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Remote Sensing (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of data processing, in particular to an ES database plaque object chain tracking method for homeland investigation, which comprises the steps of installing an ES and related plug-ins, acquiring spatial data of a region to be tracked through an inverted index of the ES, acquiring a plurality of remote sensing images of the same region in different periods, preprocessing, generating initial plaque objects of each remote sensing image, converting grid segmentation into plaque boundary vector images, obtaining tracking tracks of each plaque object by adopting a multi-target tracking technology, and the like. The design of the invention can rapidly and accurately extract the required data from massive data, improve the working efficiency and reduce the time consumption; the remote sensing images of the same area in different periods are processed, and the multi-target chain tracking technology is combined, so that the state soil resource transition condition can be accurately tracked, a large amount of manpower, material resources and time are not required to be consumed for investigation, the error occurrence condition is reduced, the data can be traced, and the working effect of state soil management is effectively improved.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to a chained tracking method for an ES database pattern spot object for homeland investigation.
Background
Homeland surveys are an important national policy. The land investigation work mainly comprises land utilization status investigation, land right investigation, special land investigation, basic farmland investigation, land utilization database construction and result summary archive arrangement and archiving. The investigation of the territory is generally carried out once every few years, the investigation of the territory has the advantages of wide range, fine and complicated work, large consumption of manpower, material resources and time, large workload, easy occurrence of work errors in the investigation process and difficult tracing and correction. With the continuous development of science and technology, the remote sensing image can display the land space condition more accurately. If the historical homeland resource data can be combined with the accurate remote sensing image in a certain period of time, the comparison and tracking of homeland resource transition can be rapidly carried out. However, there is no method for chain tracking of the plaque objects in the ES database available for homeland investigation.
Disclosure of Invention
The invention aims to provide a chained tracking method for an ES database plaque object for homeland investigation, so as to solve the problems in the background technology.
In order to solve the technical problems, one of the purposes of the present invention is to provide an ES database plaque object chain tracking method for homeland investigation, which is characterized by comprising an ES database module, a data acquisition module, a data preprocessing module, a plaque generation module, an object chain tracking analysis module and a model evaluation module, wherein for the ES database module, an ES is installed and a service is started first, and then a plurality of related third party plug-ins are installed; butting against a domestic resource database, and acquiring space data of a region to be tracked through an ES distributed inverted index; acquiring a plurality of remote sensing images of the same region to be tracked in different periods, and carrying out pretreatment such as noise reduction, correction, registration and the like on all the remote sensing images; generating an initial image spot object of each remote sensing image by adopting a topology heuristic image segmentation algorithm based on a multi-resolution topology network; grid segmentation is carried out on the remote sensing image by adopting a segmentation algorithm of mean shift to obtain a plaque object and boundary information of the remote sensing image, and the boundary information is combined and converted into a plaque boundary vector diagram; and overlapping the grid type space data of the area to be tracked and a plurality of remote sensing images corresponding to the space data according to time sequence, adopting a multi-target tracking technology to carry out chained tracking on the image spot objects of the area image, and finally obtaining the tracking track of each image spot object.
Furthermore, the ES database module is also required to be installed with ES and related plug-ins, and the specific method comprises the following steps:
(1) Installing Java;
(2) Creating a common user;
(3) Installing an ES, and placing all files chowns under an ES user group;
(4) Starting an ES service, switching to an ES user, starting in a background operation mode, checking an ES default port number and an ES process after execution, and verifying through a web, wherein the output of a node name, a cluster name and an ES version represents successful starting;
(5) Looking up the directory structure of the ES;
(6) And installing a third-party plug-in, including a word segmentation plug-in, a synchronization plug-in, a data transmission plug-in, a script plug-in, a site plug-in and other plug-ins.
Further, the directory structure of the ES in the ES database module mainly includes: bin, mainly a start file, a configuration script and an ES plug-in instruction; config, mainly configuration files, such as cluster names, node names, port numbers, etc.; data, which is a Data directory of the ES, organizing a directory structure according to the cluster-nodes; logs, a library used by ESs, mainly jar packets; plugins are mainly ES plug-ins already installed, such as the Head plug-in.
Further, the data acquisition module is configured to acquire spatial data of an area to be tracked, and the specific method includes the following steps:
(1) Docking the ES with a homeland resource database to enable the ES to access a traditional database of a homeland resource platform;
(2) Searching matched space data in a database according to the keywords by adopting an inverted index mode;
(3) Based on the grid model, the space is divided into regular grids, corresponding attribute values are given on each grid to represent geographic entities, and the space data are all converted into a grid data structure.
Further, the specific method for preprocessing the remote sensing image by the data preprocessing module comprises the following steps:
(1) Acquiring a plurality of remote sensing images in the same area in different periods, and sequencing the images according to time;
(2) Selecting the remote sensing image at the earliest time as a reference image, and using the subsequent remote sensing image as a detection image;
(3) And (3) noise reduction treatment: the noise of the image exists in the high-frequency part of the image, the high frequency and the low frequency of the image are separated, the image noise elimination is carried out by utilizing wavelet change, and the irrelevant information mixed in the image is removed;
(4) Radiation correction: adopting a statistical regression method, taking the reference image as a main image, and carrying out radiation correction on each detection image;
(5) Image registration: and (5) realizing image registration by adopting an affine invariant feature extraction algorithm.
Further, the image spot generating module is configured to generate an initial image spot object of each remote sensing image, and the specific method includes: firstly, carrying out heterogeneity criterion of feature vectors from two aspects of statistical features and geometric features; adopting a bidirectional minimum heterogeneity condition to find a pair of optimal object combinations; managing the spatial relationship among objects under different scales by adopting a hierarchical tree index structure comprising three key technologies of a hierarchical tree node data structure, a query adjacency topological relation, a scale stride and a topological network; and finally, sequentially dividing each remote sensing image by adopting a topological heuristic image segmentation method, and generating a final initial image spot object.
Further, the map spot generation module provides a generalized probability inference model, which comprises the following steps: starting from a single pixel, searching a local optimal segmentation area pair to segment a remote sensing image in a heuristic search mode; for an input high-resolution image I, a position set is defined as s= { (I, j) }, y= { Y s S represents the observed data of the image, where S represents one primitive, S represents the set of primitives, y s Representing the image characteristics of s; definition x= { X s S is the marker field, where X s Is a marker of primitive s, and X s E {1, 2..the., k }, k being the number of segmentation categories, assuming that the marker field X has markov random properties, i.e.:
P(X s |X t ,t∈S,t≠s)=P(X S |X t ,t∈N s )
where P (·) represents the probability value of the joint distribution, N s Representing all neighborhood sets adjacent to primitive s, define x= { x s S is all realizations of the marker field X, Ω is the set of all X, the segmentation resultIs the implementation that maximizes the posterior probability, namely: />Then, the mixed Gaussian distribution is adopted to model the observation characteristic field Y of the image, meanwhile, the homogeneous region in the likelihood function is assumed to have the same distribution, namely, the region of the same type in the image is assumed to obey the same distribution, and at the moment, the region is->g e {1, 2..k }, k is the number of segmentation categories, there are:
where P (y=y|x=x) is the posterior probability of the marker field x=x given the observed data y=y, μ g Representing characteristic mean value, Σ g Representing a feature covariance matrix, p being the dimension, the feature mean μ is to be estimated g And characteristic variance sigma g Estimating the two parameters by using a maximum likelihood method, wherein the characteristic average value mu of each category g And a feature covariance matrix Σ g The estimation results of (2) are as follows:
and then constructing an initial hierarchical index tree node, merging areas, maintaining dynamic update of the multi-resolution topological network, iteratively executing the process of searching the merged object sequence, creating a tree node of the upper layer step by step until the segmentation is finished, and outputting the vectorized image spot object.
Further, carrying out vectorization on the boundaries of the image spots by adopting grid segmentation, carrying out segmentation on the remote sensing image to obtain a segmentation image, extracting each image spot on the basis, then taking the image spot as a unit to extract all the contained pixels, and recording the coordinates of each pixel and the coordinates of the corner points of the pixel 4; defining pixel corner points positioned inside the boundary polygon as inner pixel corner points, removing vertexes of non-boundary polygons, and defining pixel corner point pairs which are adjacent in position and not belonging to the same effective pixel as vertexes of the boundary polygon as pseudo-adjacent pixel corner points so as to determine a real pixel corner point set to be selected; then finding out a first effective pixel of the top line of the image spot according to the pixel coordinate value, setting the upper left corner of the pixel as a starting vertex for tracking the boundary polygon, taking the upper right corner of the pixel as a second vertex of the boundary polygon, and tracking and searching the rest boundary vertices along the boundary of the image spot according to the vertex searching rule; and finally, sequentially connecting the searched vertexes in sequence to form a required boundary polygon and sealing the boundary polygon to obtain a boundary vector diagram of the image spots.
Furthermore, the object chain tracking analysis module adopts a CTracker tracking algorithm to construct a chain tracking frame based on two frames of input, realizes end-to-end joint detection tracking, and fuses target detection, feature extraction and target association together for global optimization; sequentially and simultaneously inputting two adjacent frames of remote sensing images into a network, wherein the two adjacent frames are called a node, and the network outputs all calibration frame pairs in the node, and each detection frame pair comprises two detection frames of the same target in front and rear frames, so that the detection frames of the target in the two adjacent frames and the association relation of the detection frames are obtained; for two adjacent nodes, the rear frame of the front node and the front frame of the rear node are the same frame, in the common frame, the detection frame output by the front node is basically consistent with the detection frame output by the rear node, then a simple cross-over algorithm is adopted for matching, the first node is connected to the last node, then a complete target tracking track is obtained, the target track is assumed to contain N frames of pictures, and the picture sequence is recorded as from the initial time t=1Recording the real calibration frame set as +.>The tag set is +.>The prediction calibration frame set isK t Representing the number of calibration boxes at time t, the CTracker model requires two adjacent frames as inputs, called chain nodes, the first one denoted (F 1 ,F 2 ) The last chain node is denoted (F N ,F N+1 ) The given chain node at time t=1 is denoted (F t-1 ,F t ) The prediction calibration frame pair generated by the CTracker tracking algorithm is marked as +.>Wherein n is t-1 Representing the total logarithm of the calibration frame at time t-1, the next adjacent chain node (F t ,F t+1 ) Is +.>In order to enhance the robustness of the CTracker tracking algorithm, the track of target termination and the identification thereof are reserved, and assuming that delta frames are reserved at most, the target prediction calibration frame pair is marked as +.>The calibration box at time t+τ is marked +.> If the CTracker tracking algorithm matches the calibration box +.>The target tracking trajectory will be extended by the chain node to a new calibration frame for the link point at time t (F t ,F t+1 ) Let-> Representing the content of a calibration frame of an ith node at the moment t, wherein the content comprises the abscissa of the central point of the calibration frame of the ith node at the moment t, the ordinate of the central point of the calibration frame of the ith node at the moment t, the length of the calibration frame of the ith node at the moment t, the width of the calibration frame of the ith node at the moment t, and M is used for representing a matching result, M= {0,1}, and if the ith node can be matched with the calibration frame of the jth node, M is recorded ij =1, otherwise, note M ij =0, the ith node true value tag is +.>The method is characterized by comprising the following steps:
the identity verification tag of the i-th node is marked asThe method is characterized by comprising the following steps:
prediction calibration frame for ith node at t momentNaturally also contains the content of the calibration frame, noted asLet->Representing the offset of the predicted calibration frame of the ith node at time t,representing the offset of the real calibration frame of the j-th node at the time t, for the offset content of the predicted calibration frame, there is: />And satisfy-> The loss function is defined as:
the total loss function of the CTracker tracking algorithm is noted as:
wherein alpha and beta are weight factors, +.>And->The confidence of the prediction for the ith calibration frame.
Further, the model evaluation module optimizes each image by adopting a depth evaluation method in the process of sequencing each image according to time sequence, and specifically comprises an average accuracy evaluation and a global average accuracy evaluation, wherein the average accuracy AP evaluation formula is as follows:
wherein, p (i) is the proportion of related images in the previous i time sequences, r (i) is a binary function, the value is 0 or 1, the i-th image is related to the reference image, r (i) =1, otherwise, r (i) =0, and Uni is a normalization constant;
the formula of the global average accuracy MAP evaluation is as follows:
wherein T is n The image coefficient of the nth detected image is the number of times NC is the number of times of detection.
The second object of the present invention is to provide an operation system and apparatus for the chained tracking method of the ES database pattern object for the homeland investigation, which comprises a processor, a memory and a computer program stored in the memory and running on the processor, wherein the processor is used for implementing any one of the steps of the chained tracking method of the ES database pattern object for the homeland investigation when executing the computer program.
It is a further object of the present invention to provide a computer-readable storage medium storing a computer program which, when executed by a processor, implements the steps of any of the above-described ES database plaque object chain tracking methods for homeland investigation.
Compared with the prior art, the invention has the beneficial effects that:
1. the chained tracking method for the map spots of the ES database for homeland investigation can extract required data from massive data rapidly and accurately by adopting an inverted index searching mode based on the ES database, thereby improving the working efficiency and reducing the time consumption;
2. according to the method for chain tracking of the map spot object of the ES database for homeland investigation, the remote sensing images of the same area in different periods are segmented, the map spot object is generated, the boundary vectorization is carried out on the map spot, and the multi-target chain tracking technology is combined, so that the transition condition of homeland resources can be accurately tracked, a large amount of manpower, material resources and time are not required to be consumed for investigation, the occurrence of error conditions is reduced, all data can be traced, and the working effect of homeland management is effectively improved;
3. the method for chain tracking the image spot object of the ES database for homeland investigation provides a generalized probability inference model, and the remote sensing image segmentation of anisotropic interactions among different categories provides a method, and the model not only can consider spatial interactions, but also can consider category interactions of information expected values;
4. the chained tracking method for the map spot object of the ES database for homeland investigation adopts a CTracker tracking algorithm, builds a chained tracking frame based on two frames of input, realizes end-to-end joint detection tracking, and fuses target detection, feature extraction and target association together for global optimization; sequentially and simultaneously inputting two adjacent frames of remote sensing images into a network, wherein the two adjacent frames are called a node, and the network outputs all calibration frame pairs in the node, and each detection frame pair comprises two detection frames of the same target in front and rear frames, so that the detection frames of the target in the two adjacent frames and the association relation of the detection frames are obtained; for two adjacent nodes, the rear frame of the front node and the front frame of the rear node are the same frame, in the common frame, the detection frame output by the front node is basically consistent with the detection frame output by the rear node, then a simple cross-over algorithm is adopted for matching, a complete target tracking track is obtained from the first node to the last node, the robustness of the CTracker tracking algorithm is improved by keeping the track of the target termination and the identification thereof, and when the object is in the target appearance state, the CTracker tracking algorithm can effectively track and predict the calibration frame of the object, and finally the neural network weight matrix with the minimum loss function is obtained through iteration, averaging and weighting processing.
Drawings
FIG. 1 is a block diagram of the overall process of the present invention;
FIG. 2 is a block diagram of a partial process flow of the present invention;
FIG. 3 is a second flow chart of the partial method of the present invention;
FIG. 4 is a third flow chart of the local method of the present invention;
FIG. 5 is a fourth block diagram of a partial process flow of the present invention;
FIG. 6 is a fifth block diagram of a partial process flow of the present invention;
FIG. 7 is a sixth flow chart of a partial method of the present invention;
FIG. 8 is a block diagram of a partial process flow diagram of the present invention;
FIG. 9 is a schematic diagram of an exemplary electronic computer product architecture of the present invention;
FIG. 10 is a schematic view of the structure of the present invention;
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1
As shown in fig. 1 to 9, an objective of the present embodiment is to provide an ES database plaque object chain tracking method for homeland investigation, which is characterized by comprising an ES database module, a data acquisition module, a data preprocessing module, a plaque generation module, an object chain tracking analysis module and a model evaluation module, wherein for the ES database module, an ES is installed and a service is started first, and then a plurality of related third party plug-ins are installed; butting against a domestic resource database, and acquiring space data of a region to be tracked through an ES distributed inverted index; acquiring a plurality of remote sensing images of the same region to be tracked in different periods, and carrying out pretreatment such as noise reduction, correction, registration and the like on all the remote sensing images; generating an initial image spot object of each remote sensing image by adopting a topology heuristic image segmentation algorithm based on a multi-resolution topology network; grid segmentation is carried out on the remote sensing image by adopting a segmentation algorithm of mean shift to obtain a plaque object and boundary information of the remote sensing image, and the boundary information is combined and converted into a plaque boundary vector diagram; and overlapping the grid type space data of the area to be tracked and a plurality of remote sensing images corresponding to the space data according to time sequence, adopting a multi-target tracking technology to carry out chained tracking on the image spot objects of the area image, and finally obtaining the tracking track of each image spot object.
Specifically, the ES database module further needs to install an ES and related plug-ins, and the specific method includes the following steps:
(1) Installing Java;
(2) Creating a common user;
(3) Installing an ES, and placing all files chowns under an ES user group;
(4) Starting an ES service, switching to an ES user, starting in a background operation mode, checking an ES default port number and an ES process after execution, and verifying through a web, wherein the output of a node name, a cluster name and an ES version represents successful starting;
(5) Looking up the directory structure of the ES;
(6) And installing a third-party plug-in, including a word segmentation plug-in, a synchronization plug-in, a data transmission plug-in, a script plug-in, a site plug-in and other plug-ins.
Specifically, the directory structure of the ES in the ES database module mainly includes: bin, mainly a start file, a configuration script and an ES plug-in instruction; config, mainly configuration files, such as cluster names, node names, port numbers, etc.; data, which is a Data directory of the ES, organizing a directory structure according to the cluster-nodes; logs, a library used by ESs, mainly jar packets; plugins are mainly ES plug-ins already installed, such as the Head plug-in.
Specifically, the data acquisition module is configured to acquire spatial data of an area to be tracked, and the specific method includes the following steps:
(1) Docking the ES with a homeland resource database to enable the ES to access a traditional database of a homeland resource platform;
(2) Searching matched space data in a database according to the keywords by adopting an inverted index mode;
(3) Based on the grid model, the space is divided into regular grids, corresponding attribute values are given on each grid to represent geographic entities, and the space data are all converted into a grid data structure.
Further, the specific method for preprocessing the remote sensing image by the data preprocessing module comprises the following steps:
(1) Acquiring a plurality of remote sensing images in the same area in different periods, and sequencing the images according to time;
(2) Selecting the remote sensing image at the earliest time as a reference image, and using the subsequent remote sensing image as a detection image;
(3) And (3) noise reduction treatment: the noise of the image exists in the high-frequency part of the image, the high frequency and the low frequency of the image are separated, the image noise elimination is carried out by utilizing wavelet change, and the irrelevant information mixed in the image is removed;
(4) Radiation correction: adopting a statistical regression method, carrying out radiation correction on each detection image by taking a reference image as a main image, finding out a ground object sample point which does not generate ground object type change and has stable spectral property at the same place in two-time-phase images by using a linear correlation of gray values of the ground object sample point, and correcting the ground object sample point by using a linear correlation of gray values of the ground object sample point, wherein regularized ground object reflectivity data can be obtained without the radiation calibration of a sensor and related atmospheric parameters;
(5) Image registration: the affine invariant feature extraction algorithm is adopted to realize image registration, and the specific method for image registration comprises the following steps: firstly, constructing a SIFT descriptor with affine invariance, and matching the extracted control points by using the descriptor to obtain transformation parameters to realize image registration.
Specifically, the image spot generating module is used for generating an initial image spot object of each remote sensing image, and the specific method comprises the following steps: firstly, carrying out heterogeneity criterion of feature vectors from two aspects of statistical features and geometric features; adopting a bidirectional minimum heterogeneity condition to find a pair of optimal object combinations; managing the spatial relationship among objects under different scales by adopting a hierarchical tree index structure comprising three key technologies of a hierarchical tree node data structure, a query adjacency topological relation, a scale stride and a topological network; and finally, sequentially segmenting each remote sensing image by adopting a topological heuristic image segmentation method, and generating a final initial image spot object, wherein in S4.1, the statistical characteristics mainly comprise mean values, standard deviations and information entropy, and the geometric characteristics mainly comprise density and asymmetry.
Preferably, the map spot generating module proposes a generalized probability inference model, including the following steps: starting from a single pixel, searching a local optimal segmentation area pair to segment a remote sensing image in a heuristic search mode; for an input high-resolution image I, a position set is defined as s= { (I, j) }, y= { Y s S represents the observed data of the image, where S represents one primitive, S represents the set of primitives, y s Representing the image characteristics of s; definition x= { X s S is the marker field, where X s Is a marker of primitive s, and X s E {1, 2..the., k }, k being the number of segmentation categories, assuming that the marker field X has markov random properties, i.e.:
P(X s |X t ,t∈S,t≠s)=P(X S |X t ,t∈N s )
where P (·) represents the probability value of the joint distribution, N s Representing all neighborhood sets adjacent to primitive s, define x= { x s S is all realizations of the marker field X, Ω is the set of all X, the segmentation resultIs the implementation that maximizes the posterior probability, i.e.::>then, the mixed Gaussian distribution is adopted to model the observation characteristic field Y of the image, meanwhile, the homogeneous region in the likelihood function is assumed to have the same distribution, namely, the region of the same type in the image is assumed to obey the same distribution, and at the moment, the region is->g e {1, 2..k }, k is the number of segmentation categories, there are:
where P (y=y|x=x) is the posterior probability of the marker field x=x given the observed data y=y, μ g Representing characteristic mean value, Σ g Representing a feature covariance matrix, p being the dimension, the feature mean μ is to be estimated g And characteristic variance sigma g Estimating the two parameters by using a maximum likelihood method, wherein the characteristic average value mu of each category g And a feature covariance matrix Σ g The estimation results of (2) are as follows:
and then constructing an initial hierarchical index tree node, merging areas, maintaining dynamic update of the multi-resolution topological network, iteratively executing the process of searching the merged object sequence, creating a tree node of the upper layer step by step until the segmentation is finished, and outputting the vectorized image spot object.
Specifically, grid segmentation is adopted to vectorize the boundaries of the image spots, the remote sensing image is required to be segmented to obtain a segmentation image, each image spot is extracted on the basis, all the contained pixels are extracted by taking the image spot as a unit, and the coordinates of the pixels and the coordinates of the corner points of the pixels 4 are recorded for each pixel; defining pixel corner points positioned inside the boundary polygon as inner pixel corner points, removing vertexes of non-boundary polygons, and defining pixel corner point pairs which are adjacent in position and not belonging to the same effective pixel as vertexes of the boundary polygon as pseudo-adjacent pixel corner points so as to determine a real pixel corner point set to be selected; then finding out a first effective pixel of the top line of the image spot according to the pixel coordinate value, setting the upper left corner of the pixel as a starting vertex for tracking the boundary polygon, taking the upper right corner of the pixel as a second vertex of the boundary polygon, and tracking and searching the rest boundary vertices along the boundary of the image spot according to the vertex searching rule; and finally, sequentially connecting the searched vertexes in sequence to form a required boundary polygon and sealing the boundary polygon to obtain a boundary vector diagram of the image spots.
Preferably, the object chain tracking analysis module adopts a CTracker tracking algorithm to construct a chain tracking frame based on two frames of input, realizes end-to-end joint detection tracking, and fuses target detection, feature extraction and target association together for global optimization; sequentially and simultaneously inputting two adjacent frames of remote sensing images into a network, wherein the two adjacent frames are called a node, and the network outputs all calibration frame pairs in the node, and each detection frame pair comprises two detection frames of the same target in front and rear frames, so that the detection frames of the target in the two adjacent frames and the association relation of the detection frames are obtained; for two adjacent nodes, the rear frame of the front node and the front frame of the rear node are the same frame, in the common frame, the detection frame output by the front node is basically consistent with the detection frame output by the rear node, then a simple cross-over algorithm is adopted for matching, the first node is connected to the last node, then a complete target tracking track is obtained, the target track is assumed to contain N frames of pictures, and the picture sequence is recorded as from the initial time t=1Recording the real calibration frame set as +.>The tag set is +.>The prediction calibration frame set isK t Representing the number of calibration boxes at time t, the CTracker model requires two adjacent frames as inputs, called chain nodes, the first one denoted (F 1 ,F 2 ) The last chain node is denoted (F N ,F N+1 ) The given chain node at time t=1 is denoted (F t-1 ,F t ) The prediction calibration frame pair generated by the CTracker tracking algorithm is marked as +.>Wherein n is t-1 Representing the total logarithm of the calibration frame at time t-1, the next adjacent chain node (F t ,F t+1 ) Is +.>In order to enhance the robustness of the CTracker tracking algorithm, the track of target termination and the identification thereof are reserved, and assuming that delta frames are reserved at most, the target prediction calibration frame pair is marked as +.>The calibration box at time t+τ is marked +.> If the CTracker tracking algorithm matches the calibration box +.>The target tracking trajectory will be extended by the chain node to a new calibration frame for the link point at time t (F t ,F t+1 ) Let-> Representing the content of a calibration frame of an ith node at the moment t, wherein the content comprises the abscissa of the central point of the calibration frame of the ith node at the moment t, the ordinate of the central point of the calibration frame of the ith node at the moment t, the length of the calibration frame of the ith node at the moment t, the width of the calibration frame of the ith node at the moment t, and M is used for representing a matching result, M= {0,1}, and if the ith node can be matched with the calibration frame of the jth node, M is recorded ij =1, otherwise, note M ij =0, the ith node true value tag is +.>The method is characterized by comprising the following steps:
the identity verification tag of the i-th node is marked asThe method is characterized by comprising the following steps:
prediction calibration frame for ith node at t momentNaturally also contains the content of the calibration frame, noted asLet->Representing the offset of the predicted calibration frame of the ith node at time t,representing the offset of the real calibration frame of the j-th node at the time t, for the offset content of the predicted calibration frame, there is: />And satisfy-> The loss function is defined as:
the total loss function of the CTracker tracking algorithm is noted as:
wherein alpha and beta are weight factors, +.>And->The confidence of the prediction for the ith calibration frame.
Specifically, the model evaluation module optimizes each image by adopting a depth evaluation method in the process of sequencing each image according to time sequence, specifically comprises average accuracy evaluation and global average accuracy evaluation, wherein the formula of the average accuracy AP evaluation is as follows:
wherein, p (i) is the proportion of related images in the previous i time sequences, r (i) is a binary function, the value is 0 or 1, the i-th image is related to the reference image, r (i) =1, otherwise, r (i) =0, and Uni is a normalization constant;
the formula of the global average accuracy MAP evaluation is as follows:
wherein T is n The image coefficient of the nth detected image is the number of times NC is the number of times of detection.
After the depth evaluation method is utilized for optimization, the time sequence provides further sequencing accuracy for the detected images.
The invention has the beneficial effects that:
the chained tracking method for the map spots of the ES database for homeland investigation can extract required data from massive data rapidly and accurately by adopting an inverted index searching mode based on the ES database, thereby improving the working efficiency and reducing the time consumption; the remote sensing images of the same area in different periods are segmented, image spot objects are generated, boundary vectorization is carried out on the image spots, and the multi-target chain tracking technology is combined, so that the state soil resource transition condition can be accurately tracked, a large amount of manpower, material resources and time are not required to be consumed for investigation, the error occurrence condition is reduced, all data can be traced, and the working effect of state soil management is effectively improved; a generalized probability inference model is provided, and a method is provided for remote sensing image segmentation by anisotropic interaction among different categories, wherein the model can consider not only spatial interaction, but also category interaction of information expected values; adopting a CTracker tracking algorithm to construct a chain tracking frame based on two frames of input, realizing end-to-end joint detection tracking, and fusing target detection, feature extraction and target association together for global optimization; sequentially and simultaneously inputting two adjacent frames of remote sensing images into a network, wherein the two adjacent frames are called a node, and the network outputs all calibration frame pairs in the node, and each detection frame pair comprises two detection frames of the same target in front and rear frames, so that the detection frames of the target in the two adjacent frames and the association relation of the detection frames are obtained; for two adjacent nodes, the rear frame of the front node and the front frame of the rear node are the same frame, in the common frame, the detection frame output by the front node is basically consistent with the detection frame output by the rear node, then a simple cross-over algorithm is adopted for matching, a complete target tracking track is obtained from the first node to the last node, the robustness of the CTracker tracking algorithm is improved by keeping the track of the target termination and the identification thereof, and when the object is in the target appearance state, the CTracker tracking algorithm can effectively track and predict the calibration frame of the object, and finally the neural network weight matrix with the minimum loss function is obtained through iteration, averaging and weighting processing.
As shown in fig. 9, the present embodiment also provides an operation system and an apparatus for the ES database plaque object chain tracking method for homeland investigation, where the apparatus includes a processor, a memory, and a computer program stored in the memory and running on the processor.
The processor comprises one or more than one processing core, the processor is connected with the memory through a bus, the memory is used for storing program instructions, and the steps of the ES database plaque object chain tracking method for homeland investigation are realized when the processor executes the program instructions in the memory.
Alternatively, the memory may be implemented by any type or combination of volatile or nonvolatile memory devices such as Static Random Access Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic or optical disk.
In addition, the invention also provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program realizes the steps of the ES database plaque object chain tracking method for territorial investigation when being executed by a processor.
Optionally, the present invention also provides a computer program product comprising instructions which, when run on a computer, cause the computer to perform the steps of the ES database plaque object chain tracking method of the above aspects for homeland investigation.
It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by hardware related to a program, and the program may be stored in a computer readable storage medium, where the above storage medium may be a read only memory, a magnetic disk or an optical disk, etc.
The foregoing has shown and described the basic principles, principal features and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the above-described embodiments, and that the above-described embodiments and descriptions are only preferred embodiments of the present invention, and are not intended to limit the invention, and that various changes and modifications may be made therein without departing from the spirit and scope of the invention as claimed. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (9)
1. The method is characterized by comprising an ES database module, a data acquisition module, a data preprocessing module, a pattern generation module, an object chain tracking analysis module and a model evaluation module, wherein for the ES database module, an ES is installed and started, and then a plurality of related third-party plug-ins are installed; butting against a domestic resource database, and acquiring space data of a region to be tracked through an ES distributed inverted index; acquiring a plurality of remote sensing images of the same region to be tracked in different periods, and carrying out noise reduction, correction and registration pretreatment on all the remote sensing images; generating an initial image spot object of each remote sensing image by adopting a topology heuristic image segmentation algorithm based on a multi-resolution topology network; grid segmentation is carried out on the remote sensing image by adopting a segmentation algorithm of mean shift to obtain a plaque object and boundary information of the remote sensing image, and the boundary information is combined and converted into a plaque boundary vector diagram; overlapping the grid type space data of the area to be tracked and a plurality of remote sensing images corresponding to the space data according to time sequence, adopting a multi-target tracking technology to carry out chained tracking on the image spot objects of the area image, and finally obtaining the tracking track of each image spot object;
the object chain tracking analysis module adopts a CTracker tracking algorithm to construct a chain tracking frame based on two frames of input, realizes end-to-end joint detection tracking, and fuses target detection, feature extraction and target association together for global optimization; sequentially and simultaneously inputting two adjacent frames of remote sensing images into a network, wherein the two adjacent frames are called a node, all calibration frame pairs in the network output node, and each detection frame pair comprises two detection frames of the same target in front and back framesThereby obtaining the detection frames of the target in two adjacent frames and the association relation of the detection frames; for two adjacent nodes, the rear frame of the front node and the front frame of the rear node are the same frame, in the common frame, the detection frame output by the front node is basically consistent with the detection frame output by the rear node, then a simple cross-over algorithm is adopted for matching, the first node is connected to the last node, then a complete target tracking track is obtained, the target track is assumed to contain N frames of pictures, and the picture sequence is recorded as from the initial time t=1Recording the real calibration frame set as +.>The tag set is +.>Prediction calibration frame set is +.>K t Representing the number of calibration boxes at time t, the CTracker model requires two adjacent frames as inputs, called chain nodes, the first one denoted (F 1 ,F 2 ) The last chain node is denoted (F N ,F N+1 ) The given chain node at time t=1 is denoted (F t-1 ,F t ) The prediction calibration frame pair generated by the CTracker tracking algorithm is marked as +.>Wherein n is t-1 Representing the total logarithm of the calibration frame at time t-1, the next adjacent chain node (F t ,F t+1 ) Is +.>To enhance the robustness of the CTracker tracking algorithm, the target termination trajectory and its identity are preserved, assuming maximum preservationLeaving delta frame, and marking target prediction calibration frame pair as +.>The calibration box at time t+τ is marked +.>If the CTracker tracking algorithm matches the calibration box +.>The target tracking trajectory will be extended by the chain node to a new calibration frame for the link point at time t (F t ,F t+1 ) Let-> Representing the content of a calibration frame of an ith node at the moment t, wherein the content comprises the abscissa of the central point of the calibration frame of the ith node at the moment t, the ordinate of the central point of the calibration frame of the ith node at the moment t, the length of the calibration frame of the ith node at the moment t, the width of the calibration frame of the ith node at the moment t, and M is used for representing a matching result, M= {0,1}, and if the ith node can be matched with the calibration frame of the jth node, M is recorded ij =1, otherwise, note M ij =0, the ith node true value tag is +.>The method is characterized by comprising the following steps:
the identity verification tag of the i-th node is marked asThe method is characterized by comprising the following steps:
prediction calibration frame for ith node at t momentNaturally also contains the content of the calibration frame, noted asLet->Representing the offset of the predicted calibration frame of the ith node at time t,representing the offset of the real calibration frame of the j-th node at the time t, for the offset content of the predicted calibration frame, there is: />And satisfy-> The loss function is defined as:
the total loss function of the CTracker tracking algorithm is noted as:
wherein alpha and beta are weight factors, +.>And->The confidence of the prediction for the ith calibration frame.
2. The chained tracking method of the objects of the pattern spots of the ES database for homeland investigation according to claim 1, wherein the ES database module is also required to install ES and related plug-ins, and the method comprises the following steps:
(1) Installing Java;
(2) Creating a common user;
(3) Installing an ES, and placing all files chowns under an ES user group;
(4) Starting an ES service, switching to an ES user, starting in a background operation mode, checking an ES default port number and an ES process after execution, and verifying through a web, wherein the output of a node name, a cluster name and an ES version represents successful starting;
(5) Looking up the directory structure of the ES;
(6) And installing a third-party plug-in, including a word segmentation plug-in, a synchronization plug-in, a data transmission plug-in, a script plug-in, a site plug-in and other plug-ins.
3. The method for chained tracking objects of ES database pattern spots for homeland investigation according to claim 2, wherein the directory structure of ES in the ES database module mainly comprises: bin, mainly a start file, a configuration script and an ES plug-in instruction; config, mainly configuration files, such as cluster name, node name, port number; data, which is a Data directory of the ES, organizing a directory structure according to the cluster-nodes; logs, a library used by ESs, mainly jar packets; plugins are mainly ES plug-ins already installed, such as the Head plug-in.
4. The chained tracking method of the ES database plaque object for homeland investigation according to claim 2, wherein the data acquisition module is configured to acquire spatial data of the area to be tracked, and the specific method comprises the following steps:
(1) Docking the ES with a homeland resource database to enable the ES to access a traditional database of a homeland resource platform;
(2) Searching matched space data in a database according to the keywords by adopting an inverted index mode;
(3) Based on the grid model, the space is divided into regular grids, corresponding attribute values are given on each grid to represent geographic entities, and the space data are all converted into a grid data structure.
5. The chained tracking method of the ES database plaque object for homeland investigation according to claim 3, wherein the specific method for preprocessing the remote sensing image by the data preprocessing module comprises the following steps:
(1) Acquiring a plurality of remote sensing images in the same area in different periods, and sequencing the images according to time;
(2) Selecting the remote sensing image at the earliest time as a reference image, and using the subsequent remote sensing image as a detection image;
(3) And (3) noise reduction treatment: the noise of the image exists in the high-frequency part of the image, the high frequency and the low frequency of the image are separated, the image noise elimination is carried out by utilizing wavelet change, and the irrelevant information mixed in the image is removed;
(4) Radiation correction: adopting a statistical regression method, taking the reference image as a main image, and carrying out radiation correction on each detection image;
(5) Image registration: and (5) realizing image registration by adopting an affine invariant feature extraction algorithm.
6. The method for chained tracking of ES database plaque objects for homeland investigation according to claim 4, wherein the plaque generating module is configured to generate an initial plaque object for each remote sensing image, and the method comprises: firstly, carrying out heterogeneity criterion of feature vectors from two aspects of statistical features and geometric features; adopting a bidirectional minimum heterogeneity condition to find a pair of optimal object combinations; managing the spatial relationship among objects under different scales by adopting a hierarchical tree index structure comprising three key technologies of a hierarchical tree node data structure, a query adjacency topological relation, a scale stride and a topological network; and finally, sequentially dividing each remote sensing image by adopting a topological heuristic image segmentation method, and generating a final initial image spot object.
7. The method for chained tracking of objects of pattern spots in ES database for homeland investigation according to claim 6, wherein the pattern spot generating module proposes a generalized probability inference model, comprising the steps of: starting from a single pixel, searching a local optimal segmentation area pair to segment a remote sensing image in a heuristic search mode; for an input high-resolution image I, a position set is defined as s= { (I, j) }, y= { Y s S represents the observed data of the image, where S represents one primitive, S represents the set of primitives, y s Representing the image characteristics of s; definition x= { X s S is the marker field, where X s Is a marker of primitive s, and X s E {1,2, …, k }, k being the number of segmentation categories, assuming that the marker field X has markov random properties, namely:
P(X s |X t ,t∈S,t≠s)=P(X S |X t ,t∈N s )
where P (·) represents the probability value of the joint distribution, N s Representing all neighborhood sets adjacent to primitive s, define x= { x s S is all realizations of the marker field X, Ω is the set of all X, the segmentation resultIs the implementation that maximizes the posterior probability, namely:then, the mixed Gaussian distribution is adopted to model the observation characteristic field Y of the image, meanwhile, the homogeneous region in the likelihood function is assumed to have the same distribution, namely, the region of the same type in the image is assumed to obey the same distribution, and at the moment, the region is->k is the number of divided categories, and there are:
where P (y=y|x=x) is the posterior probability of the marker field x=x given the observed data y=y, μ g Representing characteristic mean value, Σ g Representing a feature covariance matrix, p being the dimension, the feature mean μ is to be estimated g And characteristic variance sigma g Estimating the two parameters by using a maximum likelihood method, wherein the characteristic average value mu of each category g And a feature covariance matrix Σ g The estimation results of (2) are as follows:
and then constructing an initial hierarchical index tree node, merging areas, maintaining dynamic update of the multi-resolution topological network, iteratively executing the process of searching the merged object sequence, creating a tree node of the upper layer step by step until the segmentation is finished, and outputting the vectorized image spot object.
8. The chained tracking method of the image spot object of the ES database for homeland investigation according to claim 5, wherein grid segmentation is adopted to vectorize the boundary of the image spot, the remote sensing image is required to be segmented to obtain a segmentation image, each image spot is extracted on the basis, then all the contained pixels are extracted by taking the image spot as a unit, and the coordinates of the pixels and the coordinates of the 4 corner points of the pixels are recorded for each pixel; defining pixel corner points positioned inside the boundary polygon as inner pixel corner points, removing vertexes of non-boundary polygons, and defining pixel corner point pairs which are adjacent in position and not belonging to the same effective pixel as vertexes of the boundary polygon as pseudo-adjacent pixel corner points so as to determine a real pixel corner point set to be selected; then finding out a first effective pixel of the top line of the image spot according to the pixel coordinate value, setting the upper left corner of the pixel as a starting vertex for tracking the boundary polygon, taking the upper right corner of the pixel as a second vertex of the boundary polygon, and tracking and searching the rest boundary vertices along the boundary of the image spot according to the vertex searching rule; and finally, sequentially connecting the searched vertexes in sequence to form a required boundary polygon and sealing the boundary polygon to obtain a boundary vector diagram of the image spots.
9. The chained tracking method of the MAP spots of the ES database for homeland investigation according to claim 1, wherein the model evaluation module performs optimization by using a depth evaluation method in the process of sorting the images according to time sequence, specifically comprises an average accuracy rate AP evaluation and a global average accuracy rate MAP evaluation, and the formula of the average accuracy rate AP evaluation is as follows:
wherein, p (i) is the proportion of related images in the previous i time sequences, r (i) is a binary function, the value is 0 or 1, the i-th image is related to the reference image, r (i) =1, otherwise, r (i) =0, and Uni is a normalization constant;
the formula of the global average accuracy MAP evaluation is as follows:
wherein T is n The image coefficient of the nth detected image is the number of times NC is the number of times of detection.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210705310.8A CN114972440B (en) | 2022-06-21 | 2022-06-21 | Chained tracking method for ES database pattern spot objects for homeland investigation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210705310.8A CN114972440B (en) | 2022-06-21 | 2022-06-21 | Chained tracking method for ES database pattern spot objects for homeland investigation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114972440A CN114972440A (en) | 2022-08-30 |
CN114972440B true CN114972440B (en) | 2024-03-08 |
Family
ID=82965688
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210705310.8A Active CN114972440B (en) | 2022-06-21 | 2022-06-21 | Chained tracking method for ES database pattern spot objects for homeland investigation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114972440B (en) |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102254319A (en) * | 2011-04-19 | 2011-11-23 | 中科九度(北京)空间信息技术有限责任公司 | Method for carrying out change detection on multi-level segmented remote sensing image |
US8369622B1 (en) * | 2009-10-29 | 2013-02-05 | Hsu Shin-Yi | Multi-figure system for object feature extraction tracking and recognition |
CN105551028A (en) * | 2015-12-09 | 2016-05-04 | 中山大学 | Method and system for dynamically updating geographic space data based on remote sensing image |
CN108089850A (en) * | 2018-01-02 | 2018-05-29 | 北京建筑大学 | A kind of ground mulching product increment updating method based on image collaboration segmentation with ecogeography zoning ordinance storehouse |
WO2018107939A1 (en) * | 2016-12-14 | 2018-06-21 | 国家海洋局第二海洋研究所 | Edge completeness-based optimal identification method for image segmentation |
WO2021051521A1 (en) * | 2019-09-18 | 2021-03-25 | 平安科技(深圳)有限公司 | Response information obtaining method and apparatus, computer device, and storage medium |
CN113807301A (en) * | 2021-09-26 | 2021-12-17 | 武汉汉达瑞科技有限公司 | Automatic extraction method and automatic extraction system for newly-added construction land |
WO2022000838A1 (en) * | 2020-07-03 | 2022-01-06 | 南京莱斯信息技术股份有限公司 | Markov random field-based method for labeling remote control tower video target |
CN113989652A (en) * | 2021-12-27 | 2022-01-28 | 中国测绘科学研究院 | Method and system for detecting farmland change under layered multiple judgment rules |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9147260B2 (en) * | 2010-12-20 | 2015-09-29 | International Business Machines Corporation | Detection and tracking of moving objects |
-
2022
- 2022-06-21 CN CN202210705310.8A patent/CN114972440B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8369622B1 (en) * | 2009-10-29 | 2013-02-05 | Hsu Shin-Yi | Multi-figure system for object feature extraction tracking and recognition |
CN102254319A (en) * | 2011-04-19 | 2011-11-23 | 中科九度(北京)空间信息技术有限责任公司 | Method for carrying out change detection on multi-level segmented remote sensing image |
CN105551028A (en) * | 2015-12-09 | 2016-05-04 | 中山大学 | Method and system for dynamically updating geographic space data based on remote sensing image |
WO2018107939A1 (en) * | 2016-12-14 | 2018-06-21 | 国家海洋局第二海洋研究所 | Edge completeness-based optimal identification method for image segmentation |
CN108089850A (en) * | 2018-01-02 | 2018-05-29 | 北京建筑大学 | A kind of ground mulching product increment updating method based on image collaboration segmentation with ecogeography zoning ordinance storehouse |
WO2021051521A1 (en) * | 2019-09-18 | 2021-03-25 | 平安科技(深圳)有限公司 | Response information obtaining method and apparatus, computer device, and storage medium |
WO2022000838A1 (en) * | 2020-07-03 | 2022-01-06 | 南京莱斯信息技术股份有限公司 | Markov random field-based method for labeling remote control tower video target |
CN113807301A (en) * | 2021-09-26 | 2021-12-17 | 武汉汉达瑞科技有限公司 | Automatic extraction method and automatic extraction system for newly-added construction land |
CN113989652A (en) * | 2021-12-27 | 2022-01-28 | 中国测绘科学研究院 | Method and system for detecting farmland change under layered multiple judgment rules |
Non-Patent Citations (5)
Title |
---|
基于ArcGIS的耕地坡度级别赋值方法研究与应用;刘建华;地理空间信息;20220425(第04期);全文 * |
基于国土资源"一张图"的遥感监测数据处理模式的设计与实现;马艳娜等;电脑知识与技术;20170925(第27期);全文 * |
基于对象的多源数据变化检测的方法;许淑淑等;测绘通报;20200910(第S1期);全文 * |
遥感与GIS技术在土地利用动态监测及执法监察中的应用;胡黎明等;影像技术;20120615(第03期);全文 * |
遥感聚合图像的快速拓扑矢量化方法;瞿华蓥等;测绘通报;20151025(第10期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN114972440A (en) | 2022-08-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Li et al. | Learning deep semantic segmentation network under multiple weakly-supervised constraints for cross-domain remote sensing image semantic segmentation | |
CN111795704B (en) | Method and device for constructing visual point cloud map | |
Ila et al. | Information-based compact pose SLAM | |
CN113449594B (en) | Multilayer network combined remote sensing image ground semantic segmentation and area calculation method | |
CN108428220B (en) | Automatic geometric correction method for ocean island reef area of remote sensing image of geostationary orbit satellite sequence | |
Jasiewicz et al. | GeoPAT: A toolbox for pattern-based information retrieval from large geospatial databases | |
Sun et al. | Roads and Intersections Extraction from High‐Resolution Remote Sensing Imagery Based on Tensor Voting under Big Data Environment | |
Mourchid et al. | A general framework for complex network-based image segmentation | |
CN112329559A (en) | Method for detecting homestead target based on deep convolutional neural network | |
CN113223042A (en) | Intelligent acquisition method and equipment for remote sensing image deep learning sample | |
Jiang et al. | Parallel structure from motion for UAV images via weighted connected dominating set | |
Lu et al. | Detecting and tracking regional outliers in meteorological data | |
Xu et al. | Topology-based individual tree segmentation for automated processing of terrestrial laser scanning point clouds | |
CN111626311B (en) | Heterogeneous graph data processing method and device | |
Yang et al. | Non-rigid point set registration via global and local constraints | |
Li et al. | Exploring label probability sequence to robustly learn deep convolutional neural networks for road extraction with noisy datasets | |
Chen et al. | A superpixel-guided unsupervised fast semantic segmentation method of remote sensing images | |
CN114972440B (en) | Chained tracking method for ES database pattern spot objects for homeland investigation | |
Xu et al. | The image segmentation algorithm of colorimetric sensor array based on fuzzy C-means clustering | |
CN113515519A (en) | Method, device and equipment for training graph structure estimation model and storage medium | |
Madi et al. | A graph-based approach for kite recognition | |
CN113139556B (en) | Manifold multi-view image clustering method and system based on self-adaptive composition | |
Yu et al. | A lightweight ship detection method in optical remote sensing image under cloud interference | |
Pappula | A Novel Binary Search Tree Method to Find an Item Using Scaling. | |
CN112507826A (en) | End-to-end ecological change monitoring method, terminal, computer equipment and medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |