CN115601687A - Intelligent processing method for on-site survey data in insurance claim settlement process - Google Patents
Intelligent processing method for on-site survey data in insurance claim settlement process Download PDFInfo
- Publication number
- CN115601687A CN115601687A CN202211609391.8A CN202211609391A CN115601687A CN 115601687 A CN115601687 A CN 115601687A CN 202211609391 A CN202211609391 A CN 202211609391A CN 115601687 A CN115601687 A CN 115601687A
- Authority
- CN
- China
- Prior art keywords
- record
- analysis
- value
- area
- analysis window
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
- G06F40/216—Parsing using statistical methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/08—Insurance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/761—Proximity, similarity or dissimilarity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
- G06V10/763—Non-hierarchical techniques, e.g. based on statistics of modelling distributions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
Abstract
The invention relates to the technical field of data analysis and processing, in particular to an intelligent processing method for field investigation data in an insurance claim settlement process, which comprises the following steps: dividing the current site survey record into at least two analysis windows, and acquiring a height record corresponding to each analysis window; acquiring a shooting uniform value according to the height data of each analysis window; acquiring an area sequence corresponding to each analysis window, and acquiring an area floating uniform value according to the area sequence of each analysis window; obtaining the proximity degree between every two analysis windows according to the shooting uniform value and the area floating uniform value, further obtaining a referenceable data set of each analysis window, and classifying samples in the referenceable data set to obtain a plurality of types; acquiring an optimal shooting record according to the type corresponding to each analysis window, and determining supplementary shooting data according to the mean value of the optimal shooting record in each analysis interval; the fairness and the normalization of model reconstruction are guaranteed, and the evaluation result is more reliable.
Description
Technical Field
The invention relates to the technical field of data analysis and processing, in particular to an intelligent processing method for field investigation data in an insurance claim settlement process.
Background
At present, on-site investigation of insurance claims is usually in the traditional form of video, photos and the like, and if SLAM and laser radar scanning are utilized, a three-dimensional reconstruction result which is comparable to a real scene can be reconstructed.
In the prior art, a more vivid three-dimensional model can be produced for an accident scene in a continuous shooting mode, namely, a scene is generated by utilizing SLAM (simultaneous localization and mapping), for example, RTAB-Map or AR-Kit is used for carrying out three-dimensional scanning and surround shooting on the scene, so that a more precise scene can be generated; at present, on-site survey data of an insurance claim settlement process is easy to obtain, but the obtained survey data cannot ensure whether shooting is really finished according to needs, and situations such as deliberate hiding or unnatural moving may exist in the shooting process, so that models can be adhered during three-dimensional modeling, and the situations directly influence the reliability of claim settlement evaluation and are not standard and fair.
Disclosure of Invention
In order to solve the above technical problems, the present invention aims to provide an intelligent processing method for field survey data in an insurance claim settlement process, and the adopted technical scheme is specifically as follows:
one embodiment of the invention provides an intelligent processing method for field survey data in an insurance claim settlement process, which comprises the following steps:
dividing a current on-site survey record by a fixed time period to obtain at least two analysis windows, obtaining a height record corresponding to each analysis window, and corresponding the height records to different analysis intervals, wherein the analysis intervals are preset height proportion intervals;
acquiring a median of data of each analysis window in each analysis interval and a mean of data of all the analysis windows in each analysis interval, and acquiring a photographic uniform value of the corresponding analysis window according to the median of each analysis interval and the mean of the data of all the analysis windows in each analysis interval;
acquiring an area sequence corresponding to each analysis window, calculating an area median and an area variance of the area sequence, calculating a difference value between the area median of the area sequence corresponding to the current analysis window and each area data in the area sequence corresponding to the previous analysis window, and obtaining an area floating uniform value of the current analysis window according to the area variance and all the difference values;
obtaining the proximity degree between every two analysis windows according to the shooting uniform value and the area floating uniform value, and selecting a preset number of adjacent samples corresponding to each analysis window according to the proximity degree to form a referable data set of the analysis window; obtaining a difference distance between every two adjacent samples, and classifying the adjacent samples in the referenceable data set based on the difference distance to obtain at least one type;
and obtaining typical records of the site survey records according to the type corresponding to each analysis window, obtaining the optimal shooting records based on the typical records, and determining supplementary shooting data according to the average value of the optimal shooting records in each analysis interval.
Preferably, the step of obtaining the photographic uniform value of the corresponding analysis window according to the median of each analysis interval and the mean of the data of all the analysis windows in each analysis interval includes:
and acquiring a difference absolute value of the median of each analysis interval and the mean value of the data of all analysis windows in each analysis interval, constructing an exponential function by taking the negative number of the difference absolute value as a power exponent and taking a natural constant e as a base number, and accumulating the exponential functions corresponding to all the analysis intervals to form the shooting uniform value.
Preferably, the step of obtaining the area floating unity value of the current analysis window according to the area variance and all the difference values comprises:
taking the summation result of a preset constant and the area variance as a denominator; obtaining the average value of the absolute values of all the difference values, and constructing an area exponential function by taking the negative number of the average value as a power exponent and taking a natural constant e as a base number; and the ratio result of the area exponential function and the denominator is the area floating uniform value.
Preferably, the step of obtaining the proximity between each two analysis windows according to the shooting unity and the area floating unity includes:
constructing a binary group by using the shooting uniform value and the area floating uniform value corresponding to each analysis window; calculating the L2 distance between the two tuples of each two analysis windows, wherein the L2 distance is the difference degree between the two analysis windows;
the proximity is inversely related to the degree of discrimination.
Preferably, the step of obtaining the differential distance between each two adjacent samples includes:
acquiring a field semantic descriptor of each adjacent sample, and constructing a height vector corresponding to each adjacent sample; acquiring L2 distance between field semantic descriptors of every two adjacent samples and similarity between height vectors of every two adjacent samples; and calculating an accumulation result between a preset constant and the similarity, wherein the ratio of the L2 distance between the field semantic descriptors of every two adjacent samples to the accumulation result is the difference distance between the two adjacent samples.
Preferably, the step of obtaining the area sequence corresponding to each analysis window includes:
and the shot images corresponding to at least two moments in each analysis window are equivalent to obtain area data corresponding to each shot image according to the number of point clouds, and the area data of all the shot images in the analysis windows form an area sequence.
Preferably, the step of obtaining a typical record of the field survey record according to the type corresponding to each analysis window includes:
recording the current survey record as a record to be processed, acquiring at least two field survey records before the record to be processed as target records, wherein each target record comprises at least one analysis window, each analysis window corresponds to at least one type, and a reference sample is obtained according to the center point of each type;
acquiring the degree of difference between the reference samples of the analysis window in the record to be processed and the target record, and obtaining the reachable density based on the degree of difference between the record to be processed and all the reference samples of the analysis window; constructing a distribution histogram according to the reachable densities of all analysis windows in the record to be processed and the target record;
obtaining a median value of the distribution histogram, and obtaining a state uniqueness degree of the record to be processed and the target record according to the median value, wherein the state uniqueness degree and the median value of the distribution histogram are in a negative correlation relationship;
and selecting the maximum value of the state uniqueness between the record to be processed and all the target records, wherein the target record corresponding to the maximum value is the typical record.
Preferably, the step of obtaining an optimal photographing record based on the typical record includes:
acquiring shooting records of other historical colleagues, and calculating the distance between the typical record and each shooting record, wherein the distance is obtained by a distance function, and the shooting record when the distance from the typical record is the minimum is the best shooting record;
the distance function is:
wherein the content of the first and second substances,
a distribution histogram L representing the distribution between the record to be processed and the representative record V;
a distribution histogram L representing the distribution between the record to be processed and the Y-th shot record of the colleague;
representing a degree of state uniqueness between the record to be processed and the typical record V;
representing a state uniqueness between the record to be processed and the Y-th shot record of the co-worker;
histogram of representation distribution
And distribution histogram
Similarity between them;
representing state uniqueness
And state uniqueness
The absolute value of the difference between them.
Preferably, the step of obtaining the achievable density based on the degree of difference between the record to be processed and all reference samples of the analysis window comprises:
and adding the degrees of distinction between the record to be processed and all the reference samples of the analysis window to obtain an accumulated value, wherein the reciprocal of the accumulated value is the reachable density between the record to be processed and the analysis window.
Preferably, the analysis intervals are [1, 1.5), [1.5, 2), [2, 2.5) respectively; the value of the analysis interval is the height ratio, which is calculated as: and obtaining the difference between the model height and the handheld height at each moment, wherein the ratio of the total height of the model to the difference is the height proportion.
The invention has the following beneficial effects: according to the embodiment of the invention, each analysis window in the whole site survey record is analyzed, so that the local characteristics are more obvious, and a shooting uniform value is obtained by combining the shooting height at each moment in each analysis window; then analyzing the area data in the area sequence corresponding to each analysis window to obtain an area floating uniform value, calculating the proximity degree based on the shooting uniform value and the area floating uniform value of each analysis window to further find a proximity sample, classifying the proximity sample through the constructed difference distance to obtain a plurality of types, and performing combined analysis with the proximity sample to ensure that the obtained data has higher generalization; and then typical records are obtained according to a plurality of types corresponding to each analysis window, corresponding optimal shooting records are obtained according to the typical records, the optimal shooting records are obtained through multiple analyses, the result is more accurate, the effect of adjusting based on the optimal shooting records is more reliable, the analysis is only based on the area and the shooting height, the analysis is irrelevant to the accident type, the applicability is stronger, the fairness and the normalization of the data reconstructed by the model are ensured, and the evaluation result is better.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a flowchart of an intelligent processing method for field survey data in an insurance claim settlement process according to an embodiment of the present invention.
Detailed Description
To further illustrate the technical means and effects of the present invention adopted to achieve the predetermined object, the following detailed description of the method for intelligently processing field survey data for insurance claim settlement process according to the present invention, with reference to the accompanying drawings and preferred embodiments, its specific implementation, structure, features and effects are provided below. In the following description, the different references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following describes a specific scheme of the intelligent processing method for field survey data in the insurance claim settlement process, which is provided by the invention, in detail with reference to the accompanying drawings.
Referring to fig. 1, a flowchart of an intelligent processing method for field survey data in an insurance claim settlement process according to an embodiment of the present invention is shown, where the method includes the following steps:
and S100, segmenting the current site survey record in a fixed time period to obtain at least two analysis windows, obtaining a height record corresponding to each analysis window, and corresponding the height records to different analysis intervals, wherein the analysis intervals are preset height proportion intervals.
In the accident scene investigation data recording, taking one-time complete video shooting of an accident as a unit, and a photographer holds a camera to face the accident scene and moves the photographer to surround the whole scene; the orientation and three-dimensional motion trajectory of the camera can be obtained by using a gyroscope and an electronic compass of the camera, and a common SLAM system generally has a ground detection function, so that the height during shooting and the model height during model reconstruction can be obtained.
In the actual process, when different fields are surveyed, the shooting height of the operational camera pose changes, the shooting position of the camera changes, and for the SLAM system, no matter which method is used, local point clouds are added through updating for a period of time and are registered into a model, for example, RTABMAP and Apple ARkit are registered when the camera pose moves for a sufficient distance; when the point cloud is added once, the added camera height of the new point cloud during camera shooting is analyzed, and therefore whether the shooting height possibly bypasses certain visual angles or not is judged in an auxiliary mode.
In the embodiment of the invention, 20 seconds are taken as a fixed time period to analyze one-time on-site survey shooting record, namely, every 20 seconds of time corresponds to one analysis window, the analysis window corresponds to a plurality of moments of shot images, the setting of the moments is set by an implementer, and a height record of the shooting height at each moment in each analysis window is obtained, wherein the shooting height refers to the corresponding handheld height during shooting, namely the distance height between the camera height and the ground, and the obtaining method is the prior known technology and is not described again; matching analysis intervals with the height records corresponding to each analysis window, wherein each analysis interval is a preset height proportion interval, and specifically: the corresponding model heights of different accidents during model reconstruction are different, and the highest height of the point cloud model obtained in the reconstruction process is also unfixed; considering that the difference between the handheld height of the camera at each moment and the height of the model is different, the handheld height of a general person during shooting is about 1.2m, the handheld heights corresponding to different scenes are different, the handheld height refers to the height of the camera from the ground, and the height corresponding to a general accident is 2m or even higher, for example, a serious car accident and the like.
When a camera shoots different environments and different accident objects, a photographer needs to shoot from one part and then gradually complete due to shielding or glass influence, shooting modes and methods are different for different scenes and models in the completing process, and therefore different differences occur in the handheld height of the camera, so that the height ratio is obtained by the handheld height, the model height and the model total height, and the model total height is the highest height when a model of the current accident type is reconstructed; and acquiring the difference between the model height and the handheld height at each moment, wherein the ratio of the total height of the model to the difference is a height proportion, and the larger the handheld height is, the smaller the corresponding height proportion is.
In the embodiment of the invention, 4 height ratios of 1,1.5, 2 and 2.5 are set as dividing points according to experience, and the dividing points are divided into 3 analysis intervals of [1,1.5 ], [1.5,2 ] and [2,2.5 ]; in other embodiments, the setting implementer of the specific analysis interval is subject to the actual height of the site survey. The height proportion of each time corresponding to each analysis window can correspond to 3 analysis intervals, and so on, the shooting height records of each analysis interval in one shooting are obtained, and corresponding time records are added into the point cloud.
Step S200, acquiring a median of data of each analysis window in each analysis interval and a mean of data of all the analysis windows in each analysis interval, and obtaining a shooting uniform value of the corresponding analysis window according to the median of each analysis interval and the mean of data of all the analysis windows in each analysis interval.
Obtaining an analysis interval corresponding to the height record corresponding to each analysis window in one on-site survey by the step S100, wherein each data in the analysis interval is a height proportion of each moment, and counting a median value corresponding to the data in each analysis interval and a mean value of all the analysis windows in each analysis interval in the whole shooting process; because different people have different cognitions on the concerned position and the important position of the accident, certain difference exists in the height ratio, for example, the height ratio corresponding to the current moment is close to 1, and if the situation that the shooting height of the camera pose is higher is found, the situation that the important characteristic of the environment is at the top can be judged; therefore, the shooting uniform value at the moment is calculated according to the median of the analysis interval corresponding to each analysis window and the mean of each analysis interval corresponding to the whole shooting process, and is used for reflecting the uniformity of the height proportion in different analysis intervals, and the shooting uniform value is calculated as follows:
wherein the content of the first and second substances,
representing a photographic uniform value;
a median value representing data for each analysis interval within the analysis window;
representing each analysis interval within all analysis windowsMean of the data;
for calculating the median of the data for each analysis interval within the analysis window
With the mean of the data for each analysis interval over all analysis windows
The absolute value of the difference between;
an exponential function with a natural constant e as the base is shown.
When in site survey, the heights of all analysis intervals in the analysis windows tend to be consistent, namely the closer the median value in the analysis intervals is to the mean value of all the analysis intervals in the analysis windows, the smaller the variation of the shooting height is, the closer the result of the exponential function approaches to 1, and the larger the shooting uniform value is; when there is a large difference in the variation of one of the analysis intervals, the corresponding camera shooting uniform value will be small. Therefore, a shooting uniform value of the camera shooting height corresponding to each analysis window in the field survey is obtained.
Step S300, acquiring an area sequence corresponding to each analysis window, calculating an area median and an area variance of the area sequence, calculating a difference value between the area median of the area sequence corresponding to the current analysis window and each area data in the area sequence corresponding to the previous analysis window, and obtaining an area floating uniform value of the current analysis window according to the area variance and all the difference values.
Specifically, adding area data to point cloud shot at one time during site survey, and recording the corresponding area size at each moment, thereby obtaining the area size during site survey; the point cloud addition area data can be obtained simply through equivalent point cloud number, and also can be obtained through the surface area after nearest neighbor triangulation, and the specific mode is not described in detail; and obtaining the area corresponding to each time in each analysis window, thereby forming the area sequence corresponding to the analysis window.
Further, acquiring a median of the area sequence corresponding to each analysis window, and recording the median as an area median, and taking a variance of the area sequence as an area variance; acquiring an area sequence of a previous analysis window corresponding to a current analysis window, calculating a difference value of an area median of the current analysis window and each area data in the area sequence of the previous analysis window as an area change value, and acquiring an area floating uniform value of the current analysis window based on an average value and an area variance of the area change value, wherein the specific calculation is as follows:
wherein the content of the first and second substances,
representing a uniform value of area floating;
representing the area median value corresponding to the current analysis window;
representing the area sequence corresponding to the current analysis window
The larger the variance is, the larger the dispersion degree of area data in the current window is, the smaller the area floating uniform value is;
representing a variance function calculation;
area data representing an area sequence corresponding to an analysis window immediately preceding the current analysis window;
represents a natural constant;
is a predetermined constant in the embodiment of the present invention.
When the fluctuation corresponding to the area data in the current analysis window is larger, namely the variance is larger, the evaluation of the corresponding area floating uniform value is lower; similarly, when the difference between the area data of the current analysis window and the area data of the previous analysis window is larger, the evaluation of the corresponding area floating unity value is lower.
Step S400, obtaining the proximity degree between every two analysis windows according to the shooting uniform value and the area floating uniform value, and selecting a preset number of adjacent samples corresponding to each analysis window according to the proximity degree to form a referable data set of the analysis windows; and acquiring a difference distance between every two adjacent samples, and classifying the adjacent samples in the reference data set into at least one type based on the difference distance.
When a person starts to use the camera, different usage expectations, usage patterns and intentions for shooting exist in each usage record of different cameras; the proximity degree can be obtained according to the shooting uniform value and the area floating uniform value corresponding to each analysis window obtained in step S200 and step S300, and the degree of distinction between each two analysis windows is first calculated:
wherein the content of the first and second substances,
representing the degree of distinction between the analysis window p and the analysis window q;
representing a photographic uniform value corresponding to the analysis window p;
representing the area floating uniform value corresponding to the analysis window p;
representing a shooting uniform value corresponding to the analysis window q;
representing the area floating uniform value corresponding to the analysis window q;
representing a binary group formed by the shooting uniform value and the area floating uniform value corresponding to the analysis window p;
representing a binary group formed by the shooting uniform value and the area floating uniform value corresponding to the analysis window q;
indicating the L2 distance calculation.
When the binary group formed by the shooting uniform value and the area floating uniform value corresponding to the two analysis windows are more similar, namely the L2 distance between the two binary groups is smaller, the corresponding difference degree is larger; in the embodiment of the present invention, the difference degree and the proximity degree are in a negative correlation relationship, the setting of the specific negative correlation relationship can be represented by an inverse of the difference degree, and in other embodiments, other methods can be used for quantization, so that the smaller the difference degree between two tuples is, the greater the corresponding proximity degree is, that is, the closer the two analysis windows are.
In this way, for different analysis windows, the proximity degree between the analysis window and other analysis windows can be obtained, and the proximity sample corresponding to each analysis window is selected according to the proximity degree, in the embodiment of the present invention, the empirical value K =30 is set, that is, the proximity sample corresponding to K proximity to each analysis window is selected, where K proximity refers to the first 30 analysis windows with the maximum proximity degree to the analysis window, and the first 30 analysis windows with the maximum proximity degree to each analysis window are marked as the proximity samples of the analysis window, so that each analysis window corresponds to 30 proximity samples.
For different analysis windows, all the analysis windows have 30 adjacent samples, and the 30 adjacent samples corresponding to each analysis window form a referenceable data set of the analysis window. In order to analyze the characteristics among different analysis windows in a more detailed and subdivided manner, unsupervised classification is carried out on the referable data set corresponding to each analysis window so as to realize the effect of local search; the method for unsupervised classification in the embodiment of the invention is a DBSCAN clustering algorithm, and the method for classifying different adjacent samples in a referential data set comprises the following steps:
firstly, acquiring a field semantic descriptor corresponding to each adjacent sample; since the general situation of the scene can be described in a text mode during the scene survey, the scene semantic descriptor can assist in restricting the accident factor of the scene for any adjacent samples; after word segmentation and bag-of-word models are established for texts in all records in the insurance industry, feature values can be calculated through TF-IDF aiming at adjacent samples.
Then, acquiring a height vector corresponding to each adjacent sample; because the height record corresponding to each moment in the analysis window corresponding to each adjacent sample can correspond to the analysis interval, the minimum value and the maximum value of the data corresponding to the analysis interval can be correspondingly obtained in the analysis interval, and the data range from the minimum value to the maximum value of the data is divided in equal proportion, wherein the grade of the equal-proportion division in the embodiment of the invention is 4, namely the data range in each analysis interval is divided into 4 grades in equal proportion; each analysis window corresponds to 3 analysis intervals, each analysis interval comprises 4 grades, and one analysis window corresponds to 3 × 4=12 grades; the corresponding mean value can be obtained in each grade range of each analysis interval, and for an analysis window, the 12 mean values can correspond to 12, and the 12 mean values are arranged in sequence to obtain the height vector corresponding to the analysis window, that is, the height vector of the adjacent sample is obtained.
And finally, calculating the difference distance between different adjacent samples according to the field semantic descriptor and the height vector corresponding to each adjacent sample, wherein the difference distance is used as a grouping basis, and the calculation of the difference distance is as follows:
wherein the content of the first and second substances,
represents the difference distance between the adjacent sample a and the adjacent sample B;
representing the scene semantic descriptors corresponding to the adjacent samples A;
representing the corresponding scene semantic descriptor of the adjacent sample B;
representing the height vector corresponding to the adjacent sample a;
representing the corresponding height vector of the adjacent sample B;
represents a conventional L2 distance calculation;
the similarity calculation is expressed, an implementer can select a similarity acquisition method, and cosine similarity is adopted for characterization in the embodiment of the invention.
When the corresponding height vectors between two adjacent samples are more similar, the shooting height characteristics corresponding to the two adjacent samples are more similar, and the corresponding difference distance is smaller; and when the L2 distance of the field semantic descriptors between two adjacent samples is larger, the fact that the field semantic descriptors between the two adjacent samples are different is shown, and the corresponding difference distance is larger.
Therefore, all the adjacent samples are classified according to the difference distance between every two adjacent samples to obtain a plurality of types, and for each type, a corresponding central sample can be obtained.
And S500, obtaining typical records of the site survey records according to the type corresponding to each analysis window, obtaining the optimal shooting records based on the typical records, and determining supplementary shooting data according to the average value of the optimal shooting records in each analysis interval.
Obtaining a referenceable data set corresponding to each analysis window in the step S400, classifying all adjacent samples in the referenceable data set to obtain a plurality of types, and searching for the most matched type based on records of using the camera for the last few times; generally, one camera device is frequently used for one person or a small working group, so that when one person carries a certain device to carry out field investigation, typical records of the field investigation records can be searched according to the previous investigation records to obtain to-be-matched record data of the field investigation.
Recording the survey record as a to-be-processed record, acquiring at least two site survey records before the to-be-processed record as target records, wherein each target record comprises at least one analysis window, each analysis window corresponds to at least one type, and a reference sample is obtained according to the central point of each type; acquiring the degree of difference between reference samples of an analysis window in a record to be processed and a target record, and obtaining the reachable density based on the degree of difference between the record to be processed and all the reference samples of the analysis window; constructing a distribution histogram according to the reachable densities of all analysis windows in the record to be processed and the target record; acquiring a median value of the distribution histogram, and obtaining a state uniqueness degree of the record to be processed and the target record according to the median value, wherein the state uniqueness degree and the median value of the distribution histogram are in a negative correlation relationship; and selecting the maximum value of the state uniqueness between the record to be processed and all the target records, wherein the target record corresponding to the maximum value is a typical record.
Firstly, obtaining the most similar type according to the records of a plurality of previous times, selecting the site survey record of the previous 3 times of the record as a target record in the embodiment of the invention, and recording the site survey record as a record to be processed, wherein the method for obtaining the most similar type of the record to be processed comprises the following steps:
taking the distinguishing degree as a reference, acquiring a shooting uniform value and an area floating uniform value corresponding to the record to be processed, and further forming a binary group corresponding to the record to be processed; since the target record includes a plurality of analysis windows, each analysis window corresponds to a reference data set, and the reference data corresponds to a plurality of types, taking any one analysis window in a certain target record as an example, the achievable density between the record to be processed and each type of the target window is calculated, that is, the similarity between the record to be processed and a neighboring sample corresponding to the target window, and since each type of the target window corresponds to a reference sample, the reference sample may be obtained based on the same method in step S400, the calculation of the achievable density is obtained based on the degree of difference between the record to be processed and the reference sample of each type, and the calculation method of the degree of difference is the same as the method in step S400, and then the achievable density is specifically calculated as:
wherein the content of the first and second substances,
represents the achievable density;
representing records to be processed
The degree of discrimination from the nth type of reference sample of the target window;
indicating the number of all types in the reference sample set corresponding to the target window.
When the degree of difference between each type corresponding to the record to be processed and the target window is smaller, it indicates that the characteristics between the record to be processed and the target window are relatively similar, and the corresponding reachable density is relatively high.
Then, for the to-be-processed record of the current on-site survey, the reachable density of the to-be-processed record corresponding to each target window of the target record can be obtained, a distribution histogram L of the to-be-processed record is constructed according to the to-be-processed record and the reachable densities corresponding to all the target windows, wherein the horizontal axis of the distribution histogram L is 10 equally divided intervals of the upper and lower boundaries of the to-be-processed record corresponding to all the target windows, range standardization is performed on the distribution histogram L corresponding to the to-be-processed record, at the moment, each interval in the distribution histogram L represents the occurrence frequency of the reachable density range, and when the reachable density range has strong representativeness, the value of the interval is larger than the values of other intervals; in the embodiment of the present invention, the median of the corresponding values of all the intervals in the distribution histogram L is used as the typical value of the distribution histogram L, and the state uniqueness between the record to be processed and the target record is obtained according to the typical value, and specifically calculated as follows:
wherein the content of the first and second substances,
representing a state uniqueness;
the median value of the distribution histogram L is represented, i.e. is a typical value.
The larger the state uniqueness is, the more common the condition belonging to the target record is indicated, thereby representing the state uniqueness under the influence of field factors and handheld factors during operation.
Obtaining a distribution histogram L and a state uniqueness degree between the record to be processed and a target record based on the above
The same method is used for obtaining a distribution histogram L and a state uniqueness degree between the record to be processed and the other two target records
Selecting a degree of uniqueness of a state
And recording the target record with the maximum value as the data of the record to be matched of the field survey record, and recording the data as the most typical record of the record to be processed.
Further, the closest record between the typical record V and the shooting records of other colleagues is obtained as the nearest neighbor match, and the shooting height corresponding to the record of the nearest neighbor match is used as the current timeJudging basis of field survey records; each shooting record and the record to be processed of the colleagues can obtain a corresponding distribution histogram L and a state uniqueness degree
Based on distribution histogram L and state uniqueness
Setting a distance function of
(ii) a Wherein the content of the first and second substances,
a distribution histogram L representing the distribution between the record to be processed and the representative record V;
a distribution histogram L representing the distribution between the record to be processed and the Y-th shot record of the colleague;
representing a state uniqueness between the record to be processed and the typical record V;
representing a state uniqueness between the pending record and the Y-th shot record of the co-worker.
Similarity between distribution histograms
The calculation of (1) is to divide each distribution histogram into corresponding vectors, that is, the numerical values of the longitudinal axes of the distribution histograms correspond to the elements of the vectors, and the cosine similarity between the vectors is used to characterize the similarity between the corresponding distribution histograms, and the specific method for obtaining the cosine similarity is the prior art and is not repeated;
means for indicating absoluteFor the function of the value(s),
representing the degree of uniqueness of a computation state
And state uniqueness
The smaller the value of the absolute value of the difference value is, the more consistent the state uniqueness between the two records is, although the upper and lower boundaries corresponding to the distribution histogram are different, the upper and lower boundaries of the distribution histogram are constrained to be similar as much as possible due to the similarity of the state uniqueness, and the shooting record with the minimum distance is found out more accurately according to the shooting records of the distance function which are distributed all over the colleagues.
Recording the best shooting record corresponding to the minimum distance function as Q, taking the best shooting record Q as the best normative reference, wherein the best shooting record Q corresponds to a complete site survey record, and therefore the mean value of the normative shooting height can be obtained based on the best shooting record Q
(ii) a Recording the mean value corresponding to Q based on the best shot
And checking and contrasting, then giving suggestions to shooting personnel, and performing supplementary shooting on a part of improper shooting processes to avoid the problem of adhesion of the reconstructed model caused by cheating and mistaken card visual angle or shooting angle.
Firstly, for the typical record P, the maximum shooting height difference in the record is limited to be the largest one-level shooting height of the optimal shooting record Q, if the limiting condition is met, scanning reconstruction is carried out again, and therefore the camera cannot cause the problem that some objects are shielded by foreign matters on the bottom surface due to the fact that the shooting height is too low, and excessive looking-up cannot occur. Then, the default shooting height is given as the average value of the best shooting record Q
For a shooting process, it is required that the camera trajectory needs to maintain at least the mean value continuously while moving around the object
The minimum shot height is 1 minute to prevent the camera from missing some views, resulting in foreground protrusions that obscure detail while modeling. After an analysis interval is reached, the minimum shooting height is limited to be the minimum first-level height difference of the corresponding interval of the optimal shooting record Q, so that the problem that the ground scene or a lower scene is not scanned sufficiently is solved, and the requirement that the shooting is carried out for 1 minute around field scanning in the height range is met after scanning, so that the operation normative and the fairness during claim evaluation are improved.
Therefore, according to the modeling process of automatically analyzing the site survey data, operations needing to be supplemented are intelligently restricted and pointed out, the modeling quality of the accident site can be improved, unnecessary repeated modeling time is reduced, and in consideration of the fact that the height of an object of the accident site is uncertain, the method and the device have no direct relation to different accident sites, and therefore have strong applicability.
In summary, in the embodiment of the present invention, the current site survey record is segmented by a fixed time period to obtain at least two analysis windows, a height record corresponding to each analysis window is obtained, and the height records are corresponding to different analysis intervals, where the analysis intervals are preset height proportion intervals; acquiring a median of data of each analysis window in each analysis interval and a mean value of data of all the analysis windows in each analysis interval, and acquiring a shooting uniform value of the corresponding analysis window according to the median of each analysis interval and the mean value of the data of all the analysis windows in each analysis interval; acquiring an area sequence corresponding to each analysis window, calculating an area median and an area variance of the area sequence, calculating a difference value of the area median of the area sequence corresponding to the current analysis window and each area data in the area sequence corresponding to the previous analysis window, and obtaining an area floating uniform value of the current analysis window according to the area variance and all the difference values; obtaining the proximity degree between every two analysis windows according to the shooting uniform value and the area floating uniform value, and selecting a preset number of adjacent samples corresponding to each analysis window according to the proximity degree to form a referenceable data set of the analysis windows; acquiring a difference distance between every two adjacent samples, and classifying the adjacent samples in the referenceable data set based on the difference distance to obtain at least one type; obtaining typical records of the site survey records according to the type corresponding to each analysis window, obtaining optimal shooting records based on the typical records, and determining supplementary shooting data according to the average value of the optimal shooting records in each analysis interval; the normative and the fairness when the field investigation data are analyzed and claim are improved, and the method is high in applicability and high in efficiency.
It should be noted that: the sequence of the above embodiments of the present invention is only for description, and does not represent the advantages or disadvantages of the embodiments. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that are within the spirit of the present invention are intended to be included therein.
Claims (10)
1. An intelligent processing method for on-site survey data in an insurance claim settlement process is characterized by comprising the following steps:
the method comprises the steps of segmenting a current site survey record by a fixed time period to obtain at least two analysis windows, obtaining a height record corresponding to each analysis window, and enabling the height records to correspond to different analysis intervals, wherein the analysis intervals are preset height proportion intervals;
acquiring a median of data of each analysis window in each analysis interval and a mean of data of all the analysis windows in each analysis interval, and acquiring a photographic uniform value of the corresponding analysis window according to the median of each analysis interval and the mean of the data of all the analysis windows in each analysis interval;
acquiring an area sequence corresponding to each analysis window, calculating an area median and an area variance of the area sequence, calculating a difference value between the area median of the area sequence corresponding to the current analysis window and each area data in the area sequence corresponding to the previous analysis window, and obtaining an area floating uniform value of the current analysis window according to the area variance and all the difference values;
obtaining the proximity degree between every two analysis windows according to the shooting uniform value and the area floating uniform value, and selecting a preset number of adjacent samples corresponding to each analysis window according to the proximity degree to form a referable data set of the analysis window; acquiring a difference distance between every two adjacent samples, and classifying the adjacent samples in the referenceable data set based on the difference distance to obtain at least one type;
and obtaining typical records of the site survey records according to the type corresponding to each analysis window, obtaining the optimal shooting records based on the typical records, and determining supplementary shooting data according to the average value of the optimal shooting records in each analysis interval.
2. The method of claim 1, wherein the step of obtaining the photographic uniform value of the corresponding analysis window according to the median value of each analysis interval and the mean value of the data of all analysis windows in each analysis interval comprises:
and acquiring a difference absolute value of the median of each analysis interval and the mean value of the data of all analysis windows in each analysis interval, constructing an exponential function by taking the negative number of the difference absolute value as a power exponent and taking a natural constant e as a base number, and accumulating the exponential functions corresponding to all the analysis intervals to form the photographic uniform value.
3. The method of claim 1, wherein the step of deriving a floating unity value for the area of the current analysis window from the area variance and all of the differences comprises:
taking a summation result of a preset constant and the area variance as a denominator; obtaining the average value of the absolute values of all the difference values, and constructing an area exponential function by taking the negative number of the average value as a power exponent and taking a natural constant e as a base number; and the ratio result of the area exponential function and the denominator is the area floating uniform value.
4. The method of claim 1, wherein the step of deriving a proximity between each two analysis windows based on the photographic uniform value and the area floating uniform value comprises:
constructing a binary group by using the shooting uniform value and the area floating uniform value corresponding to each analysis window; calculating the L2 distance between the two tuples of each two analysis windows, wherein the L2 distance is the difference degree between the two analysis windows;
the proximity is inversely related to the degree of discrimination.
5. The intelligent processing method for the field survey data of the insurance claim settlement process, according to claim 3, wherein the step of obtaining the difference distance between each two adjacent samples comprises:
acquiring a field semantic descriptor of each adjacent sample, and constructing a height vector corresponding to each adjacent sample; acquiring the L2 distance between the field semantic descriptors of every two adjacent samples and the similarity between the height vectors of every two adjacent samples; and calculating an accumulation result between a preset constant and the similarity, wherein the ratio of the L2 distance between the field semantic descriptors of every two adjacent samples to the accumulation result is the difference distance between the two adjacent samples.
6. The intelligent processing method for the on-site survey data of the insurance claim settlement process, wherein the step of obtaining the area sequence corresponding to each analysis window comprises:
and the shot images corresponding to at least two moments in each analysis window are equivalent to obtain area data corresponding to each shot image according to the number of point clouds, and the area data of all the shot images in the analysis windows form an area sequence.
7. The intelligent processing method for the field investigation data of the insurance claims process, wherein the step of obtaining the typical record of the field investigation record according to the type corresponding to each analysis window comprises:
recording the current survey record as a record to be processed, acquiring at least two field survey records before the record to be processed as target records, wherein each target record comprises at least one analysis window, each analysis window corresponds to at least one type, and a reference sample is obtained according to the center point of each type;
acquiring the difference degree between the record to be processed and the reference sample of the analysis window in the target record, and acquiring the reachable density based on the difference degree between the record to be processed and all the reference samples of the analysis window; constructing a distribution histogram according to the reachable densities of all analysis windows in the record to be processed and the target record;
obtaining a median value of the distribution histogram, and obtaining a state uniqueness degree of the record to be processed and the target record according to the median value, wherein the state uniqueness degree and the median value of the distribution histogram are in a negative correlation relationship;
and selecting the maximum value of the state uniqueness between the record to be processed and all the target records, wherein the target record corresponding to the maximum value is the typical record.
8. The intelligent processing method for the on-site survey data of the insurance claims process as claimed in claim 7, wherein the step of obtaining the best shot record based on the typical record comprises:
acquiring shooting records of other historical colleagues, and calculating the distance between the typical record and each shooting record, wherein the distance is obtained by a distance function, and the shooting record with the minimum distance from the typical record is the best shooting record;
the distance function is:
wherein the content of the first and second substances,
a distribution histogram L representing the distribution between the record to be processed and the typical record V;
a distribution histogram L representing the distribution between the record to be processed and the Y-th shot record of the colleague;
representing a degree of state uniqueness between the record to be processed and the typical record V;
representing a state uniqueness between the record to be processed and the Y-th shot record of the co-worker;
histogram of representation distribution
And distribution histogram
The similarity between them;
representing state uniqueness
And state uniqueness
The absolute value of the difference between them.
9. The intelligent processing method for the on-site survey data of the insurance claims process as claimed in claim 7, wherein the step of obtaining the achievable density based on the degree of difference between the record to be processed and all the reference samples of the analysis window comprises:
and adding the degrees of distinction between the record to be processed and all the reference samples of the analysis window to obtain an accumulated value, wherein the reciprocal of the accumulated value is the reachable density between the record to be processed and the analysis window.
10. The intelligent processing method for the on-site survey data of the insurance claim settlement process, according to claim 1, wherein the analysis intervals are [1,1.5 ], [1.5,2 ], [2, 2.5); the value of the analysis interval is the height ratio, which is calculated as: and obtaining the difference between the model height and the handheld height at each moment, wherein the ratio of the total height of the model to the difference is the height proportion.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211609391.8A CN115601687B (en) | 2022-12-15 | 2022-12-15 | Intelligent processing method for on-site survey data in insurance claim settlement process |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211609391.8A CN115601687B (en) | 2022-12-15 | 2022-12-15 | Intelligent processing method for on-site survey data in insurance claim settlement process |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115601687A true CN115601687A (en) | 2023-01-13 |
| CN115601687B CN115601687B (en) | 2023-03-07 |
Family
ID=84854309
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211609391.8A Active CN115601687B (en) | 2022-12-15 | 2022-12-15 | Intelligent processing method for on-site survey data in insurance claim settlement process |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115601687B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116704843A (en) * | 2023-06-07 | 2023-09-05 | 广西茜英信息技术有限公司 | Virtual simulation training platform based on communication engineering investigation design |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005057343A (en) * | 2003-08-05 | 2005-03-03 | Tokio Marine & Nichido Fire Insurance Co Ltd | Apparatus and method of registering image data |
| CN102542293A (en) * | 2011-12-27 | 2012-07-04 | 上海交通大学 | Class-I extraction and classification method aiming at high-resolution SAR (Synthetic Aperture Radar) image scene interpretation |
| CN106296816A (en) * | 2016-08-01 | 2017-01-04 | 清华大学深圳研究生院 | Unmanned plane determining method of path and device for reconstructing three-dimensional model |
| KR101741960B1 (en) * | 2016-10-13 | 2017-05-30 | 홍병진 | System and method for claiming insurance based on online environment |
| CN108009314A (en) * | 2017-11-03 | 2018-05-08 | 中国科学院遥感与数字地球研究所 | A kind of ruins area early stage landform three-dimensional rebuilding method based on archaeological data |
| US20190139274A1 (en) * | 2016-05-27 | 2019-05-09 | Korea Advanced Institute Of Science And Technology | Method and apparatus for generating magnetic resonance image |
| CN110751646A (en) * | 2019-10-28 | 2020-02-04 | 支付宝(杭州)信息技术有限公司 | Method and device for identifying damage by using multiple image frames in vehicle video |
| CN110853073A (en) * | 2018-07-25 | 2020-02-28 | 北京三星通信技术研究有限公司 | Method, device, equipment and system for determining attention point and information processing method |
| CN111476771A (en) * | 2020-04-03 | 2020-07-31 | 中山大学 | Domain self-adaptive method and system for generating network based on distance countermeasure |
| CN113160410A (en) * | 2021-04-19 | 2021-07-23 | 云南云能科技有限公司 | Real scene three-dimensional refined modeling method and system |
| CN115456316A (en) * | 2022-11-11 | 2022-12-09 | 山东交通职业学院 | Ship multi-energy power supply system management method |
-
2022
- 2022-12-15 CN CN202211609391.8A patent/CN115601687B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005057343A (en) * | 2003-08-05 | 2005-03-03 | Tokio Marine & Nichido Fire Insurance Co Ltd | Apparatus and method of registering image data |
| CN102542293A (en) * | 2011-12-27 | 2012-07-04 | 上海交通大学 | Class-I extraction and classification method aiming at high-resolution SAR (Synthetic Aperture Radar) image scene interpretation |
| US20190139274A1 (en) * | 2016-05-27 | 2019-05-09 | Korea Advanced Institute Of Science And Technology | Method and apparatus for generating magnetic resonance image |
| CN106296816A (en) * | 2016-08-01 | 2017-01-04 | 清华大学深圳研究生院 | Unmanned plane determining method of path and device for reconstructing three-dimensional model |
| KR101741960B1 (en) * | 2016-10-13 | 2017-05-30 | 홍병진 | System and method for claiming insurance based on online environment |
| CN108009314A (en) * | 2017-11-03 | 2018-05-08 | 中国科学院遥感与数字地球研究所 | A kind of ruins area early stage landform three-dimensional rebuilding method based on archaeological data |
| CN110853073A (en) * | 2018-07-25 | 2020-02-28 | 北京三星通信技术研究有限公司 | Method, device, equipment and system for determining attention point and information processing method |
| CN110751646A (en) * | 2019-10-28 | 2020-02-04 | 支付宝(杭州)信息技术有限公司 | Method and device for identifying damage by using multiple image frames in vehicle video |
| CN111476771A (en) * | 2020-04-03 | 2020-07-31 | 中山大学 | Domain self-adaptive method and system for generating network based on distance countermeasure |
| CN113160410A (en) * | 2021-04-19 | 2021-07-23 | 云南云能科技有限公司 | Real scene three-dimensional refined modeling method and system |
| CN115456316A (en) * | 2022-11-11 | 2022-12-09 | 山东交通职业学院 | Ship multi-energy power supply system management method |
Non-Patent Citations (2)
| Title |
|---|
| 刘钢等: "基于多幅图像的场景交互建模系统", 《计算机辅助设计与图形学学报》 * |
| 向怀坤等: "超低空无人机交通事故现场三维建模勘测系统设计", 《深圳职业技术学院学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116704843A (en) * | 2023-06-07 | 2023-09-05 | 广西茜英信息技术有限公司 | Virtual simulation training platform based on communication engineering investigation design |
| CN116704843B (en) * | 2023-06-07 | 2024-02-23 | 广西茜英信息技术有限公司 | Virtual simulation training platform based on communication engineering investigation design |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115601687B (en) | 2023-03-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xie et al. | Multilevel cloud detection in remote sensing images based on deep learning | |
| CN111104867B (en) | Recognition model training and vehicle re-recognition method and device based on part segmentation | |
| US9846946B2 (en) | Objection recognition in a 3D scene | |
| Kandaswamy et al. | Efficient texture analysis of SAR imagery | |
| EP1233374A2 (en) | Apparatus and method for extracting objects based on feature matching between segmented regions in images | |
| US9626585B2 (en) | Composition modeling for photo retrieval through geometric image segmentation | |
| CN104978567B (en) | Vehicle checking method based on scene classification | |
| CN103530638B (en) | Method for pedestrian matching under multi-cam | |
| CN110866079A (en) | Intelligent scenic spot real scene semantic map generating and auxiliary positioning method | |
| Benini et al. | Shot scale distribution in art films | |
| CN113129335B (en) | Visual tracking algorithm and multi-template updating strategy based on twin network | |
| CN102346854A (en) | Method and device for carrying out detection on foreground objects | |
| CN115131760B (en) | Lightweight vehicle tracking method based on improved feature matching strategy | |
| CN108073940B (en) | Method for detecting 3D target example object in unstructured environment | |
| CN115601687B (en) | Intelligent processing method for on-site survey data in insurance claim settlement process | |
| Wang et al. | Combining semantic scene priors and haze removal for single image depth estimation | |
| CN111695373A (en) | Zebra crossing positioning method, system, medium and device | |
| CN112053371A (en) | Water body extraction method and device in remote sensing image | |
| CN111709317A (en) | Pedestrian re-identification method based on multi-scale features under saliency model | |
| CN114743139A (en) | Video scene retrieval method and device, electronic equipment and readable storage medium | |
| CN110636248B (en) | Target tracking method and device | |
| CN107145820B (en) | Binocular positioning method based on HOG characteristics and FAST algorithm | |
| CN116862832A (en) | Three-dimensional live-action model-based operator positioning method | |
| CN112015937B (en) | Picture geographic positioning method and system | |
| CN115063698A (en) | Automatic identification and information extraction method and system for slope surface deformation crack |
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 |