A Random Error Analysis System and Method for Target Features Based on Cloud Computing

Abstract

The present invention discloses a system and method for analyzing random errors in target features based on cloud computing, which relates to the technical field of image processing. It records the flight state of an unmanned aerial vehicle (UAV), correlates the flight state of the UAV with the movement trajectory of the same object in the picture in the on-board camera, collects and obtains a picture transfer database, intercepts video segments with non-random errors in the pictures from the shooting records, fits a function of the picture features with non-random errors changing over time, extracts picture frames from the pictures taken by the on-board camera, extracts image regions with abnormal contrast changes in the video frames, obtains the record of the real-time flight state of the UAV, the region of the shooting object in the image region in the shooting picture, respectively obtains image slices of the two image regions, and judges whether there is random interference in the picture from two aspects: the change characteristics of the picture being disturbed and the comparison with the monitoring model.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and specifically to a target feature random error analysis system and method based on cloud computing. Background Art

[0002] When a drone takes pictures, due to the high-speed flight of the target, there is often a situation of blurred motion, resulting in a blurred effect in consecutive frame images. When a blurred effect occurs in consecutive frame images, noise of varying degrees will inevitably be introduced during the image acquisition, storage, and transmission of the drone, such as salt-and-pepper noise, Gaussian noise, etc., which will cause image errors. Due to the rapid changes and ambiguity of high-speed moving objects, traditional pose estimation methods often cannot provide accurate results. Therefore, in the prior art, a large number of images disturbed by random errors will be generated in the video images recorded or transmitted back by the drone camera system, which brings difficulties to image processing and image analysis. Summary of the Invention

[0003] The purpose of the present invention is to provide a target feature random error analysis system and method based on cloud computing to solve the problems raised in the prior art.

[0004] To achieve the above purpose, the present invention provides the following technical solution: A target feature random error analysis method based on cloud computing, the method comprising:

[0005] Step S100: During the flight of the drone, record the flight state of the drone, correspond the flight state of the drone with the motion trajectory of the same object in the picture of the on-board camera, and collect to obtain a picture transfer database;

[0006] Step S200: Obtain the historical shooting records of the on-board camera, intercept the video segments in the picture with non-random errors from the shooting records, fit the function of the picture features of the non-random errors changing with time, and collect them into an error change database;

[0007] Step S300: Extract picture frames from the pictures taken by the on-board camera, and extract the image regions with abnormal contrast changes in the video frames;

[0008] Step S400: Obtain the record of the real-time flight state of the drone, extract the region of the shooting object in the shooting picture one unit time ago in the image region, and respectively obtain the image slices of the two image regions;

[0009] Step S500: Compare the image errors of the two image slices, and when the difference between the image error and the non-random error is greater than the threshold, send the image slice to the relevant personnel.

[0010] Further, step S100 includes:

[0011] Step S101: Shoot an object through the on-board camera of the drone. Take a certain object as the target object. Divide the shooting image of the on-board camera of the drone into several unit areas. Establish a plane coordinate system for the shooting image. Number each unit area according to the coordinates of the unit area;

[0012] Step S102: Obtain the video record of the target object shot by the on-board camera, and intercept a section of the video record as the first target video record;

[0013] Step S103: Obtain the flight attitude and flight speed of the drone when shooting the first target video record and record them in the first state set R1. From two frames in the first target video record, obtain the coordinates ref of the unit area where the target object is located in the first frame 1 , and the corresponding coordinates ref in the second frame 2 ;

[0014] Step S104: In the plane coordinate system, calculate the vector coordinates α of the indication vector, α = ref 2 - ref 1 , associate the indication vector with the first state set R1 to form a picture transfer data group, adjust the flight attitude and flight speed of the drone, and collect several picture transfer data groups to obtain a picture transfer database.

[0015] Further, step S200 includes:

[0016] Step S201: Obtain the video segment with error pictures from the video record shot by the on-board camera, and intercept a section of the video record as the second target video record;

[0017] Step S202: Obtain n frames of pictures from the second target video record, mark the unit areas with picture errors in each frame, collect the unit areas with error pictures in the kth frame to form an error area. The error area is composed of adjacent unit areas. Obtain the image of the error area, denoted as the error image im of the kth error area k ;

[0018] Step S203: Collect the error images of each frame, arrange the errors in the order of the pictures in the second target video record to obtain an error image sequence, calculate the image quality evaluation index of each error image, and arrange them in the order of the error images to obtain an error value sequence;

[0019] Step S204: Perform function fitting on the sequence of image quality evaluation metrics to obtain a function of the image quality evaluation metric varying with time, and denote the function as the error variation function. Obtain the flight attitude and flight speed of the drone when shooting the second target video record and record them in the second state set R2. Combine the error variation function with the second state set R2 to form an error variation group. Collect several sets of corresponding relationships between the error variation function and the second state set R2, and gather the error variation groups to obtain an error variation database.

[0020] Further, the calculation method of the image quality evaluation metric includes:

[0021] 2-1: Convert the image into a grayscale image, set u gray-level judgment levels to quantify the grayscale image, and count the occurrence probabilities of each gray level in the grayscale image;

[0022] 2-2: Obtain the probability p of the e-th gray level, and calculate the information entropy H of the grayscale image, e ;

[0023] ;

[0024] 2-3: Use the information entropy as the image quality evaluation metric of the grayscale image.

[0025] Further, step S300 includes:

[0026] Step S301: From the real-time video captured by the on-board camera of the drone, obtain the first image and the second image in the shooting order of the video, calculate the contrast of the first image and the second image respectively, and denote them as CR1 and CR2. When CR1 > CR2, calculate the contrast threshold M, M = CR1 - CR2;

[0027] Step S302: Obtain the image slice corresponding to the j-th unit area in the first image and denote it as psc 1j , and the image slice corresponding to the j-th unit area in the second image and denote it as psc 2j , obtain the contrast of psc 1j and denote it as r 1j , and the contrast of psc 2j and denote it as r 2j ;

[0028] Step S303: When r 1j - r 2j > M, mark the j-th unit area in the second image as the first target unit area;

[0029] When the video image is interfered, a main feature that appears is the decrease in the contrast of the picture. Taking the change in the overall contrast of the two captured pictures as a reference value, compare, unit area by unit area, the speed at which the contrast decreases in the unit areas where the decrease exceeds the reference value;

[0030] In the video recording, the first image is recorded before the second image. Therefore, the areas with a more severe decrease in contrast in the second image may be interfered by random errors.

[0031] Furthermore, step S400 includes:

[0032] Step S401: Obtain the timestamps of the first image and the second image captured by the on-board camera of the drone. Denote the time period from capturing the first image to capturing the second image as the target time period. Obtain the flight attitude and flight speed of the drone during the target time period, and record them in the target state set RT;

[0033] Step S402: Obtain the first state set identical to the target state set RT in the picture transfer database, and obtain the vector coordinates β of the indication vector associated with the first state set;

[0034] Step S403: Obtain the coordinates ref t1 of the first target unit area, calculate the coordinates ref t2 of the second target unit area, ref t2 = ref t1 -β, and the second target unit area is in the first image;

[0035] Step S404: Obtain the image slice of the first target unit area, denoted as psc 1 and the image slice of the second target unit area, denoted as psc 2 , respectively obtain the image quality evaluation indicators of psc 1 and psc 2 , and denote them as rq 1 and rq 2 ;

[0036] By tracing back to the first target area, find the corresponding area captured before the unit time in the first image for the first target area, and further compare the first target area with the second target area to confirm whether the on-board camera is interfered by random errors.

[0037] Furthermore, step S500 includes:

[0038] Step S501: Obtain the second state set identical to the target state set RT in the error change database, and obtain the error change function associated with the second state set;

[0039] Step S502: Mark the data points in the error change function whose values are equal to the image quality evaluation index rq 2 and record the data points as the first data points. In the error change function, obtain the data points passing through the time length of the target time period and record them as the second data points;

[0040] Step S503: Obtain the image quality evaluation index corresponding to the second data point and use the image quality evaluation index as the error reference value;

[0041] Step S504: Set the error threshold γ. When |Q - rq 1 | > γ, mark the image slice psc 1 and send it to the relevant management personnel.

[0042] To better execute the above method, a target feature random error analysis system based on cloud computing is also proposed. The system includes:

[0043] Furthermore, a screen movement management module, an error management module, a contrast comparison module, an image slice management module, and an image judgment module. Among them, the screen movement management module is used to manage the relative motion relationship between the objects and the viewing angle in the images captured by the airborne camera under different flight states. The error management module is used to manage the screen features of non-random errors. The contrast comparison module is used to compare the contrast changes in the screen frames to obtain the first target unit area. The image slice management module is used to obtain the aircraft state records of the UAV and manage the image slices. The image judgment module is used to judge whether there are non-random errors in the image slices;

[0044] Furthermore, the screen movement management module includes: a screen management unit, an indication vector management unit, and a screen transfer database management unit. Among them, the screen management unit is used to manage the unit management area in the images captured by the airborne camera. The indication vector management unit is used to mark the relative motion relationship of the objects captured in the UAV aircraft state to obtain the indication vector. The screen transfer database management unit is used to manage the screen transfer database;

[0045] Furthermore, the error management module includes: an error image management unit, an image quality evaluation unit, a function fitting unit, and an error change database management unit. Among them, the error image management unit is used to manage the images with non-random errors in the historical shooting records. The image quality evaluation unit is used to calculate the image quality evaluation index in the images. The function fitting unit is used to perform function fitting on the image quality evaluation index of the error images to obtain the function of the image quality evaluation index changing with time. The error change database management unit is used to manage the error change database;

[0046] Further, the contrast comparison module includes: a global contrast comparison unit and a local contrast comparison unit. Among them, the global contrast comparison unit is used to compare the contrast of the complete image to obtain a contrast threshold, and the local contrast comparison unit is used to compare the contrast in a unit area of the image to obtain a first target unit area;

[0047] Further, the image slicing management module includes: a status management unit and an image area backtracking unit. Among them, the status management unit is used to obtain the status record of the drone's real-time flight, and the image area backtracking unit is used to backtrack the image of the first target unit area to obtain the image of the second target unit area;

[0048] Further, the image judgment module includes: a reference value management unit, an error judgment unit, and an information feedback unit. Among them, the reference value management unit is used to obtain an error reference value, the error judgment unit is used to judge the difference between the image quality evaluation index and the error reference value, and the information feedback unit is used to mark the image slices that meet the judgment conditions and send them to relevant management personnel.

[0049] Compared with the prior art, the beneficial effects of the present invention are as follows: By marking the corresponding relationship between the flight state of the drone, the movement of the picture, and the picture interference, a monitoring model for known non-random error interference is established. The picture of the on-board camera is monitored, and the picture quality is evaluated. Through two comparisons, from two aspects of the change characteristics of the picture affected by interference and the comparison with the monitoring model, it is judged whether there is random interference in the picture, improving the accuracy of the judgment. Mark the picture areas that may have random interference, which provides convenience for further processing of video information. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a schematic structural diagram of a target feature random error analysis system based on cloud computing according to the present invention;

[0051] Figure 2 It is a schematic flow diagram of a target feature random error analysis method based on cloud computing according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0053] Embodiment: As Figure 1 and Figure 2As shown in the figure, the present invention provides a technical solution, a method for analyzing random errors of target features based on cloud computing:

[0054] Step S100: During the flight of the drone, record the flight state of the drone, correspond the flight state of the drone with the movement trajectory of the same object in the picture of the on-board camera, and collect to obtain a picture transfer database;

[0055] Among them, step S100 includes:

[0056] Step S101: Shoot a certain object through the on-board camera of the drone, take a certain object as the target object, divide the shooting picture of the on-board camera of the drone into several unit areas, establish a plane coordinate system for the shooting picture, and number each unit area according to the coordinates of the unit area;

[0057] Step S102: Obtain the video record of the on-board camera shooting the target object, and intercept a section of the video record as the first target video record;

[0058] Step S103: Obtain the flight attitude and flight speed of the drone when shooting the first target video record and record them in the first state set R1. From two frames in the first target video record, obtain the coordinates ref of the unit area where the target object is located in the first frame 1 , and the corresponding coordinates ref in the second frame 2 ;

[0059] Step S104: In the plane coordinate system, calculate the vector coordinates α of the indication vector, α = ref 2 - ref 1 , associate the indication vector with the first state set R1 to form a picture transfer data group, adjust the flight attitude and flight speed of the drone, and collect several picture transfer data groups to obtain a picture transfer database;

[0060] In the embodiment, in the coordinate system, a unique rectangular coordinate is given to each unit area, and a two-dimensional vector is used to mark the exponential vector, ref 2 =(5, 5), ref 1 =(3, 2), so α=(2, 3).

[0061] Step S200: Obtain the historical shooting record of the on-board camera, intercept the video segment with non-random errors in the picture from the shooting record, fit the function of the picture features of the non-random errors changing with time, and collect them into the error change database;

[0062] Among them, step S200 includes:

[0063] Step S201: obtaining a video clip with an error picture from a video record shot by an onboard camera, and intercepting a section of the video record as a second target video record;

[0064] Step S202: Obtain n frames from the second target video record, mark the unit area with picture errors in each frame, collect the unit areas with picture errors in the kth frame to form an error area, the error area is composed of adjacent unit areas, obtain the image of the error area, and record it as the error image im of the kth error area k ;

[0065] Step S203: Gathering error images of each frame, arranging the errors according to the order of the pictures in the second target video record to obtain an error image sequence, calculating the image quality evaluation index of each error image, and arranging the error images according to the order to obtain an error value sequence;

[0066] Among them, the calculation method of the image quality evaluation index includes:

[0067] 2-1: Convert the image into a grayscale image, set u grayscale level judgment levels to quantify the grayscale image, and count the occurrence probability of each grayscale level in the grayscale image;

[0068] 2-2: Get the probability p of the e-th gray level e , calculate the information entropy H of the grayscale image,

[0069]

[0070] 2-3: Use information entropy as the image quality evaluation indicator of grayscale images.

[0071] Step S204: Perform function fitting on the image quality evaluation index sequence to obtain a function of the image quality evaluation index changing with time, record the function as an error change function, obtain the flight posture and flight speed of the drone when shooting the second target video record and record them in the second state set R2, form an error change group with the error change function and the second state set R2, collect the corresponding relationship between several groups of error change functions and the second state set R2, collect the error change groups, and obtain an error change database.

[0072] Step S300: extracting a picture frame from the picture taken by the onboard camera, and extracting an image area with abnormal contrast change in the video frame;

[0073] Wherein, step S300 includes:

[0074] Step S301: Obtain a first image and a second image from the real-time video captured by the on-board camera of the drone in the shooting order of the video. Calculate the contrast ratios of the first image and the second image respectively, denoted as CR1 and CR2. When CR1 > CR2, calculate the contrast threshold M, where M = CR1 - CR2;

[0075] In the embodiment, the method for calculating the contrast ratio includes:

[0076] By obtaining the brightness values of each pixel point in the image, where the brightness value at the image position (x, y) is denoted as I(x, y);

[0077] Calculate the mean value μ of the brightness in the image, , where N represents the total number of pixels in the x-axis direction of the image, and M represents the total number of pixels in the y-axis direction of the image;

[0078] Calculate the standard deviation σ of the pixels in the image, ;

[0079] Take the standard deviation σ as the contrast ratio of the image;

[0080] Step S302: Obtain the image slice corresponding to the j-th unit area in the first image, denoted as psc 1j , and the image slice corresponding to the j-th unit area in the second image, denoted as psc 2j , obtain the contrast ratio of psc 1j , denoted as r 1j , and the contrast ratio of psc 2j , denoted as r 2j ;

[0081] Step S303: When r 1j - r 2j > M, mark the j-th unit area in the second image as the first target unit area.

[0082] Step S400: Obtain the record of the real-time flight state of the drone, extract the area of the shooting object in the shooting frame in the image area one unit time ago, and obtain the image slices of the two image areas respectively;

[0083] Among them, Step S400 includes:

[0084] Step S401: Obtain the timestamps of the first image and the second image captured by the on-board camera of the drone. Denote the time period from shooting the first image to shooting the second image as the target time period. Obtain the flight attitude and flight speed of the drone during the target time period and record them in the target state set RT;

[0085] Step S402: Obtain the first state set that is the same as the target state set RT in the screen transfer database, and obtain the vector coordinates β of the indication vector associated with the first state set;

[0086] Step S403: Obtain the coordinates ref of the first target unit area t1 , calculate the coordinates ref of the second target unit area t2 , ref t2 = ref t1 - β, and the second target unit area is in the first image;

[0087] For example, when the coordinates ref t1 of the first target unit area = (7, 5), and the obtained indication vector is (2, 3), ref t2 = ref t1 - β = (5, 2), take the unit area with coordinates (5, 2) in the first image as the second unit area;

[0088] Step S404: Obtain the image slice of the first target unit area and denote it as psc 1 and the image slice of the second target unit area and denote it as psc 2 , respectively obtain the image quality evaluation indexes of psc 1 and psc 2 , and denote them as rq 1 and rq 2 .

[0089] Step S500: Compare the image errors of the two image slices. When the difference between the image error and the non-random error is greater than the threshold, send the image slices to the relevant personnel;

[0090] Among them, Step S500 includes:

[0091] Step S501: Obtain the second state set that is the same as the target state set RT in the error change database, and obtain the error change function associated with the second state set;

[0092] Step S502: Mark the data points with values equal to the image quality evaluation index rq 2 in the error change function, denote the data points as the first data points, and in the error change function, obtain the data points passing through the time length of the target time period and denote them as the second data points;

[0093] Step S503: Obtain the image quality evaluation index corresponding to the second data point, and use the image quality evaluation index as the error reference value;

[0094] Step S504: Set the error threshold γ. When |Q - rq 1 | > γ, the image slice psc1 Make markings and send them to relevant management personnel.

[0095] The system includes: a screen movement management module, an error management module, a contrast comparison module, an image slicing management module, and an image judgment module;

[0096] Among them, the screen movement management module is used to manage the relative motion relationship between the objects and the viewing angle in the images captured by the airborne camera under different flight states. Among them, the screen movement management module is used to manage the relative motion relationship between the objects and the viewing angle in the images captured by the airborne camera under different flight states. The error management module is used to manage the screen features of non-random errors. The contrast comparison module is used to compare the contrast changes in the image frames to obtain the first target unit area. The image slicing management module is used to obtain the aircraft state records of the UAV and manage the image slices. The image judgment module is used to judge whether there are non-random errors in the image slices;

[0097] Among them, the error management module is used to manage the screen features of non-random errors. Among them, the error management module includes: an error image management unit, an image quality evaluation unit, a function fitting unit, and an error change database management unit. Among them, the error image management unit is used to manage the images with non-random errors in the historical shooting records. The image quality evaluation unit is used to calculate the image quality evaluation index in the image. The function fitting unit is used to perform function fitting on the image quality evaluation index of the error image to obtain the function of the image quality evaluation index changing with time. The error change database management unit is used to manage the error change database;

[0098] Among them, the contrast comparison module is used to compare the contrast changes in the image frames to obtain the first target unit area. Among them, the contrast comparison module includes: a global contrast comparison unit and a local contrast comparison unit. Among them, the global contrast comparison unit is used to compare the contrast of the complete image to obtain the contrast threshold. The local contrast comparison unit is used to compare the contrast in the unit area of the image to obtain the first target unit area;

[0099] Among them, the image slicing management module is used to obtain the aircraft state records of the UAV and manage the image slices. Among them, the image slicing management module includes: a state management unit and an image area backtracking unit. Among them, the state management unit is used to obtain the real-time flight state records of the UAV. The image area backtracking unit is used to backtrack the image of the first target unit area to obtain the image of the second target unit area;

[0100] Among them, the image judgment module is used to judge whether there are non-random errors in the image slices. Among them, the image judgment module includes: a reference value management unit, an error judgment unit, and an information feedback unit. Among them, the reference value management unit is used to obtain the error reference value, the error judgment unit is used to judge the difference between the image quality evaluation index and the error reference value, and the information feedback unit is used to mark the image slices that meet the judgment conditions and send them to the relevant management personnel.

[0101] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above-described exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention. Any reference signs in the claims should not be construed as limiting the claims involved.

Claims

1. A target feature random error analysis method based on cloud computing, characterized by: Methods include: Step S100: During the flight of the drone, the flight state of the drone is recorded, and the flight state of the drone is matched with the motion trajectory of the same object in the picture of the onboard camera to obtain a picture transfer database; Step S200: Obtain historical shooting records of the onboard camera, capture video clips with non-random errors in the pictures from the shooting records, fit a function of the picture features of the non-random errors changing over time, and compile them into an error change database; Step S300: extracting picture frames from the pictures taken in real time by the onboard camera, and extracting image areas with abnormal contrast changes in the video frames; Step S300 includes: Step S301: from the real-time video captured by the onboard camera of the drone, a first image and a second image are obtained in the order of video capture, and the contrasts of the first image and the second image are calculated, respectively, and are recorded as CR1 and CR2. When CR1>CR2, a contrast threshold M is calculated, where M=CR1-CR2. Step S302: Obtain the image slice corresponding to the jth unit area in the first image, denoted as psc 1j , the image slice corresponding to the jth unit area in the second image is denoted as psc 2j , get psc 1j The contrast is denoted as r 1j ,psc 2j The contrast is denoted as r 2j ; Step S303: When r 1j -r 2j >M, the jth unit area in the second image is marked as the first target unit area, and the first target unit area is regarded as the image area with abnormal contrast change; Step S400: obtaining a record of the real-time flight status of the drone, extracting the area of ​​the photographed object in the image area in the photographed picture one unit time ago, and obtaining image slices of the two image areas respectively; Step S400 includes: Step S401: obtaining the timestamps of the first image and the second image captured by the drone's onboard camera, recording the time period from capturing the first image to capturing the second image as the target time period, obtaining the flight attitude and flight speed of the drone in the target time period, and recording them in the target state set RT; Step S402: acquiring a first state set identical to the target state set RT in a picture transfer database, and acquiring a vector coordinate β of an indicator vector associated with the first state set; Step S403: Obtain the coordinates ref of the first target unit area t1 , calculate the coordinates of the second target unit area ref t2 , ref t2 = ref t1 -β, the second target unit area is in the first image; Step S404: obtaining an image slice of the first target unit area, denoted as psc1, and an image slice of the second target unit area, denoted as psc2, and obtaining image quality evaluation indicators of psc1 and psc2, denoted as rq1 and rq2 respectively; Step S500: comparing the image errors of two image slices, and when the difference between the image error and the non-random error is greater than a threshold, sending the image slices to relevant personnel.

2. The target feature random error analysis method based on cloud computing according to claim 1 is characterized by: Step S100 includes: Step S101: photographing an object through an onboard camera of a drone, taking the object as a target object, dividing the photographing picture of the onboard camera of the drone into a plurality of unit areas, establishing a plane coordinate system for the photographing picture, and numbering each unit area according to the coordinates of the unit area; Step S102: obtaining a video record of a target object shot by an onboard camera, and intercepting a segment of the video record as a first target video record; Step S103: Obtain the flight attitude and flight speed of the drone when shooting the first target video record and record them in the first state set R1, wherein the flight attitude of the drone includes: roll, pitch and yaw. From two frames of the first target video record, obtain the coordinates ref1 of the unit area where the target object is located in the first frame and the corresponding coordinates ref2 in the second frame; Step S104: In the plane coordinate system, calculate the vector coordinate α of the indication vector, α= ref2- ref1, associate the indication vector with the first state set R1 to form a picture transfer data group, adjust the flight attitude and flight speed of the drone, collect several picture transfer data groups, and obtain a picture transfer database.

3. The target feature random error analysis method based on cloud computing according to claim 2 is characterized by: Step S200 includes: Step S201: acquiring a video segment with an error picture from a video record shot by an onboard camera, and intercepting a segment of the video record as a second target video record; Step S202: Obtain n frames from the second target video record, mark the unit area with picture errors in each frame, collect the unit areas with picture errors in the kth frame to form an error area, the error area is composed of adjacent unit areas, obtain the image of the error area, and record it as the error image im of the kth error area k ; Step S203: Gathering error images of each frame, arranging the errors according to the order of the pictures in the second target video record to obtain an error image sequence, calculating the image quality evaluation index of each error image, and arranging the error images according to the order to obtain an error value sequence; Step S204: Perform function fitting on the image quality evaluation index sequence to obtain a function of the image quality evaluation index changing with time, record the function as the error change function, obtain the flight posture and flight speed of the drone when shooting the second target video record and record them in the second state set R2, form an error change group with the error change function and the second state set R2, collect the corresponding relationship between several groups of error change functions and the second state set R2, collect the error change groups, and obtain the error change database.

4. The target feature random error analysis method based on cloud computing according to claim 3 is characterized by: The calculation method of image quality evaluation index includes: 2-1: Convert the image into a grayscale image, set u grayscale level judgment levels to quantify the grayscale image, and count the occurrence probability of each grayscale level in the grayscale image; 2-2: Get the probability p of the e-th gray level e , calculate the information entropy H of the grayscale image, ； 2-3: Use information entropy as an image quality evaluation indicator for the grayscale image.

5. The target feature random error analysis method based on cloud computing according to claim 4 is characterized in that: Step S500 includes: Step S501: acquiring a second state set identical to the target state set RT in an error change database, and acquiring an error change function associated with the second state set; Step S502: Marking a data point whose value is equal to the image quality evaluation index rq2 in the error change function, recording the data point as a first data point, and obtaining a data point of a time length of a target time period in the error change function as a second data point; Step S503: obtaining an image quality evaluation index corresponding to the second data point, and using the image quality evaluation index as an error reference value Q; Step S504: setting an error threshold γ, when |Q-rq1|>γ, marking the image slice psc1 and sending it to the relevant management personnel.

6. A target feature random error analysis system based on cloud computing, used to execute a target feature random error analysis method based on cloud computing according to any one of claims 1 to 5, characterized in that: The system includes: The invention relates to a picture movement management module, an error management module, a contrast comparison module, an image slice management module and an image judgment module, wherein the picture movement management module is used to manage the relative motion relationship between objects and viewing angles in the pictures taken by the onboard camera under different flight states, the error management module is used to manage the picture features of non-random errors, the contrast comparison module is used to compare the changes in contrast in the picture frames and obtain the first target unit area, the image slice management module is used to obtain the aircraft status record of the UAV and manage the image slices, and the image judgment module is used to judge whether there are non-random errors in the image slices.

7. The target feature random error analysis system based on cloud computing according to claim 6, characterized in that: The picture movement management module includes: a picture management unit, an indication vector management unit and a picture transfer database management unit, wherein the picture management unit is used to manage the unit management area in the picture taken by the airborne camera, the indication vector management unit is used to mark the relative motion relationship of the object taken in the state of the drone relative to the picture taken, and obtain the indication vector, and the picture transfer database management unit is used to manage the picture transfer database; The error management module includes: an error image management unit, an image quality evaluation unit, a function fitting unit and an error change database management unit, wherein the error image management unit is used to manage the pictures with non-random errors in the historical shooting records, the image quality evaluation unit is used to calculate the image quality evaluation index in the image, the function fitting unit is used to perform function fitting on the image quality evaluation index of the error image to obtain a function of the image quality evaluation index changing with time, and the error change database management unit is used to manage the error change database.

8. The target feature random error analysis system based on cloud computing according to claim 6, characterized in that: The contrast comparison module includes: a global contrast comparison unit and a local contrast comparison unit, wherein the global contrast comparison unit is used to compare the contrast of the complete image and obtain a contrast threshold, and the local contrast comparison unit is used to compare the contrast in a unit area in the image and obtain a first target unit area; The image slice management module includes: a state management unit and an image area tracing unit, wherein the state management unit is used to obtain the real-time flight state record of the UAV, and the image area tracing unit is used to trace the image of the first target unit area to obtain the image of the second target unit area; The image judgment module includes: a reference value management unit, an error judgment unit and an information feedback unit, wherein the reference value management unit is used to obtain an error reference value, the error judgment unit is used to judge the difference between the image quality evaluation index and the error reference value, and the information feedback unit is used to mark the image slices that meet the judgment conditions and send them to relevant managers.

Images