Wide dynamic unmanned aerial vehicle imaging screening evaluation method and system

Abstract

The invention discloses a wide dynamic unmanned aerial vehicle imaging screening evaluation method and system, and the method comprises the steps: building an unmanned aerial vehicle image evaluation system based on a monitoring image preset standard, and carrying out the image collection of a monitoring region through an unmanned aerial vehicle; performing OECF, resolution, color, shadow and distortion evaluation on the acquired image through an evaluation system; inputting the preprocessed image data into a neural network, and performing feature extraction on the image data acquired in real time based on the neural network; normalization processing is carried out on the system extraction characteristic values, and an image grade model is built. Through evaluation results of five indexes of ECF, resolution, color, shadow and distortion of an unmanned aerial vehicle image, screening transmission is carried out on the unmanned aerial vehicle shot image in combination with multi-index weight analysis, so that the data volume required to be processed by subsequent analysis can be greatly reduced, the overall working process is accelerated, and the working efficiency is improved. And the hard disk capacity of the unmanned aerial vehicle or cloud storage resources in other forms can be saved.

Description

Technical Field

[0001] The present invention relates to the technical field of visible light imaging evaluation, and in particular to a wide dynamic range unmanned aerial vehicle imaging screening evaluation method and system. Background Art

[0002] At present, with the widespread application of drones and other aircraft in various industries, air-to-ground visual imaging has become one of the core functions of drones, and its imaging quality is the basis for drones to perform intelligent functions such as target recognition and tracking. The air-to-ground visual imaging process may encounter mist, clouds and poor lighting conditions. The dynamic range of the imaging environment is wide, and the target imaging will be over-exposed, under-exposed and color imbalanced. In addition, affected by aerodynamic forces, air-to-ground imaging is accompanied by severe jitter, which makes the collected images produce a lot of random noise. Therefore, the images collected by air-to-ground sensors need to be screened and processed to obtain high-quality images. The quality of the processed images depends on the imaging clarity under different illuminations. The clarity here refers to high-quality images with good resolution, accurate color, wide dynamic range and small imaging distortion. Among them, the wide dynamic range enables the normal capture of targets at different time periods during the day, the accurate color can find more target image information, the good resolution is to find more texture details of the target, and the small imaging distortion is to ensure that the target image is not deformed or distorted, so as to better serve target recognition and tracking. Therefore, an objective testing and evaluation method is proposed for the evaluation of UAV wide dynamic range air-to-ground imaging. Summary of the invention

[0003] The wide dynamic range unmanned aerial vehicle imaging screening and evaluation method and system proposed in the present invention can solve at least one of the technical problems in the background technology.

[0004] To achieve the above object, the present invention adopts the following technical solutions:

[0005] A wide dynamic range unmanned aerial vehicle imaging screening and evaluation method, comprising the steps of:

[0006] S100, build a drone image evaluation system based on the preset standards of monitoring images, and use drones to collect images of the monitoring area;

[0007] S200, respectively evaluating the OECF, resolution, color, shadow, and distortion of the collected image through an evaluation system;

[0008] S300, inputting the preprocessed image data into a neural network, and performing feature extraction on the real-time collected image data based on the neural network;

[0009] S400, normalizing the feature values ​​extracted by the system, and building an image grade model based on the normalization results.

[0010] Furthermore, the method for building a drone image evaluation system based on preset standards of surveillance images in step S100 of the present invention is: performing preset standard detection through OECF test model, resolution test model, color test model, shadow test model, and distortion test model, and building a drone image evaluation system based on the detection results.

[0011] Furthermore, the color test model of the present invention performs a detection method for a preset mark as follows:

[0012] The corresponding test environment and TE42 chart are used for testing. The color blocks in the chart are distributed around the grayscale blocks. Each color block has a default value for test calculation. Under the influence of light sources in different environments, colors will appear to be the same color but different spectra, causing people to see incorrect colors. The graphic results intuitively reflect the camera's ability to truly restore each color block image.

[0013] Furthermore, the method for evaluating the shadow test model described in the present invention is: by verifying and evaluating the brightness shadow and color shadow of the drone image, measuring 24 ROIs evenly distributed on the image of the TE42 chart, and obtaining the shadow intensity in the aperture and the color shadow in ΔEab; the graphic results intuitively reflect the brightness shadow and color shadow, and the imaging color and shadow can be tested according to the correction coefficient.

[0014] Furthermore, the distortion evaluation method in step S200 of the present invention is specifically as follows: using a TE251 chart, which contains 13 rows×23 columns of small crosses and 4 mark points, for geometric distortion testing of normal angle lenses;

[0015] The geometric distortion of the lens is divided into four quadrants based on the center of the image: Q1, Q2, Q3, and Q4. When testing the distortion, the cross is divided into the four quadrants of the image;

[0016]

[0017] Where H is the position of the point without distortion, and H' is the distance from the point to the center of the image. The geometric distortion of a grid position is the value between the radial distance H' of the actual grid position divided by the ideal grid position H and the ideal grid position H. H' less than H indicates negative distortion (barrel distortion), and H' greater than H indicates positive distortion (pincushion distortion).

[0018] Furthermore, in step S400 of the present invention, the system extracts feature values ​​for normalization processing, and the method for building an image level model is:

[0019] S410, mapping the characteristic values ​​extracted by the evaluation system to [0,1] through the Min-Max Scaling formula;

[0020] The Min-Max Scaling formula is:

[0021]

[0022] Among them, x i is the test index value series, x min is the minimum value of the test index, x max is the maximum value of the test index, X i 'It is the characteristic value after the test index value is mapped;

[0023] S420, normalizing the normalized data using a Z-score formula;

[0024]

[0025] where w i is the standardized index, S is the standard deviation of each index series;

[0026] S430, weighting the processing results of each normalized index to build a grade model;

[0027] Based on the subjective judgment of human vision on image quality, a quantitative weighted evaluation model is established, in which resolution accounts for 40%, OECF accounts for 25%, color accounts for 15%, shadow accounts for 10%, and distortion accounts for 10%. The evaluation model is as follows:

[0028]

[0029] Among them, q i is the weight of the ith indicator, w i is the i-th normalized index.

[0030] In yet another aspect, a wide dynamic range unmanned aerial vehicle imaging screening and evaluation system includes a data collection module, a data preprocessing module, an index calculation module, and a visualization module;

[0031] The data collection module is used to collect image data collected by the drone and transmit it to the data preprocessing module;

[0032] The data preprocessing module processes the data collected by the drone and transmits the processed data to the index calculation module;

[0033] The indicator calculation module transmits the data to the visualization module to generate a data evaluation curve chart.

[0034] It can be seen from the above technical scheme that the present invention discloses a wide dynamic UAV imaging screening and evaluation method and system. The method builds a UAV image evaluation system based on the preset standard of monitoring images, uses a UAV to collect images of the monitoring area; evaluates the OECF, resolution, color, shadow, and distortion of the collected images through the evaluation system; inputs the pre-processed image data into the neural network, and extracts features of the real-time collected image data based on the neural network; normalizes the feature values ​​extracted by the system, and builds an image grade model. By combining the evaluation results of the five indicators of ECF, resolution, color, shadow, and distortion of the UAV image with multi-indicator weight analysis to screen and transmit the UAV-shot images, the amount of data required for subsequent analysis can be greatly reduced, thereby speeding up the overall workflow and helping to save the UAV's hard disk capacity or other forms of cloud storage resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flowchart of the UAV imaging assessment method of the present invention;

[0036] Figure 2 It is a wide dynamic COMS imaging index system;

[0037] Figure 3 It is a schematic diagram of the dynamic range curve;

[0038] Figure 4 It is the CMOS wide dynamic index test environment;

[0039] Figure 5 This is a schematic diagram of the picture card test and the test effect of the present invention;

[0040] Figure 6 This is a schematic diagram of the test chart and test environment;

[0041] Figure 7 Schematic diagram of the Siemens star used for measurement;

[0042] Figure 8 Schematic diagram of the bevel edge used for measurement;

[0043] Fig. 9 Identify schematic diagrams for color patches;

[0044] Fig.10 Numerical results and visual comparison for color testing;

[0045] Fig.11 This is the TE42 card and identification diagram;

[0046] Fig.12 This is a distortion chart and identification diagram;

[0047] Fig.13 This is a schematic diagram of the distortion quadrant division;

[0048] Fig.14 Schematic diagram of lens geometric distortion;

[0049] Fig.15 This is a schematic diagram of the optical axis calibration environment;

[0050] Fig.16 Schematic diagram of UAV indoor imaging test. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments.

[0052] like Figure 1 As shown, the wide dynamic range drone imaging screening and evaluation method and system described in this embodiment performs the following steps through a computer device:

[0053] S100, build a drone image evaluation system based on the preset standards of monitoring images, and use drones to collect images of the monitoring area;

[0054] S200, respectively evaluating the OECF, resolution, color, shadow, and distortion of the collected image through an evaluation system;

[0055] S300, inputting the preprocessed image data into a neural network, and performing feature extraction on the real-time collected image data based on the neural network;

[0056] S400, normalizing the feature values ​​extracted by the system, and building an image grade model based on the normalization results.

[0057] The following is a detailed description of each step:

[0058] S100, build a drone image evaluation system based on the preset standards of monitoring images, and use drones to collect images of the monitoring area;

[0059] The clarity standard of surveillance images is usually determined by multiple factors, including but not limited to resolution, frame rate, compression ratio, etc. These parameters jointly determine the quality of the image and its applicability in specific application scenarios. The present invention builds a drone image evaluation system through OECF, resolution, color, shadow, and distortion test models to evaluate the drone surveillance area images. The specific imaging index system is as follows: Figure 2 shown.

[0060] The optoelectronic conversion function (OECF) describes the characteristics of digital cameras in converting brightness into numerical values. It can verify and evaluate the wide dynamic performance of drone camera imaging. The calculation models of its various indicators are as follows:

[0061] (1) Dynamic range (DR)

[0062] Dynamic range (DR) describes the maximum scene contrast that a camera can restore. The brightest point that a camera can reach is the maximum output value. According to ISO15739, the SNR value of the darkest point is 1. The dynamic range is defined on the basis of SNR 1. The dynamic range is the contrast between the brightest point and the darkest point.

[0063] like Figure 3 As shown in the figure, the dotted line is the signal-to-noise ratio (SNR) curve, the solid line is the RGB curve, and the two vertical lines are the dynamic range curves. The left vertical line is obtained by the logarithmic value of the brightness when the signal-to-noise ratio is 1, and the right vertical line is obtained by the logarithmic value of the brightness at the intersection of the RGB curve fitting. The dynamic range value of the camera is obtained according to the two red lines.

[0064] The dynamic range calculation formula has the following relationship:

[0065]

[0066] Among them, Lsat and Lmin are the set brightness value and the minimum brightness value.

[0067] (2) Signal-to-noise ratio (SNR)

[0068] The signal-to-noise ratio describes the relationship between the signal and the noise and is defined as the ratio of the signal value to the standard deviation of the signal value. The noise is displayed as the standard deviation. The standard deviation is defined as the root mean square of the absolute value of the sum of the standard deviations within the signal range. The larger the SNR, the lower the camera noise. The signal-to-noise ratio is determined by the following formula:

[0069]

[0070]

[0071] L SNR =0.13×R ref

[0072] R ref =S -1 (I)| I＝245

[0073] Among them, σ is the standard deviation of noise, gain is the increment, L SNR For the reference exposure R at a brightness value of 13% ref The time signal under measurement noise ratio, R refis the reference exposure, which is the logarithmic value of the exposure brightness corresponding to a pixel value of 245 on the focal plane.

[0074] (3) Visual Noise (VN)

[0075] Compared to the signal-to-noise ratio, different observers have different perceptions of visual noise, and the CIE-Luv color space has better visual perception characteristics than the CIE-Lab color space. To this end, the RGB image is converted to the XYZ color space, and the color receptors are analyzed based on the results of the human eye and the processing of color signals. The visual noise value VN is the weighted sum of the standard deviations of the △L*, △u*, and △v* channels in the CIE-Luv color space.

[0076] In the visual noise calculation, the default observation conditions are 100% observation on the display, observation distance 0.5m, and display resolution 96ppi. VN is calculated according to the following formula:

[0077] VN=1.000ΔL*+0.8520Δu*+0.3230Δv*

[0078] Based on the above indicator measurement model, an ideal outdoor simulation test environment is built. Figure 4 As shown:

[0079] The test environment is a dark room without stray light, the light box is an LG3 transmissive light box with adjustable brightness, the TE270X V2D (100000:1) chart contains a circle of 20 grayscale blocks, and the dynamic test range can reach 80db. The test process is as follows:

[0080] S131: Place the chart card in the card slot of the LG3 light box, turn on the light box, set the light box to the NORM working mode, and the light source illumination unit is lux;

[0081] S132: Fix the camera on the mobile bracket, adjust the relative position of the camera and the image card, make the optical axis perpendicular to the image card, and make the image of the image card located at the center of the field of view and occupy about 1 / 2 to 2 / 3 of the camera field of view;

[0082] S133: placing the shading plate between the chart and the camera so that the camera field of view just passes through the light hole in the middle of the shading plate to prevent other stray light from affecting the test results;

[0083] S134: Observe the chart image, with the bright area of ​​the chart on the top, and adjust the polarization angle in the middle of the chart to change the center brightness. The top is the darkest, and 90 degrees is the brightest. By changing the center brightness, the exposure can be normal.

[0084] S135: Collect the image of the picture card, and detect the background value of the picture card in real time according to the set illumination value, and check the test results;

[0085] S136: Adjust the illumination value of the light box, repeat S134 to S135, and obtain multiple groups of test results.

[0086] The resolution test method is specifically as follows:

[0087] Resolution test includes Siemens star / MTF, slant edge / SFR. TE42 chart can cover all the above test indicators. The test environment layout is shown in Figure 6:

[0088] (1) Siemens Star Test

[0089] like Figure 7 As shown in the figure, during the test, a full Siemens star chart is used in the center of the chart and half Siemens stars are used in the four corners. A Siemens star is divided into multiple parts, with 8 parts in the center and 3 parts in the corners. The average value of each part of each star contains information on imaging sharpness and resolution at different positions and directions, which is used to check the optical center of the camera, analyze optical errors such as astigmatism, verify image processing in a specific direction or at a specific frequency, and determine the limiting resolution.

[0090] During the test, the limiting resolution is calculated from the MTF data of the specified frequency modulation. The modulation transfer function (MTF) describes the modulation loss of the spatial frequency. It is a response of the imaging system on the spatial frequency, representing that the modulation starts from 0 line pairs and continues to decrease at a given spatial frequency. As the line pair density increases, the difference between the black and white lines of the line pair decreases, and the MTF decreases. The Siemens Star is simulated by radial lines of different frequencies.

[0091] (2) Bevel

[0092] like Figure 8 As shown in the figure, the bevel used for measurement is mainly used to detect the sharpness of the camera. The TE42 chart has 4 bevels, with two different contrasts for horizontal and vertical. Different contrasts are used in the sharpening algorithm in image signal processing to detect the contrast of the bevel and adjust the sharpening based on this information. The test results include two modulation contrast deviations with different contrasts.

[0093] The color evaluation method is specifically as follows:

[0094] like Fig. 9 As shown, the color test mainly verifies and evaluates the camera's color reproduction ability. Figure 6 The environment layout and TE42 chart are used for testing. The color blocks in the chart are distributed around the grayscale blocks. Each color block has a default value for test calculation.

[0095] like Fig.10As shown in the figure, under the influence of light sources in different environments, colors will appear metamerism, causing the human eye to see incorrect colors. This phenomenon can be alleviated by adaptively adjusting the standard. The graphic results can intuitively reflect the camera's ability to truly restore each color block image.

[0096] The shadow evaluation method is specifically as follows:

[0097] The shadow test is mainly used to verify and evaluate the brightness shadow and color shadow of the camera. It measures 24 ROIs evenly distributed on the image of the TE42 chart to obtain the shadow intensity in the aperture and the color shadow in ΔEab.

[0098] The graphic results directly reflect the brightness shadow and color shadow, and the imaging color and shadow can be tested according to the correction coefficient.

[0099] The distortion evaluation method is specifically as follows:

[0100] The distortion test is mainly used to verify and evaluate the distortion and chromatic aberration of camera imaging. The TE251 chart is used. The chart contains 13 rows × 23 columns of small crosses and 4 mark points. It is used for geometric distortion testing of normal-angle lenses.

[0101] like Fig.13 As shown, the geometric distortion of the lens is divided into four quadrants with the center of the image: Q1, Q2, Q3, and Q4.

[0102] When testing for distortion, the crosses are divided into 4 quadrants of the image.

[0103]

[0104] Where H is the position of the point without distortion, and H' is the distance from the point to the center of the image. The geometric distortion of a grid position is the value between the radial distance H' of the actual grid position divided by the ideal grid position H and the ideal grid position H. H' less than H indicates negative distortion (barrel distortion), and H' greater than H indicates positive distortion (pincushion distortion).

[0105] S151: Adjust the position of the guide rail and the focal length of the lens so that the camera faces the chart directly and ensures that the camera does not tilt in any direction;

[0106] S152: Cooperate with the illuminance meter to adjust the light source in real time to provide uniform fill light for the chart, to ensure that the illumination and intensity of the four corners and the center of the chart are not much different, and the difference between the center and the four corners cannot exceed 10%, so that the image is clear and discernible;

[0107] S153: Fix the camera on the moving guide rail, turn on the power supply and the display (set the cross reference), adjust the relative position and posture of the camera and the chart card, so that the center of the chart card image coincides with the center reference of the display, and the image frame meets the chart card test requirements;

[0108] S154: Collect the image of the picture card and check the test results;

[0109] S155: Adjust the light source to different illuminations, repeat the measurement, and obtain multiple groups of test results;

[0110] S156: Selecting image results that meet the requirements as a standard model for the evaluation system based on multiple groups of test results.

[0111] By testing and verifying the OECF, resolution, color, shadow, and distortion of drone camera imaging, we can obtain various indicators that affect camera imaging, and use this to verify and evaluate the imaging effect. The wide dynamic drone camera test indicator system is as follows:

[0112] Among them, OECF includes three indicators: dynamic range DR, signal-to-noise ratio SNR, and visual noise VN. The larger the DR value, the wider the dynamic range; the larger the SNR, the less noise the camera has; the smaller the VN value, the less noise. Resolution is evaluated by Siemens Star and Slant Edge test indicators. The closer the modulation transfer function MTF and visual modulation transfer function vMTF are to 1, the better, and the smaller the edge width Edge-Width, the better. Color is measured by color distance ΔE, brightness ΔL, chromaticity ΔC, hue ΔH, and visual noise VN. The smaller the average color distance ΔE value, the better the result and the more standard the color. The other indicators are similar. Shadow is measured by brightness shadow Shading, color shadow average CIE ΔEab, maximum difference in visual noise ΔVN, and optical center. The smaller the measurement value, the better. The distortion is evaluated by lens geometric distortion LGD, chromatic aberration CA, and longitudinal chromatic aberration LCA. The closer LGD is to 0, the smaller the distortion. Less than 0 is barrel distortion, and greater than 0 is pincushion distortion. The average chromatic aberration of the GR / GB channel is closer to 0, the better.

[0113] First, use a collimator to calibrate the optical axis of the camera's CMOS target surface and the lens, so that the two optical axes coincide and are parallel to the mechanical axis, ensure that the camera is correctly installed, adjust the camera focal length, and meet the imaging effect required by the test scene. System errors are negligible. During the calibration process, ensure that the crosshairs in the monitor coincide with the center of the "BMW image" in the graphic field of view, and subjectively clear imaging.

[0114] Secondly, the various indicators of the wide dynamic range drone camera are tested using transmissive and reflective charts.

[0115] like Fig.16 As shown, in order to complete the indoor imaging test of the drone, an indoor test environment is built using a sand table and simulated fluorescent lamps. The sand table has a fog-making function, and the simulated fluorescent lamp has functions such as adjusting color temperature and hue. The surrounding environment is decorated with light-absorbing paint to reduce the impact of external ambient light.

[0116] S200, use drones to collect images of the monitored area, and use the evaluation system to evaluate the OECF, resolution, color, and shadow of the collected images;

[0117] The OECF, resolution, color, and shading evaluation method includes the following steps:

[0118] S210, selecting a monitoring area, and using a drone to collect images of the monitoring area;

[0119] S220, the evaluation system pre-processes the image collected by the UAV based on the standard model;

[0120] The preprocessing steps for drone-collected images include:

[0121] S221, filtering the image data collected by the drone;

[0122] Set the center of the image to Q (center pixel), set the neighborhood set S of the center pixel, and p to the neighborhood set of the center pixel;

[0123]

[0124]

[0125] Where e is a natural constant; Ip represents the grayscale value of point p = (x1, y1) in image I, (x1, y1) represents the coordinates of the pixel point, Iq represents the grayscale value of point q = (u1, v1) in image I; σs represents the standard deviation of the spatial distance in the Gaussian function, σr represents the standard deviation of the pixel value in the Gaussian function; Gs represents the spatial distance weight, Gr represents the pixel value weight; BF Q represents the gray value of point Q in the filtered image, and Wq represents the weighted sum of the pixel values ​​at point q;

[0126] S222, establishing a linear mapping relationship of the image data based on the brightness in the filtered image data;

[0127] L OUT (x,y)=k×L in (x,y)

[0128] Among them, L in (x, y) represents the pixel value of the local pixel block in the image data before contrast enhancement, L OUT (x, y) represents the pixel value of the local pixel block in the image data after contrast enhancement, and k is the contrast gain parameter;

[0129] Set the brightness threshold in the image data;

[0130]

[0131] Among them, L avg represents the average brightness of the local pixel block in the image data; when the brightness in the image is lower than the set threshold, 0 < Q < 1 is satisfied, and when the brightness in the image is higher than or equal to the set threshold, Q > 1 is satisfied;

[0132] S223. Set a sliding window and enhance the contrast of pixels based on the average brightness and variance within the window;

[0133]

[0134] The average brightness and variance within the sliding window are respectively:

[0135]

[0136]

[0137] The enhancement coefficient η is:

[0138]

[0139] Among them, k represents a set constant, and η is the enhancement coefficient;

[0140] S300. Input the preprocessed image data into the neural network and extract features from the real-time collected image data based on the neural network;

[0141] S400. Normalize the feature values extracted by the system and build an image grading model based on the normalization result.

[0142] Since the measurement units of each measurement index are different, the Min-Max Scaling method is used for data normalization to eliminate the influence of the measurement units of each index. The specific method is as follows:

[0143] S410. Map the feature values extracted by the evaluation system to [0, 1] through the Min-Max Scaling formula;

[0144] The Min-Max Scaling formula is:

[0145]

[0146] Among them, x i is the sequence of test index values, x min is the minimum value of the test index, x max is the maximum value of the test index, and X i ’ is the feature value after mapping the test index value.

[0147] S420, normalizing the normalized data using a Z-score formula;

[0148] In the multi-index evaluation system, due to the different nature of each evaluation index, they usually have different orders of magnitude and units. If the original data is used directly, the role of the index with high values ​​in the analysis will be highlighted, and the role of the index with low values ​​will be relatively weakened. In order to ensure the reliability of the results, the mapped eigenvalues ​​are standardized;

[0149]

[0150] where w i is the standardized index, and S is the standard deviation of each index series.

[0151] S430, weighting the processing results of each normalized index to build a grade model:

[0152] After normalization of each test index, a dynamic evaluation system for wide dynamic range drone camera imaging quality was built. Based on the subjective judgment of human vision on image quality, a quantitative weighted evaluation model was established, in which resolution accounted for 40%, OECF accounted for 25%, color accounted for 15%, shadow accounted for 10%, and distortion accounted for 10%. The evaluation model is as follows:

[0153]

[0154] Among them, q i is the weight of the ith indicator, w i is the i-th normalized index.

[0155] On the other hand, the present invention also discloses a wide dynamic range unmanned aerial vehicle imaging screening and evaluation system, including a data collection module, a data preprocessing module, an index calculation module, and a visualization module;

[0156] The data collection module is used to collect image data collected by the drone and transmit it to the data preprocessing module;

[0157] The data preprocessing module processes the data collected by the drone and transmits the processed data to the index calculation module;

[0158] The indicator calculation module transmits the data to the visual module to generate a data evaluation curve chart.

[0159] In summary, the present invention provides a wide dynamic range unmanned aerial vehicle imaging screening and evaluation method and system. The unmanned aerial vehicle image evaluation system is built by monitoring image preset standards, and the unmanned aerial vehicle is used to collect images of the monitoring area; the collected images are evaluated by the evaluation system in terms of OECF, resolution, color, shadow, and distortion; the pre-processed image data is input into a neural network, and feature extraction is performed on the real-time collected image data based on the neural network; the feature values ​​extracted by the system are normalized to build an image grade model. By combining the evaluation results of the five indicators of ECF, resolution, color, shadow, and distortion of the unmanned aerial vehicle image with multi-indicator weight analysis to screen and transmit the images taken by the unmanned aerial vehicle, the amount of data required for subsequent analysis can be greatly reduced, thereby speeding up the overall workflow and helping to save the hard disk capacity of the unmanned aerial vehicle or other forms of cloud storage resources.

[0160] It is understandable that the system provided by the embodiment of the present invention corresponds to the method provided by the embodiment of the present invention, and the explanation, examples and beneficial effects of the relevant contents can refer to the corresponding parts of the above method.

[0161] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive Solid State Disk (SSD)), etc.

[0162] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0163] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0164] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A wide dynamic range unmanned aerial vehicle imaging screening and evaluation method, characterized in that: The specific steps include S100, build a drone image evaluation system based on the preset standards of monitoring images, and use drones to collect images of the monitoring area; S200, respectively evaluating the OECF, resolution, color, shadow, and distortion of the collected image through an evaluation system; S300, inputting the preprocessed image data into a neural network, and performing feature extraction on the real-time collected image data based on the neural network; S400, normalizing the feature values ​​extracted by the system, and building an image grade model based on the normalization results.

2. The wide dynamic range unmanned aerial vehicle imaging screening and evaluation method according to claim 1, characterized in that: The method for building a drone image evaluation system based on preset standards for surveillance images in step S100 is: performing preset standard detection through an OECF test model, a resolution test model, a color test model, a shadow test model, and a distortion test model, and building a drone image evaluation system based on the detection results.

3. The wide dynamic range unmanned aerial vehicle imaging screening and evaluation method and system according to claim 1, characterized in that: The color test model performs a detection method for a preset mark as follows: The corresponding test environment and TE42 chart are used for testing. The color blocks in the chart are distributed around the grayscale blocks. Each color block has a default value for test calculation. Under the influence of light sources in different environments, colors will appear to be the same color but different spectra, causing people to see incorrect colors. The graphic results intuitively reflect the camera's ability to truly restore each color block image.

4. The wide dynamic range unmanned aerial vehicle imaging screening and evaluation method and system according to claim 2, characterized in that: The shadow test model is evaluated by verifying and evaluating the brightness shadow and color shadow of the drone image, measuring 24 ROIs evenly distributed on the image of the TE42 chart, and obtaining the shadow intensity in the aperture and the color shadow in ΔEab; the graphic results intuitively reflect the brightness shadow and color shadow, and the imaging color and shadow can be tested according to the correction coefficient.

5. The wide dynamic range unmanned aerial vehicle imaging screening and evaluation method and system according to claim 2, characterized in that: The distortion evaluation method in step S200 is specifically as follows: using a TE251 chart, which contains 13 rows×23 columns of small crosses and 4 mark points, for testing the geometric distortion of a normal-angle lens; The geometric distortion of the lens is divided into four quadrants based on the center of the image: Q1, Q2, Q3, and Q4. When testing the distortion, the cross is divided into the four quadrants of the image; Where H is the position of the point without distortion, H′ is the distance from the point to the center of the image; the geometric distortion of a grid position is the value between the radial distance H′ of the actual grid position divided by the ideal grid position H and the ideal grid position H; H′ less than H indicates negative distortion (barrel distortion), and H′ greater than H indicates positive distortion (pincushion distortion).

6. The wide dynamic range unmanned aerial vehicle imaging screening and evaluation method and system according to claim 1, characterized in that: In step S400, the system extracts the feature values ​​for normalization processing, and the method for building the image level model is: S410, mapping the characteristic values ​​extracted by the evaluation system to [0,1] through the Min-Max Scaling formula; The Min-Max Scaling formula is: Among them, x i is the test index value series, x min is the minimum value of the test index, x max is the maximum value of the test index, X i 'It is the characteristic value after mapping the test index value; S420, normalizing the normalized data using a Z-score formula; where w i is the standardized index, S is the standard deviation of each index series; S430, weighting the processing results of each normalized index to build a grade model; Based on the subjective judgment of human vision on image quality, a quantitative weighted evaluation model is established, in which resolution accounts for 40%, OECF accounts for 25%, color accounts for 15%, shadow accounts for 10%, and distortion accounts for 10%. The evaluation model is as follows: Among them, qi is the i-th indicator weight, and wi is the i-th normalized indicator.

7. A wide dynamic range unmanned aerial vehicle imaging screening and evaluation system, characterized in that: It includes data collection module, data preprocessing module, indicator calculation module and visualization module; The data collection module is used to collect image data collected by the drone and transmit it to the data preprocessing module; The data preprocessing module processes the data collected by the drone and transmits the processed data to the index calculation module; The indicator calculation module transmits the data to the visualization module to generate a data evaluation curve chart.

Images