An urban surveillance image recognition system
Through the combination of light detection and feature analysis, frequency analysis and abnormal detection modules, image processing is optimized, and the recognition accuracy of the urban monitoring system in complex lighting environments is solved, the recognition accuracy and response speed are improved, and the monitoring capability is enhanced.
Patent Information
- Application Number
- CN202411301090.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-18
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-09-18
AI Technical Summary
The existing urban surveillance image recognition system has reduced recognition accuracy and availability under complex lighting environments and image quality issues, resulting in misjudgment and misjudgment, affecting the effectiveness of urban safety monitoring.
The light detection and feature analysis module is used to locate the brightness uneven area through histogram technology, calculate the light uneven index, adjust the brightness parameters of the target area, combine the frequency analysis and abnormality detection module to identify image abnormalities through Fourier transform, and use the Kriging interpolation method to repair it. Damaged area, optimize overall image details.
It improves the accuracy and response speed of image recognition, enhances the system's real-time monitoring and early warning capabilities, reduces the loss of key information, and improves the quality of urban management and security.
Smart Images

Figure CN118823292B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of image processing, and in particular to an urban monitoring image recognition system. Background Art
[0002] The field of image processing technology includes various technologies ranging from image enhancement and restoration to image recognition and analysis. The field focuses on how to improve image quality and how to extract useful information from images. Specific applications include image compression, denoising, edge detection, feature extraction, etc. In addition, by combining machine learning and artificial intelligence technologies, image processing is not limited to enhancing visual effects, but can also perform situational understanding, content classification, and dynamic event recognition. Image processing technology is widely used in many fields such as medical diagnosis, satellite remote sensing, industrial detection, and security monitoring.
[0003] Among them, the urban surveillance image recognition system refers specifically to a system that uses image processing technology to automatically analyze and identify videos or images captured in urban surveillance. The main purpose of the system is to improve urban safety and support urban management and security through functions such as automatically detecting abnormal behavior, tracking criminal activities, and managing traffic flow. By integrating image recognition algorithms, the system can identify crowd density, vehicle type, and even the behavior patterns of specific individuals, thereby providing timely response and intervention when needed.
[0004] Although existing technologies are capable of basic video and image analysis, they are deficient in dealing with complex lighting environments and image quality issues. For example, under uneven or extreme lighting conditions, the recognition accuracy and availability of images are usually greatly reduced, resulting in misjudgments and missed judgments, affecting the overall effect of urban security monitoring. In addition, the ability to capture details of image quality and diagnose abnormal conditions is limited, which can easily lead to the loss of key information in rapidly changing monitoring scenarios, reducing the speed and accuracy of response to emergencies, and affecting the overall quality of urban management and security. Summary of the invention
[0005] The purpose of the present invention is to solve the shortcomings in the prior art and to propose an urban monitoring image recognition system.
[0006] In order to achieve the above object, the present invention adopts the following technical solution: A city monitoring image recognition system comprises:
[0007] The illumination detection and feature analysis module collects image brightness data in the scene based on urban surveillance videos, analyzes the illumination distribution of the entire surveillance scene, uses histogram technology to locate areas with uneven brightness, calculates the illumination unevenness index based on the brightness distribution, and generates illumination feature data;
[0008] The target illumination adjustment module determines the brightness value of the key target area based on the illumination feature data, adjusts the brightness parameter of the target area according to the ambient illumination through a nonlinear mapping function, and changes the ratio of the bright area to the dark area of the overall image to obtain optimized target recognition image data;
[0009] The frequency analysis and anomaly detection module performs Fourier transform analysis on the frequency characteristics of the image based on the optimized target recognition image data, identifies the abnormal frequency in the image, locates the abnormal image quality changes in the image by comparing the frequency characteristics, including blurring and distortion of the image, and obtains the image anomaly detection result;
[0010] The abnormal area repair module uses the Kriging interpolation method to restore the damaged image area based on the image abnormality detection result, and simultaneously performs frequency smoothing and edge enhancement processing to optimize the overall image details to obtain an adjusted overall recognition image.
[0011] As a further solution of the present invention, the step of obtaining the area with uneven brightness is specifically as follows:
[0012] Extract the brightness information of each frame from the city surveillance video and integrate the data into an array of brightness values , using the formula: Calculate brightness value array Average brightness ;in, Indicates The brightness value of each pixel, Indicates the total number of pixels; according to the brightness value array Average brightness , using the formula: Calculate light level Frequency , get the brightness histogram ;in, is the brightness level, is an indicative function, if The value is equal to ,but ,if The value is not equal to , the result is .
[0013] As a further solution of the present invention, the step of acquiring the illumination feature data is specifically as follows:
[0014] According to the brightness histogram , using the formula:
[0015]
[0016] Calculate brightness histogram The average frequency ;
[0017] in, is the total number of brightness levels;
[0018] According to the brightness histogram The average frequency , using the formula:
[0019]
[0020] Calculate the uneven light index ;
[0021] According to the uneven illumination index , using the formula:
[0022]
[0023] Perform standardization operations to obtain lighting feature data ;
[0024] in, It is the maximum uneven illumination index.
[0025] As a further solution of the present invention, the step of obtaining the brightness parameter of the adjustment target area is specifically as follows:
[0026] According to the illumination characteristic data , using the formula:
[0027]
[0028] Determine the average brightness of key target areas ;
[0029] in, Represents lighting feature data Middle The brightness value of the data point, Indicates the total number of data points within the key target area;
[0030] According to the average brightness of the key target area , using the formula:
[0031]
[0032] Adjust the brightness parameters of the key target area through a nonlinear mapping function ;
[0033] in, Represents the maximum adjustment, adjusted through experiment or experience, Represents the brightness adjustment offset, which is adjusted according to the ambient lighting conditions and application scenarios. Represents the sensitivity of the adjustment, controls the steepness of the brightness adjustment curve, and is set by analyzing the changes in ambient light. is the base of natural logarithms.
[0034] As a further solution of the present invention, the step of acquiring the optimized target recognition image data is specifically as follows:
[0035] According to the brightness parameters of the key target area and lighting feature data , using the formula:
[0036]
[0037] Calculate the ratio adjustment factor of the bright and dark areas of the overall image ;
[0038] in, Dynamically adjust according to current lighting conditions and target recognition requirements;
[0039] The ratio of the bright area to the dark area of the overall image is adjusted by the factor , using the formula:
[0040]
[0041] Change the ratio of bright and dark areas of the overall image to obtain optimized target recognition image data ;
[0042] in, is a normalized value used to preserve the dark details of the image. Representation and Complementary ratio.
[0043] As a further solution of the present invention, the step of obtaining the abnormal frequency in the recognition image is specifically:
[0044] Image data is recognized according to the optimized target , using the formula:
[0045]
[0046] Perform Fourier transform to extract frequency features of the image ;
[0047] in, Represents the function Along the variables To perform integration, is a spatial domain variable, representing the position in the image data, is a frequency domain variable used to analyze the frequency content in the image, is the complex power of the base of the natural logarithm, used to convert spatial signals to frequency signals, represents the imaginary unit, Express The integral of is used to calculate the frequency distribution of the entire image data;
[0048] According to the frequency characteristics , using the formula:
[0049]
[0050] Calculate the amplitude of each frequency ;
[0051] in, Indicates frequency The amplitude in the frequency domain is the absolute value of the amplitude, which is used to determine whether the frequency is abnormal. It is the amplitude threshold used to distinguish normal from abnormal frequencies, set based on experience or statistical data.
[0052] As a further solution of the present invention, the steps of obtaining the image anomaly detection result are specifically as follows:
[0053] According to the amplitude of each frequency , using the formula:
[0054]
[0055] Add the frequency square term to get the weighted frequency contribution ;
[0056] According to the weighted frequency contribution , using the formula:
[0057]
[0058] Calculate the overall anomaly index of the image , and obtain the image anomaly detection results.
[0059] As a further solution of the present invention, the step of acquiring the adjusted overall recognition image is specifically as follows:
[0060] According to the overall abnormality index of the image , using the formula:
[0061]
[0062] The damaged area in the image is repaired by Kriging interpolation method to obtain the repaired image data function ;
[0063] in, is the restored image value, which is the output of the Kriging interpolation method. are weights, determined based on spatial autocorrelation, is the image anomaly detection result, at position The original image value of the damaged data point, It is the total number of neighboring points, which determines the range and accuracy of interpolation and is dynamically determined according to the distribution and density of the damaged area;
[0064] According to the restored image value , using the formula:
[0065]
[0066] Get frequency The smoothed image value at ;
[0067] in, is the frequency point being processed, is the cutoff frequency, which is adjusted according to the image requirements and quality indicators to control the cutoff limit of the filter;
[0068] According to the frequency The smoothed image value at , using the formula:
[0069]
[0070] Get in position The enhanced image value , get the adjusted overall recognition image;
[0071] in, is the interpolation process at position The image value of is smoothed at position The image value of is the enhancement factor, which is usually set based on visual effects and image quality requirements. are the row and column indices in the image data.
[0072] Compared with the prior art, the advantages and positive effects of the present invention are:
[0073] In the present invention, by collecting the image brightness data in the scene and analyzing the illumination distribution, the histogram technology is used to locate the area of uneven brightness and calculate the uneven illumination index, so that the scene analysis is more accurate. By adjusting the brightness parameters of the key target area and changing the ratio of the bright area to the dark area of the overall image, higher accuracy and response speed can be achieved in the process of recognizing the image. In addition, Fourier transform is performed to analyze the frequency characteristics of the image and identify abnormal frequencies in the image, which effectively identifies abnormal changes in blur and distortion in the image quality and enhances the real-time monitoring and early warning capabilities of the system. BRIEF DESCRIPTION OF THE DRAWINGS
[0074] Figure 1 is a system flow chart of the present invention;
[0075] Figure 2 A flow chart of the steps for locating an area with uneven brightness according to the present invention;
[0076] Figure 3 A flowchart of the steps for obtaining illumination characteristic data of the present invention;
[0077] Figure 4 A flow chart of the steps for obtaining the brightness parameters of the target area according to the present invention;
[0078] Figure 5 A flow chart of the steps for obtaining target recognition image data optimized for the present invention;
[0079] Figure 6 A flow chart of the steps of obtaining abnormal frequencies in an image according to the present invention;
[0080] Figure 7 A flowchart of the steps for obtaining the image anomaly detection result of the present invention;
[0081] Figure 8 The flowchart is a step flow of obtaining the adjusted overall recognition image of the present invention. DETAILED DESCRIPTION
[0082] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0083] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.
[0084] See also Figure 1 , an urban monitoring image recognition system comprises:
[0085] The illumination detection and feature analysis module collects image brightness data in the scene based on urban surveillance videos, analyzes the illumination distribution of the entire surveillance scene, uses histogram technology to locate areas with uneven brightness, calculates the illumination unevenness index based on the brightness distribution, and generates illumination feature data;
[0086] The target illumination adjustment module measures the brightness value of the key target area based on the illumination feature data, adjusts the brightness parameters of the target area according to the ambient illumination through a nonlinear mapping function, and changes the ratio of the bright area to the dark area of the overall image to obtain optimized target recognition image data;
[0087] The frequency analysis and anomaly detection module performs Fourier transform analysis on the frequency characteristics of the image based on the optimized target recognition image data, identifies the abnormal frequency in the image, and locates the abnormal image quality changes in the image by comparing the frequency characteristics, including image blur and distortion, to obtain the image anomaly detection results;
[0088] The abnormal area repair module uses the Kriging interpolation method to restore the damaged image area based on the image abnormality detection results, and simultaneously performs frequency smoothing and edge enhancement processing to optimize the overall image details to obtain the adjusted overall recognition image;
[0089] The lighting feature data include regional brightness level, position coordinates of main bright areas and secondary dark areas, and lighting values of key visual channels. The optimized target recognition image data include the adjusted brightness level of the core observation area, the contrast ratio of the overall image, and the brightness adjustment value of key details. The image anomaly detection results include the identified frequency anomaly range, the frequency that affects the image quality, and the abnormal image quality area that needs attention. The adjusted overall recognition image includes the repaired image blocks, the area with improved smoothness of image details, and the range of enhanced edge clarity.
[0090] See also Figure 2, the specific steps for locating the area with uneven brightness are:
[0091] Extract the brightness information of each frame from the city surveillance video and integrate the data into an array of brightness values , using the formula:
[0092]
[0093] Calculate brightness value array Average brightness ;
[0094] in, Indicates The brightness value of each pixel, Indicates the total number of pixels;
[0095] According to the brightness value array Average brightness , using the formula:
[0096]
[0097] Calculate light level Frequency , get the brightness histogram ;
[0098] in, is the brightness level, which is also a bucket or interval in the histogram. Is an indicative function used to calculate the brightness value equal If The value is equal to ,but ,if The value is not equal to , the result is ;
[0099] The calculation process is as follows:
[0100] For the formula:
[0101]
[0102] Setting parameters is 100;
[0103] The brightness value of 100 pixels in the surveillance video is , , , directly measured and recorded by the sensors of the monitoring system.
[0104] Calculate the sum of the brightness values of all pixels:
[0105]
[0106] Assume the sum is 15,000 (the actual value depends on all the specific brightness data).
[0107] Use the above summation to calculate the average brightness:
[0108]
[0109] Average brightness value Represents the average brightness level of the monitored scene at the considered time point.
[0110] For the formula:
[0111]
[0112] If the brightness level of interest is from 140 to 160;
[0113] Calculate the frequency of brightness value 150:
[0114] Calculate the number of pixels with a brightness value equal to 150:
[0115]
[0116] There are 20 pixels with a brightness of 150;
[0117] Calculate the frequency of brightness value 150:
[0118]
[0119] Brightness Histogram It means that among all detected pixels, 20% of the pixels have a brightness value of 150.
[0120] See also Figure 3 , the specific steps for obtaining illumination feature data are as follows:
[0121] According to the brightness histogram , using the formula:
[0122]
[0123] Calculate brightness histogram The average frequency ;
[0124] in, is the total number of brightness levels, indicating the kind of brightness gradation considered in the image;
[0125] According to the brightness histogram The average frequency , using the formula:
[0126]
[0127] Calculate the uneven light index ;
[0128] Calculate the uneven light index ;
[0129] According to the uneven light index , using the formula:
[0130]
[0131] Perform standardization operations to obtain lighting feature data ;
[0132] in, is the maximum illumination unevenness index, used for normalization , depending on the most extreme light distribution in theory;
[0133] The calculation process is as follows:
[0134] Set a simplified brightness histogram , the histogram records the pixel frequency of different brightness levels in a surveillance image. The following is the histogram data:
[0135] ;
[0136] ;
[0137] ;
[0138] ;
[0139] ;
[0140] Total brightness levels;
[0141] Calculate the average frequency :
[0142]
[0143] Calculate the uneven light index :
[0144] Total pixels (i.e. the total number of pixels in the image).
[0145]
[0146] Set the maximum illumination unevenness index is the theoretical value in the case of an extremely uneven distribution, where each pixel is at its maximum possible deviation (for example, when When , all pixels are either 0 or 400, that is, a bimodal extreme distribution).
[0147] Calculate the maximum uneven illumination index :
[0148]
[0149] Set the maximum deviation to 400 (brightness from 0 to a maximum of 400):
[0150]
[0151] Standardized lighting characteristics data:
[0152]
[0153] result This indicates that the illumination distribution of this image is relatively uniform relative to the theoretical maximum unevenness, and can be used to evaluate the improvement effect of image processing algorithms on uneven illumination or to compare the illumination conditions of different images.
[0154] See also Figure 4 , the specific steps for obtaining the brightness parameters of the target area are:
[0155] According to the light characteristic data , using the formula:
[0156]
[0157] Determine the average brightness of key target areas ;
[0158] in, Represents lighting feature data Middle The brightness value of the data point is used to calculate the light level of the area. Indicates the total number of data points within the key target area;
[0159] Based on the average brightness of the key target area , using the formula:
[0160]
[0161] Adjust the brightness parameters of the key target area through a nonlinear mapping function ;
[0162] in, Represents the maximum adjustment range, which is adjusted through experiments or experience to meet the lighting needs of different monitoring scenarios. Represents the brightness adjustment offset, which is usually adjusted according to ambient lighting conditions and specific application scenarios. Represents the sensitivity of the adjustment, controls the steepness of the brightness adjustment curve, and is set by analyzing changes in ambient light. is the base of the natural logarithm, used to construct the exponential function of nonlinear mapping;
[0163] The calculation process is as follows:
[0164] For the formula:
[0165]
[0166] The key target area contains 10 data points with specific brightness values. : 200, 210, 220, 230, 240, 250, 260, 270, 280, 290;
[0167] calculate Values:
[0168]
[0169] Represents the average brightness level of the key target area at the considered time point.
[0170] For the formula:
[0171]
[0172] Setting parameters is 1.5, is 240, is 10;
[0173]
[0174]
[0175]
[0176] Value Represents the adjusted brightness parameter. This value is used in actual image processing to adjust the brightness of key areas to adapt to different ambient lighting conditions.
[0177] See also Figure 5 , the steps for obtaining optimized target recognition image data are as follows:
[0178] According to the brightness parameters of key target areas and lighting feature data , using the formula:
[0179]
[0180] Calculate the ratio adjustment factor of the bright and dark areas of the overall image ;
[0181] in, Dynamically adjust according to current lighting conditions and target recognition requirements;
[0182] Adjust the factor based on the ratio of bright and dark areas of the overall image , using the formula:
[0183]
[0184] Change the ratio of bright and dark areas of the overall image to obtain optimized target recognition image data ;
[0185] in, It is a standardized value used to preserve the dark details of the image. It is usually set as a ratio or fixed value of the overall brightness of the image. Representation and The proportion of complementation is used to calculate the proportion of dark area retention;
[0186] The calculation process is as follows:
[0187] For the formula:
[0188]
[0189] Set up lighting characteristics data 100, 150, 200, 250, 300 (5 data points in total), the adjusted brightness parameter is 200;
[0190]
[0191]
[0192] For the formula:
[0193]
[0194] Setting parameters , is 250.
[0195]
[0196] Analysis results Calculation results Indicates the brightness of the image data after adjustment. This means that the bright areas of the image are not adjusted because , saying that the original ratio of bright area to dark area is already ideal, or the adjustment factor is completely biased towards the bright area.
[0197] See also Figure 6 , the specific steps for obtaining the abnormal frequency in the recognition image are:
[0198] Recognize image data according to optimized target , using the formula:
[0199]
[0200] Perform Fourier transform to extract frequency features of the image ;
[0201] in, Represents the function Along the variables To perform integration, is a spatial domain variable, representing the position in the image data, is a frequency domain variable used to analyze the frequency content in the image, It is the complex power of the base of the natural logarithm, representing the core calculation part of the Fourier transform, which is used to convert spatial signals into frequency signals. Represents the imaginary unit, used for complex number operations, Express The integral of is used to calculate the frequency distribution of the entire image data;
[0202] According to the frequency characteristics , using the formula:
[0203]
[0204] Calculate the amplitude of each frequency ;
[0205] in, Indicates frequency The amplitude in the frequency domain is the absolute value of the amplitude, which is used to determine whether the frequency is abnormal. It is the amplitude threshold used to distinguish normal from abnormal frequencies, usually set based on experience or statistical data;
[0206] The calculation process is as follows:
[0207] For the formula:
[0208]
[0209] like , set the parameters Hz, corresponding image data ;
[0210] limit The value range is arrive The integral calculation is performed in seconds, taking into account is the exponential form of a complex number. , the expression can be written as , calculate the following integral:
[0211]
[0212] Calculate the integral:
[0213] Will Replace with its equivalent expression , substituting this expression into the integral:
[0214]
[0215] The calculation of the integral is divided into two parts:
[0216]
[0217] simplify:
[0218]
[0219] The first integral is a constant The integral result is , the second term involves the integration of complex exponentials, and the result is .
[0220]
[0221] Therefore, the result of Fourier transform shows that the complex form of the component with a frequency of 10 Hz is approximately , the amplitude is .
[0222] For the formula:
[0223]
[0224] Known ;
[0225] For plural , its actual part is (no real term), the imaginary part is ;
[0226] For plural The modulus (absolute value) is calculated by the following formula:
[0227]
[0228] Substituting the values into the calculation will and Substituting into the above formula, we get:
[0229]
[0230] It means that when the frequency is 10Hz, the amplitude of the signal is ;
[0231] Setting Thresholds for , this value is obtained based on historical data analysis and usually reflects the maximum amplitude under normal circumstances. Greater than threshold , so the frequency is marked as abnormal.
[0232] See also Figure 7 , the specific steps for obtaining image anomaly detection results are:
[0233] According to the amplitude of each frequency , using the formula:
[0234]
[0235] Add the frequency square term to get the weighted frequency contribution ;
[0236] in, Used to locate quality changes in an image by dividing the amplitude of the frequency The square of the frequency Multiplying together, it highlights the impact of high-frequency components on image quality, especially those that may cause image blur or distortion;
[0237] According to the weighted frequency contribution , using the formula:
[0238]
[0239] Calculate the overall anomaly index of the image , get the image anomaly detection result;
[0240] The calculation process is as follows:
[0241] Set the parameters to an amplitude of 20 Hz , at an amplitude of 50Hz , at an amplitude of 100Hz ;
[0242] Calculate weighted frequency contribution :
[0243]
[0244]
[0245] Here, the weighted values are intended to emphasize those frequencies that are higher in frequency and larger in amplitude, as these are often more associated with image quality issues such as blurring and distortion.
[0246] Calculate the cumulative weighted frequency contribution according to the formula:
[0247]
[0248] Includes all significant frequency contributions:
[0249]
[0250] Set a baseline value , this benchmark may be determined through historical data analysis and experience to represent the general level of frequency anomalies in the image. Significantly higher than this baseline value indicates that there are more serious quality problems in the image than usual, especially in the contribution of high-frequency components, indicating that the image is excessively blurred or distorted.
[0251] See also Figure 8 , the steps for obtaining the adjusted overall recognition image are as follows:
[0252] According to the overall abnormality index of the image , using the formula:
[0253]
[0254] The damaged area in the image is repaired by Kriging interpolation method to obtain the repaired image data function ;
[0255] in, is the restored image value, which is the output of the Kriging interpolation method. is a weight, determined based on spatial autocorrelation, that affects the contribution of each neighboring point in the interpolation. is the image anomaly detection result, at position The original image value of the damaged data point, It is the total number of neighboring points, which determines the range and accuracy of interpolation and is dynamically determined according to the distribution and density of the damaged area;
[0256] According to the restored image value , using the formula:
[0257]
[0258] Get frequency The smoothed image value at ;
[0259] in, is the frequency point being processed, is the cutoff frequency, which is adjusted according to the image requirements and quality indicators to control the cutoff limit of the filter;
[0260] According to frequency The smoothed image value at , using the formula:
[0261]
[0262] Get in position The enhanced image value , get the adjusted overall recognition image;
[0263] in, is the interpolation process at position The image value is used as the basis for the enhancement calculation. is smoothed at position The image value provides a smooth background to highlight the edges. is the enhancement factor, which adjusts the strength of edge enhancement and is usually set based on visual effects and image quality requirements. It is the row index and column index in the image data, which is used to locate the specific pixel points, and the enhancement processing is performed on these positions;
[0264] The calculation process is as follows:
[0265] For the formula:
[0266]
[0267] Set the damaged point location and its outlier value As follows, it is taken from the actual monitoring image anomaly detection results:
[0268] exist Value ;
[0269] exist Value ;
[0270] exist Value ;
[0271] Set each weight for , determined based on the inverse square of the distance or statistical methods.
[0272] calculate exist The value at:
[0273]
[0274] For the formula:
[0275]
[0276] Select frequency The cut-off frequency is 60Hz. Set to 50Hz (based on analysis of the image data, this cutoff frequency was determined to optimize the filtering effect):
[0277]
[0278] It means that the image value at the frequency of 60 Hz is reduced to about 41% of the original value after processing.
[0279] For the formula:
[0280]
[0281] from arrive The transformation of is used in image processing and data interpolation. Position is usually represented abstractly, but in practice, especially when two-dimensional images are involved, two coordinates are required. (line) and (column) to specifically describe this position, accurately specifying and accessing each pixel in the image matrix.
[0282] For example, if in The position needs to be calculated or repaired. In fact, in the specific image matrix, this point is specified as (line 25) and (column 25).
[0283] Setting parameters is 130, is 0.41, (Set according to image processing requirements):
[0284]
[0285] The result means that The image value at the position is significantly increased after enhancement, which improves the local details and contrast of the image.
[0286] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.
Claims
1. A city monitoring image recognition system, characterized in that: The system comprises: The illumination detection and feature analysis module collects image brightness data in the scene based on urban surveillance videos, analyzes the illumination distribution of the entire surveillance scene, uses histogram technology to locate areas with uneven brightness, calculates the illumination unevenness index based on the brightness distribution, and generates illumination feature data; The acquisition step of locating the area with uneven brightness is specifically as follows: extracting the brightness information of each frame from the urban surveillance video, and integrating the data into a brightness value array , using the formula: Calculate brightness value array Average brightness ;in, Indicates The brightness value of each pixel, Indicates the total number of pixels; according to the brightness value array Average brightness , using the formula: Calculate light level Frequency , get the brightness histogram ;in, is the brightness level, is an indicative function, if The value is equal to ,but ,if The value is not equal to , the result is ; The step of acquiring the illumination feature data is specifically as follows: , using the formula: Calculate brightness histogram The average frequency ;in, is the total number of brightness levels; according to the brightness histogram The average frequency , using the formula: Calculate the uneven light index According to the uneven illumination index , using the formula: Perform standardization operations to obtain lighting feature data ;in, is the maximum uneven illumination index; The target illumination adjustment module determines the brightness value of the key target area based on the illumination feature data, adjusts the brightness parameter of the target area according to the ambient illumination through a nonlinear mapping function, and changes the ratio of the bright area to the dark area of the overall image to obtain optimized target recognition image data; The frequency analysis and anomaly detection module performs Fourier transform analysis on the frequency characteristics of the image based on the optimized target recognition image data, identifies the abnormal frequency in the image, locates the abnormal image quality changes in the image by comparing the frequency characteristics, including blurring and distortion of the image, and obtains the image anomaly detection result; The abnormal area repair module uses the Kriging interpolation method to restore the damaged image area based on the image abnormality detection result, and simultaneously performs frequency smoothing and edge enhancement processing to optimize the overall image details to obtain an adjusted overall recognition image.
2. The urban monitoring image recognition system according to claim 1, characterized in that: The steps of obtaining the brightness parameter of the target area are specifically as follows: According to the illumination characteristic data , using the formula: Determine the average brightness of key target areas ;in, Represents illumination feature data Middle The brightness value of the data point, Represents the total number of data points within the key target area; based on the average brightness of the key target area , using the formula: Adjust the brightness parameters of the key target area through a nonlinear mapping function ;in, Represents the maximum adjustment range, adjusted through experiments, Represents the brightness adjustment offset, which is adjusted according to the ambient lighting conditions and application scenarios. Represents the sensitivity of the adjustment, controls the steepness of the brightness adjustment curve, and is set by analyzing the changes in ambient light. is the base of natural logarithms.
3. The urban monitoring image recognition system according to claim 2, characterized in that: The steps for obtaining the optimized target recognition image data are specifically as follows: According to the brightness parameters of the key target area and lighting feature data , using the formula: Calculate the ratio adjustment factor of the bright and dark areas of the overall image ;in, Dynamically adjust according to current lighting conditions and target recognition requirements; adjust the factor according to the ratio of bright and dark areas of the overall image , using the formula: Change the ratio of bright and dark areas of the overall image to obtain optimized target recognition image data ;in, is a normalized value used to preserve the dark details of the image. Representation and Complementary ratio.
4. The urban monitoring image recognition system according to claim 3 is characterized in that: The step of obtaining the abnormal frequency in the recognition image is specifically as follows: Image data is recognized according to the optimized target , using the formula: Perform Fourier transform to extract frequency features of the image ;in, Represents the function Along the variables To perform integration, is a spatial domain variable, representing the position in the image data, is a frequency domain variable used to analyze the frequency content in the image, is the base of the natural logarithm, used to convert spatial signals to frequency signals, represents the imaginary unit, Express The integral of is used to calculate the frequency distribution of the entire image data; according to the frequency characteristics , using the formula: Calculate the amplitude of each frequency ;in, Indicates frequency The amplitude in the frequency domain is the absolute value of the amplitude, which is used to determine whether the frequency is abnormal. It is the amplitude threshold used to distinguish normal and abnormal frequencies, set based on statistical data.
5. The urban monitoring image recognition system according to claim 4, characterized in that: The steps for obtaining the image anomaly detection result are specifically as follows: According to the amplitude of each frequency , using the formula: Add the frequency square term to get the weighted frequency contribution ; According to the weighted frequency contribution , using the formula: Calculate the overall anomaly index of the image , and obtain the image anomaly detection results.
6. The urban monitoring image recognition system according to claim 5, characterized in that: The steps for obtaining the adjusted overall recognition image are specifically as follows: According to the overall abnormality index of the image , using the formula: The damaged area in the image is repaired by Kriging interpolation method to obtain the repaired image data function ;in, is the restored image value, which is the output of the Kriging interpolation method. are weights, determined based on spatial autocorrelation, is the image anomaly detection result, at position The original image value of the damaged data point, is the total number of neighboring points, which determines the range and accuracy of interpolation and is dynamically determined according to the distribution and density of the damaged area; according to the image value after repair , using the formula: Get frequency The smoothed image value at ;in, is the frequency point being processed, is the cutoff frequency, which is adjusted according to the image requirements and quality indicators to control the cutoff limit of the filter; according to the frequency The smoothed image value at , using the formula: Get in position The enhanced image value , and obtain the adjusted overall recognition image; where, is the interpolation process at position The image value of is smoothed at position The image value of is the enhancement factor, which is usually set based on visual effects and image quality requirements. are the row and column indices in the image data.
