TW201524188A - Detecting method of abnormality of image capturing by camera - Google Patents

Abstract

The present invention discloses a detecting method of an abnormality of an image capturing by a camera. Two absolute backgrounds are set and quickly judging if a frame is shaking or not.If the frame is not shaking, the background subtraction is used to check if two absolute backgrounds obtained from different time have large difference or not. If there is a large difference, both grayscale histogram distribution and information of edges are utilized to judge if the abnormal image belongs to the frame shift, covering or paint spraying. If the difference is not large, both a luminance and an edge detail are employed for judging if there is fogging. Discrete cosine transform is performed in order to detect if there is defocus problem. Color cast is detected by analyzing histogram distribution of color channels (RGB). Change of the luminance is used to check if the frame is interfered. When the frame dose not has any abnormal condition, the absolute background will be updated. Otherwise, an alarm is given if any abnormal condition mentioned above is detected.

Description

Camera screen abnormality detection method

The invention relates to a camera screen abnormality detecting method, in particular to a camera screen for detecting whether a camera photographing screen is blocked, painted, shaken, lens shifted, out-focus, color shift, dense fog, screen flicker and the like. The detection method of abnormality.

In recent years, the public has gradually attached importance to its own safety, and the demand for electronic safety equipment has increased year by year. In addition, in order to maintain law and order and deter crime, it is a common way to install surveillance cameras on city corners, roads or buildings, which in turn will lead to the rapid development of smart video security software. With the evolution of technology, the related security surveillance industry has also evolved from the early stage of the closed-circuit television system to the Digital Video Recorder (DVR) and IP Network Video Recorder (NVR).

However, although the monitoring system is moving toward digitalization, in terms of security monitoring and management, it still has to rely on the management of personnel, and often the security management personnel must monitor multiple monitoring screens at the same time. In addition, surveillance systems often face problems such as aging, abnormality, or deliberate destruction. Deliberate damage, including shadowing, painting, out-of-focus, color shift, or image shift, can result in the inability to record normal images and produce blind spots. When the incident needs to be clarified by the surveillance screen, it is likely to cause great losses and regrets.

For the above various abnormal conditions, the main method principle of the prior art is to reconstruct an absolute background, and judge whether the picture is subject to steering, displacement, or out of focus by the absolute background information. The camera is turned or shifted by the absolute background update delay method to judge the image is turned and displaced, and the number of pixel differences in the whole picture is counted. After a slow update, the absolute background at n sheets and the absolute before k sheets are absolute. The background judges the difference, and when the difference is greater than the threshold, it is judged that the picture is turned. When the camera is covered, the difference between the histogram of the absolute background and the histogram of the current image is judged by the difference between the histogram of the absolute image and the current image. The condition 1 is that the image is biased when the lens is blocked. Dark, the number of times the histogram is distributed in some ranges will become higher. When the number of image histograms is the highest and the sum of adjacent gray levels is greater than the threshold, this condition is satisfied; condition 2 is when the camera has not been obscured. The histogram of the absolute background image and the histogram of the subtraction of the current image histogram, the sum of the dark areas should be less than the sum of the cameras when they are obscured. When the sum of the dark areas i~32 is greater than the sum of the first i~k gray levels, That is, the condition is satisfied; when both conditions are satisfied, it is judged that the camera is obscured.

However, when the camera is obscured by a translucent object such as white paper, the screen does not necessarily belong to the darkness, which will cause the condition 2 to be unsuccessful, and thus the screen cannot be detected to be shielded.

Then, the frame out-of-focus detection uses Fourier Transform (Fourier Transform) to analyze the difference between the current image high-frequency information and the absolute background high-frequency information. When the current image high-frequency information is smaller than the absolute background high-frequency information, the judgment is made. Out of focus for the picture.

However, the shortcoming of this algorithm is that if multiple anomaly detections are performed at the same time, the time complexity may be too high, and instant detection cannot be achieved.

In view of the above-mentioned problems of the prior art, it is an object of the present invention to provide a method for detecting an abnormality of a camera screen to solve the problems faced by conventional cameras for detecting abnormal images of a camera.

According to an object of the present invention, a method for detecting an abnormality of a camera screen is provided, which comprises the steps of: capturing a plurality of consecutive pictures. Whether the plurality of pictures are stable is determined by the adjacent subtraction method. According to the plurality of pictures, the two absolute backgrounds are established, and the two absolute backgrounds are separated by time, and are updated every preset time. The background subtraction method is used to determine the difference between each updated absolute background and its corresponding absolute background before the update. When at least one absolute background before updating and the corresponding updated absolute background difference are greater than the first threshold, determining whether the camera has image shift or based on the grayscale histogram distribution and edge pixels of the plurality of pictures and the two absolute backgrounds The abnormal state of being obscured, and the warning message is sent corresponding to the abnormal state and the troubleshooting step of executing the corresponding abnormal state is performed. When the at least one absolute background before the update and the corresponding updated absolute background difference are less than the first threshold, the method includes the following steps: determining whether the camera screen is based on the brightness (ie, grayscale value) information and the edge information of the plurality of pictures It has an abnormal state of dense fog. According to the high frequency information loss amount of the plurality of pictures, it is determined whether the camera picture has an abnormal state of out of focus. According to the average brightness value of each of the RGB channels of the plurality of pictures and the local area average brightness value of each channel of the RGB, it is determined whether the camera picture has an abnormal state of color shift. According to the average brightness value of the plurality of pictures, it is determined whether the camera screen has an abnormal state of flicker. When the camera screen does not have an abnormal state of dense fog, out-of-focus, color shift, flicker or a combination thereof, each absolute background is updated every preset time, and when the camera screen has an abnormal state, a warning message is issued.

Preferably, determining whether the camera is stable or more may include the step of obtaining a difference image by one of the pictures and another adjacent picture. When the number of pixels with a large difference in the difference image is lower than the total number of pixels of the preset percentage, it is determined that the camera picture is in a stable state, and a two absolute background is established. When the number of pixels with a large difference in the difference image is higher than the total number of pixels of the preset percentage, it is determined that the camera screen is in a shaking state and a warning message is issued.

Preferably, the detecting method of the abnormality of the camera screen may further comprise the step of: determining whether the total number of pixels of the grayscale histogram of the plurality of pictures in a specific section is greater than the total number of pixels of the grayscale histogram of each absolute background. When the total number of pixels of the grayscale histogram of the plurality of pictures is greater than or equal to the total number of pixels of the grayscale histogram of the two absolute backgrounds, it is determined whether the total number of edge pixels of the plurality of pictures is less than or equal to the preset of each absolute background. The total number of edge pixels in the scale.

Preferably, the detection method of the abnormality of the camera picture may further include the following steps: when the absolute background difference is large, and the total number of pixels of the grayscale histogram of the plurality of pictures is smaller than the total number of pixels of the gray level histogram of each absolute background. , it is judged that the camera is in an abnormal state of displacement, and the warning message is issued. When the absolute background difference is large, and the total number of pixels of the grayscale histogram of a plurality of pictures in a specific section is not less than the total number of pixels of the grayscale histogram of each absolute background, and the total number of edge pixels of the plurality of pictures is greater than the preset of each absolute background. When the total number of edge pixels is proportional, the camera is judged to be in an abnormal state of displacement, and a warning message is issued. When the absolute background difference is large, and the total number of pixels of the grayscale histogram of a plurality of pictures in a specific section is not less than the total number of pixels of the grayscale histogram of each absolute background, the total number of edge pixels of the plurality of pictures is not greater than the absolute background. When the total number of edge pixels is set, it is judged that the camera is in an abnormal state of being shielded and a warning message is issued.

Preferably, the detecting method of the abnormality of the camera screen further comprises the following steps: determining whether the difference between the highest peak-to-peak value of the left side of the average brightness value of the plurality of pictures and the highest peak-to-peak value of the right side of the average brightness value is less than a preset gap threshold. It is determined whether the number of 250 to 255 grayscale values and 255 grayscale values of the edge image grayscale histogram of the plurality of pictures is less than a preset number threshold. Obtain the grayscale average value of the edge image by Sobel edge detection, and judge whether there is fog and fog concentration according to the grayscale average value of the edge image of the plurality of images and the grayscale image standard deviation of the original image, and issue a warning containing the fog concentration. message.

Preferably, the detecting method of the abnormality of the camera screen may further include the following steps: when the difference between the highest peak value and the highest peak value of the right side of the average brightness value of the gray level histogram of the plurality of pictures is less than the preset gap threshold, Then, it is initially determined that the camera screen is in an abnormal state of fogging.

Preferably, the detecting method of the camera screen abnormality further comprises the following steps: when the number of the 250 to 255 grayscale value and the 255 grayscale value of the edge image histogram are less than the preset number threshold, it is determined that the camera screen is in the starting position. The abnormal state of the fog.

Preferably, the detecting method of the camera screen abnormality further comprises the steps of: setting a high frequency information amount reference value, and setting a preset loss amount by using the high frequency information amount reference value. Determining whether the high-frequency information loss of the plurality of pictures is greater than the preset loss amount, and when it is greater than the preset loss amount, determining that the camera picture is in an abnormal state of out-of-focus, and issuing a warning message; when not greater than the preset loss amount, Then, the high frequency information reference value is updated, and the comparison of the high frequency information loss amount and the preset loss amount of the next plurality of pictures is performed.

Preferably, the method for detecting abnormality of the camera screen further comprises the steps of: determining whether a difference between a maximum average brightness of each color channel (RGB) of the plurality of pictures and a minimum average brightness of each color channel is greater than a preset average brightness difference. . The difference between the average value of the bright portion of the color channel that determines the maximum color shift and the average of the bright portion ranges of the other color channels is greater than a predetermined average difference. The difference between the average of the dark portion of the color channel that determines the maximum color shift and the average of the dark portion of the other color channels is greater than the preset average difference. When the difference between the maximum average brightness and the minimum average brightness of each color channel is greater than the average brightness difference of the preset channel, and the average of the bright portion range of the color channel with the largest color shift and the average of the bright portion of the other color channels If the difference between the value and the average of the dark portion of the color channel with the largest color shift and the average of the dark portion of the other color channels are greater than the preset average difference, the camera screen is judged to be in an abnormal state of the color shift. And issue a warning message.

Preferably, the detecting method of the abnormality of the camera screen further comprises the following steps: determining the difference value of the adjacent frames of the plurality of pictures. When the difference value of the adjacent frames is greater than the second threshold, it is determined that the image state changes. When the number of image state changes is greater than the preset accumulated number, it is determined that the camera screen is in an abnormal state of image flicker, and a warning message is issued.

As described above, the detection method of the camera screen abnormality of the present invention can automatically and automatically detect most of the abnormal images in real time, so as to facilitate subsequent monitoring and maintenance processing, and can ensure the quality of the monitoring screen so as to avoid accidents. It is impossible to clarify the regrets of the event by means of the monitoring screen; in addition, the present invention can also save a lot of manpower management costs, and can properly handle the abnormal screen when the first time is generated.

Steps S101 to S111, S201 to S204, S301 to S307‧‧

Fig. 1 is a first flow chart showing a method for detecting an abnormality of a camera screen of the present invention.
Fig. 2 is a second flow chart showing a method for detecting an abnormality of a camera screen of the present invention.
Fig. 3 is a third flow chart showing a method for detecting an abnormality of a camera screen of the present invention.

The technical features, contents, and advantages of the present invention and the efficacies thereof can be understood by the present inventors. The present invention will be described in conjunction with the drawings and will be described in detail with reference to the embodiments. The subject matter is only for the purpose of illustration and description. It is not intended to be a true proportion and precise configuration after the implementation of the present invention. Therefore, the scope and configuration relationship of the attached drawings should not be interpreted or limited. First described.

Please refer to FIG. 1 to FIG. 3; FIG. 1 is a first flowchart of a method for detecting an abnormality of a camera screen according to the present invention; and FIG. 2 is a second flowchart of a method for detecting an abnormality of a camera screen according to the present invention; The figure is a third flow chart of the method for detecting an abnormality of a camera screen of the present invention. As shown in the figure, the present invention provides a method for detecting an abnormality of a camera screen, which comprises the following steps:

A plurality of consecutive pictures are captured (step S101). Whether or not the plurality of pictures are stable is determined by the adjacent subtraction method (step S102). The two absolute backgrounds are established according to the plurality of pictures, and the two absolute backgrounds have a time difference (step S103), and are updated every preset time. The difference between each updated absolute background and its corresponding absolute background before the update is judged by the background subtraction method (step S104). When at least one absolute background before updating and the corresponding updated absolute background difference are greater than the first threshold, determining whether the camera has image shift or based on the grayscale histogram distribution and edge pixels of the plurality of pictures and the two absolute backgrounds The abnormal state of being shielded (step S105), and a warning message is issued corresponding to the abnormal state and a troubleshooting step of executing the corresponding abnormal state (step S106). When the at least one absolute background before the update and the corresponding updated absolute background difference are less than the first threshold, the method includes the following steps: determining whether the camera screen has an abnormal state of the fog according to the brightness information and the edge information of the plurality of pictures (step S107). Based on the high frequency information loss amount of the plurality of pictures, it is judged whether or not the camera screen has an abnormal state of out of focus (step S107). The camera screen determines whether the camera screen has an abnormal state of color shift based on the average luminance value of each of the RGB channels of the plurality of pictures and the local area average luminance value of each of the RGB channels (step S107). Based on the average brightness value of the plurality of pictures, it is determined whether or not the camera screen has an abnormal state of flicker (step S107). When the camera screen does not have an abnormal state of dense fog, out-of-focus, color shift, flicker or a combination thereof, each absolute background is updated every preset time (step S108), and when the camera screen has an abnormal state, a warning is issued. Message (step S109).

In other words, before performing various anomaly detections, you should confirm that the camera has been erected steadily, because the established absolute backgrounds and stable images ensure the correct rate of subsequent anomaly detection to avoid too many false positives. The invention uses the background subtraction method to filter the moving object, and uses the time difference to quickly determine whether the pixel is a background pixel. As shown in the formula (1), if there is a moving object on the screen, after the n frames, the original position should be left. The position, if there is no change in the pixel value after passing n frames, is determined as the background.

(1)

Where t is a time variable; F _t+n (x, y), F _t (x, y) are respectively represented as pixel values at (x, y) at time t+n and t; B _t (x, y ) is expressed as the pixel value of the absolute background (x, y) when time is t. In this way, using the time difference picture, most of the absolute background can be quickly constructed, and too many operations can be avoided. After establishing a basic absolute background, the subsequent background information can be continuously updated by updating the absolute background.

The main purpose of the background update is to create a reliable background information from the input screen so that the moving object detection can be executed correctly. However, as time changes or the moving target enters the screen, the background image will inevitably produce some changes. If the initial background is used to detect the anomaly, there may be a misjudgment. In order to reduce this error situation, the update of the background needs to be established. The method of this case proposes an idea and concept of background update. When the pixel value in the picture has not changed any time after a certain period of time, the current pixel value is updated to the background according to the proportional principle, and the judgment rule is as in formula (2). And formula (3):

(2)

(3)

Where Count _{t +1} (x,y) is the counter, k is the image difference threshold, B _th is the updated threshold, B _{t +1} (x,y) is the new background after the background update, B _t (x,y For the current reference background, F _t (x, y) is the current image, and the α parameter is to adjust the background update rate. The value can be between 0 and 1. This value is used to control the rate of background update. The larger the value, the closer the background image is to the current image; the smaller the value, the closer the image after the background update is to the background image that was originally created, and then the latest background information can be obtained.

For the judgment of the absolute background difference, before performing various anomaly detections, it is first determined whether the input image and the constructed absolute background are greatly different. When the original shooting scene is different from the original setting, It is very likely that the picture is obscured, painted or the lens is turned, shifted, etc.; however, if only the input image is compared with the absolute background, the short-term blockage, such as the short-term difference caused by the pedestrian walking in front of the lens, will Warning messages may be posted continuously.

Therefore, when establishing an absolute background, two absolute backgrounds with time difference will be constructed at the same time. The two absolute backgrounds will have different update time due to time difference, and the image that stays for a short time will not be updated to the background, which can avoid short-lived The difference caused misjudgment. Therefore, under normal circumstances, the difference between the two absolute backgrounds will not be too large. When the two absolute backgrounds cause great differences, it means that the absolute background of the earlier update is updated to a corrupted or abnormal picture, resulting in a situation that has not been updated before. Absolute background has a great difference. At this point, it can be judged that the camera may be damaged or abnormal. When the absolute background judgment does not find any abnormality, it does not mean that the picture is normal, and some abnormal conditions will not be The original preset screens are different, so it is still necessary to detect whether the screen is normal.

The detection of whether the picture is stable or swaying will be explained in this paragraph. When the fixed camera is shaken, it is also one of the abnormal states. However, there are many factors in the shaking. In addition to the camera being deliberately pushed and shaken by natural phenomena such as earthquakes, real life, such as large vehicles passing through, causes road vibration. Or strong winds, etc., may also cause the picture to shake, but they are all within the tolerable range of shaking. Therefore, it is necessary to further control the adjustment detection sway sensitivity when detecting sloshing.

In the image stabilization detection, the present invention uses a temporal difference method, and then determines whether it is stable according to the difference ratio. A stable video has a larger number of pixels with a smaller difference in the difference image of the adjacent picture. Conversely, a picture with a large sway is a larger number of pixels in the difference image of the adjacent picture, so according to the number of pixels Whether the difference ratio determination screen is stable, as shown in formula (4), if SP _{FD is} larger than S _th ×S_image, it means that the picture is shaking, and if SP _{FD is} not larger than S _th ×S_image, the picture is stable; SP _FD is a large difference image difference number of pixels, the ratio S _TH as shaking, N_image the total number of pixels in the image, it is possible by setting S S _TH to adjust the sensitivity setting (in this case a method of detecting shake _th = 0.5 ~ 0.7 And finally, if the video screen is stable, the screen is directly output; otherwise, if it is detected that the screen is unstable, a warning message is issued (step S110).

SP _FD > S _th ×N_image(4)

When the absolute background difference is large, it is preliminarily judged that the camera may be damaged or abnormal. The abnormality or damage that occurs when it is different from the absolute background is most likely that the picture is subject to steering or shifting. Both make the camera unable to monitor effectively. Furthermore, the method of the present invention sets two conditions to determine whether the lens is subject to steering or shifting or that the picture is obscured or painted.

First, gray scale histogram analysis is performed on a plurality of pictures and absolute backgrounds (step S201). When the shots have not been masked, the gray scale histograms of the plurality of pictures will exhibit a normal distribution, and the absolute background gray level histograms constructed. The difference is not large, so the histogram distribution can be used to judge whether the picture is normal. When the camera is covered or painted, the picture may be covered by the cover (such as palm or paint), and the gray scale of the picture is straight. The graph will have a tendency to concentrate on the distribution.

Determining whether the gray-scale histogram of the plurality of pictures and the absolute background is abnormally distributed (step S202), and satisfying the condition when the total number of pixels of the plurality of picture gray-scale histograms in the specific interval is greater than or equal to the total number of pixels of the absolute background gray-scale histogram First, as shown in formula (5):

Condition 1 (Condition1): (5)

Among them, CF is a multi-picture gray-scale histogram; BI is an absolute background gray-scale histogram; G _max is the gray-scale value corresponding to the highest peak of the multi-picture gray-scale histogram of the masking picture, and n is a parameter (the method is set in this case) 2, but should not be limited to this), θ _occlusion is the proportional value (this method is set to 1.5, but should not be limited to this).

After the condition 1 is satisfied, the edge detection of the gray image of the plurality of pictures and the absolute background is performed by the Sobel operator (step S203), and most of the edge information of the original plurality of pictures is caused when the lens is masked or painted. It will be lost. Therefore, the resulting image processed by the Sobel operator will be extremely lacking in edge information. Therefore, it is possible to determine whether the original plurality of pictures are masked or painted by the amount of missing edge information (step S204). Formula (6):

Condition 2 (Condition2):

S _CF ≦S _BI ×θ _edge (6)

Where S _CF is the total pixel edge of the plurality of pictures; S _BI is the total edge pixel of the absolute background, and θ _edge is the proportional value (the method is set to 0.7, but should not be limited thereto).

Finally, when the condition one is not satisfied, or the condition one is satisfied but the condition two is not satisfied, it is determined that the camera is turned or shifted, and a warning message is sent (step S111); if the condition one and the condition two are satisfied, then the judgment is made. The camera screen is shielded or painted, and a warning message is issued (step S111).

Regarding the detection of dense fog of the picture, the present invention uses the Lambert-Beer law to propose three judgment conditional expressions to determine whether the picture is a foggy picture.

Further, the brightness of the image is initially utilized to distinguish whether the captured image is likely to be foggy. The contrast generated by the image generated in the fog environment will be reduced. By analyzing the distribution of the histogram, the difference between the highest peak value on the left side of the average brightness value (ie, the gray level value) and the highest peak value on the right side will be smaller than the valve. The value PD _th needs to calculate the highest peak value and the highest peak value on the left side of the average brightness value, f _avg is the average brightness value of the image, M×N is the image size (the total number of image pixels), and f(x, y) is The pixel value at position (x, y) (abbreviated as f), n _f is the number of gray scale values in the histogram, f _L is the highest peak-to-peak value on the left side of the image average luminance value (f _avg ), and f _R is the image The average brightness value (f _avg ) is the highest peak-to-peak value on the right side. By analysis, the method determines the peak-to-peak difference of the average brightness value of the atomized image as a gap range, if the difference is lower than the threshold PD _th (this method is set If it is 30, but should not be limited thereto, then Condition 1 (Step S301) is satisfied, and as shown in Formula (7), an image which may be atomized may be initially determined.

Condition one: ∣f _L - f _R ∣≦ PD _th (7)

among them,

　　　　

　　　　

Condition 2 (Condition2) takes into account the smooth area of the picture, so it is for Sobel image (the result image processed by Sobel operator is called Sobel image, and this image will retain the edge information of the original image, so it can also be called The edge gradient change of the Sobel edge image is used to analyze the difference between the foggy and fog-free images (step S302), and the number of pixels of the Sobel edge image histogram in the grayscale value 250~255 and the grayscale value 255 is analyzed (step S303). When the number of grayscale values 250~255 and grayscale value 255 of the Sobel edge image histogram becomes higher, the image can be regarded as a fog-free phenomenon; when the number becomes lower, the image is regarded as foggy. If the formula (8) is met, it is determined that the image may be fogged, where n _{250≦f≦255} is the total number of grayscale values 250 to 255 of the Sobel image histogram, and n _{f = 255} is the grayscale value of the Sobel image histogram. The total number of 255, θ _Range is the threshold (this method is set to 0.001 × M × N, but not limited to this).

Condition 2: (Steps S304 and S305)

((n _{250≦f≦255} ) < θ _Range ) AND((n _{f =255} ) < θ _Range ) (8)

The atomized environment is a low-visibility image, so an atomized image is blurred and unclear. If the image is viewed from the perspective of the image, it is a smooth image of the property, while a clear image is used. Image that belongs to complex attributes. Therefore, the method of this method uses Sobel edge image detection to take its Sobel image as its average, as shown in formula (9), where S _μ represents the average value of the entire Sobel image, f _E (x, y) is expressed as the grayscale value of the position (x, y) in the Sobel image, and the gradient change of an atomized image is analyzed. A foggy image has a smaller S _{μ of} the Sobel edge image detection result; The S _μ of the Sobel edge image detection results in a fog-free environment is large.

(9)

In general, the standard deviation can be regarded as the degree of deviation of the image content. The larger the value of the standard deviation is, the larger the change of the image content is, which means that the contrast of the image is larger, and the information content is richer; the smaller the standard deviation is. , the degree of change is small, the contrast is relatively small, in general, the information content is more monotonous. Therefore, the method of this case uses the standard deviation (σ) as the basis for the classification of the fog concentration, as shown in equation (10), where σ represents the standard deviation of the original grayscale image, μ represents the average of the entire image, and f( x, y) is each pixel in the original image.

(10)

Condition 3 (Condition 3) is to determine the fog concentration by formula (9) and formula (10), and establish the attribution function of the fog level, as shown in formula (11):

Condition three: z = S _μ + σ(11)

Among them, the value of z is related to the fog concentration. If z = 0 to 44, it is set to "dense fog"; z = 45 to 75 is set to "mid fog"; z = 76 to 90 is set to "light fog"; > 90 is set to "no fog" (step S306), which can give the monitor as a reference for the degree of fog affecting the picture.

If the above three conditions are met (condition 1 to condition 3), the detected image has an abnormal state of fogging, and then a warning message is issued (step S307), and the warning message should include the concentration information of the fog.

In terms of detecting whether the picture is out of focus, the cause of the picture being out of focus may be that the device is damaged or mist, water drops fall in front of the lens, and the picture captured by the water droplets causes blurring, which may cause the picture to be out of focus. Since the image is out of focus mainly causes the image to be blurred, affecting the edge details of the image and causing the loss of high frequency information. The method uses Discrete Cosine Transform (DCT) to convert the image into the frequency domain, and then calculates the high frequency information to detect The picture is out of focus.

When the picture is out of focus, the amount of high-frequency information will be greatly reduced. The method of this method can determine whether the picture is out of focus by counting the loss of high-frequency information. First, the high frequency information reference value Base _{HF is} set by using a plurality of previous normal pictures, and then the threshold value α _{HF_Loss of the} high frequency information loss amount is set by the reference value, and α _{HF_Loss} = Base _HF × β _U , β _U is a proportional parameter ( The method in this case is set to β _U = 0.7, but should not be limited to this). When the high frequency information loss of the image taken by the camera is greater than α _{HF_Loss} , it is judged to be out of focus; otherwise, the high frequency information reference value is updated, and the screen out of focus program is continuously detected. (This method uses the number of coefficients in the DCT high-frequency coefficient region to be 0 to calculate the high-frequency information loss)

However, the color cast abnormal state of the screen is detected, and the color shift may occur because the scene color temperature causes the camera white balance error, or the input signal is abnormal, the signal is weak, the color is strange, or the screen is damaged due to the screen damage. Color shift, according to the acuity of the human eye to the color, the slight difference in color can not recognize the difference. When the channels of the picture have great differences, the overall picture can be clearly observed to a certain color. The method of the present invention proposes two judgment conditions to determine whether the picture has color cast. When both judgment conditions are satisfied, it is confirmed as the screen color shift.

When the color shift has not occurred in the picture, the average brightness of each channel of RGB is not too different from each other; on the contrary, when the color shift occurs on the picture, the average brightness of each channel of the picture is different, and the average brightness of each channel is calculated. The values (R _avg , G _avg , B _avg ), when the difference between the maximum average brightness L _{max of} each channel and the minimum average brightness L _min is greater than the threshold θ _cd (the method is set θ _cd = 30, but should not be used as this Limit), condition 1 is satisfied, as shown in formula (12), it can be preliminarily judged that the screen may have color cast.

Condition one:

(12)
among them ,
, ,
M×N represents the size of the image, and R(x, y), G(x, y), and B(x, y) respectively represent pixel values per point of each channel of the original image.

When it is initially determined that the RGB channels of the input image are different, in order to avoid misjudgment because the picture is only rich or poor in single color information, the second condition is to determine whether there is color cast by the local area average brightness value. First, look for the channel C _cd with the largest color shift, which is defined as the channel with the largest difference between the difference between one channel and the other. As shown in equation (13), B _diff , G _diff , R _diff are the other two channels. The difference. Use this step to find the channel with the largest color shift.

(13)

among them

C _cd identify the channel, and then determining the range of the bright portion C _cd channel (lighter range) and the range of a dark portion (darker range), the range for the bright portions of the channel C _cd brightest gray scale of (graylevel) to the first n gradations Range, the range of the dark part is the darkest gray scale of the C _cd channel to the next n gray scale range (the method is set to n = 5, which should not be limited to this). Calculate the average grayscale value of each channel in the bright range and the dark part range, and use the brightness range of the channel C _cd with the largest color shift and the average gray scale value of the dark part range, and the brightness range of the other channels and the average gray of the dark part range. The order value is subjected to a difference operation, as shown in equation (14). for versus The difference between the subtractions, for versus The difference between the subtractions, of which Is the average grayscale value of the C _cd dark range, Is the average grayscale value of the C _cd highlight range, The average grayscale value of each channel in the range of R, G, or B highlights, For the average grayscale value of each channel in the range of the R, G, or B dark portion, the above superscript c represents the channel: R, G, B.

;
(14)

Such as Greater than the threshold θ _cd and If it is greater than the threshold θ _cd (the method is set to θ _cd = 30, which should not be limited to this), the condition 2 is satisfied. As shown in the formula (15), it can be judged that the color shift occurs on the screen. Condition 1 is satisfied first, and if condition 2 is satisfied again, the warning message of the release color deviation is confirmed.

Condition 2: (15)

When the signal is subjected to high-frequency interference or the signal line is damaged, the most common abnormal condition is the flickering of the screen. The serious situation may cause the screen to be lost, and no signal can be accepted, resulting in a black screen phenomenon, making it impossible to smoothly monitor and monitor. Screen, a surveillance vulnerability has occurred. The method of the present invention utilizes the calculated average brightness value of the picture (ie, the gray level value), as shown in the formula (16), where L _avg represents the average value of the grayscale value of the image frame, and M×N is the total pixel of the image frame, f(x, y). ) is the grayscale value of the pixel (x, y). Since the average value of the grayscale value of the flashing picture and the normal stable picture will be significantly different, that is, the average value of the grayscale value of the blinking picture will be in a messy state, and the average value of the grayscale value of the normal stable picture will be A smooth and stable state is presented, so this difference can be used to detect whether the picture has flickering abnormalities.

(16)

The method of the present invention uses the difference between the grayscale value average curve of the flickering picture and the normal stable picture to determine whether the picture is flickering abnormal. However, there are many methods for judging the difference of the two graphs, and only one of them is as follows.

The invention utilizes the curve of the average value of the grayscale value of the above picture to calculate the number of state changes to determine whether the picture is flickering abnormality, and the state setting is as shown in the formula (17), wherein the state _t is the current state 0, 1 or unchanged. When the difference between the L _{avg of the} adjacent frames is greater than the threshold C _th (the method in this case sets C _th = 5, which should not be limited to this), the state changes; otherwise, the current state is maintained. When the state changes from 0 to 1 or from 1 to 0, the counter performs an accumulation operation, by judging the accumulated value of the counter (ACC), if the state is changed by successive n rounds, and the counter accumulated value is greater than the threshold value θ _A When it is judged that the screen has a flickering abnormality, a warning message is issued, as shown in the formula (18).
(17)

If (18)
among them

In view of the above, the detection method of the abnormality of the camera screen of the present invention is not possible by the prior art, and has indeed achieved the desired effect, and is not familiar with the skill of the artist, and is progressive. Applicability, obviously has met the requirements of the patent application, 提出 filed a patent application according to law, and asks your office to approve the application for this invention patent, in order to encourage creation, to the sense of virtue.

no

no

S101 to S111‧‧‧ steps

Claims (10)

[Item 1] A method for detecting an abnormality of a camera screen, comprising the following steps:
Draw a continuous number of pictures;
Determining whether the plurality of pictures are stable by the adjacent subtraction method;
Establishing two absolute backgrounds according to the plurality of pictures, the two absolute backgrounds are separated by a time difference, and are updated every predetermined time;
Determining, by background subtraction, the difference between each updated absolute background and its corresponding absolute background before updating;
When at least one of the absolute backgrounds before the update and the corresponding updated absolute background difference are greater than a first threshold, determining the camera according to the plurality of pictures and the gray-scale histogram distribution of the two absolute backgrounds and the edge pixels Whether the picture has an abnormal state of image shifting or being masked, and issuing a warning message corresponding to the abnormal state and performing a troubleshooting step of the corresponding abnormal state; and updating the absolute background and the corresponding updated one of the absolute before the update When the background difference is less than the first threshold, the following steps are included:
Determining whether the camera screen has an abnormal state of dense fog according to brightness information and edge information of the plurality of pictures;
Determining whether the camera screen has an abnormal state of out-of-focus according to the high-frequency information loss of the plurality of pictures;
Determining whether the camera screen has an abnormal state of color shift according to an average brightness value of each of the RGB channels of the plurality of pictures and an average brightness value of a local area of each of the RGB channels;
Determining whether the camera screen has an abnormal state of flicker according to the average brightness value of the plurality of pictures; and when the camera screen does not have an abnormal state of dense fog, out of focus, color shift, flicker or a combination thereof, respectively Each of the absolute backgrounds is updated by a preset time, and the warning message is issued when the camera screen has an abnormal state. [Item 2] The method for detecting an abnormality of a camera screen according to the first aspect of the patent application, wherein determining whether the camera screen is stable or not includes the following steps:
Obtaining a difference image by one of the pictures and another adjacent picture;
When the difference image has a larger number of pixels than a predetermined percentage of the total number of pixels, it is determined that the camera picture is in a stable state, and the two absolute backgrounds are established; and when the difference image is poor When the number of pixels with a larger value is higher than the total number of pixels of the preset percentage, it is determined that the camera screen is in a shaking state, and the warning message is issued. [Item 3] The method for detecting an abnormality of a camera screen as described in claim 1 of the patent application further includes the following steps:
Determining whether the total number of pixels of the grayscale histogram of the plurality of pictures is greater than the total number of pixels of the grayscale histogram of the absolute background; and when the total number of pixels of the grayscale histogram of the plurality of pictures is greater than or in a specific interval When the total number of pixels of the gray-scale histogram of the two absolute backgrounds is equal to, the total number of edge pixels of the plurality of pictures is determined to be less than or equal to a total number of edge pixels of a preset ratio of the absolute background. [Item 4] The method for detecting an abnormality of a camera screen as described in claim 3 of the patent application further includes the following steps:
When the absolute background difference is large, and the total number of pixels of the grayscale histogram of the plurality of pictures in a specific section is smaller than the total number of pixels of the grayscale histogram of the absolute background, it is determined that the camera screen is in an abnormal state of displacement, and is issued The warning message;
When the absolute background difference is large, and the total number of pixels of the grayscale histogram of the plurality of pictures in a specific section is not less than the total number of pixels of the grayscale histogram of the absolute background, and the total number of edge pixels of the plurality of pictures is greater than the absolute When the total number of edge pixels of the preset ratio is determined, the camera screen is judged to be in an abnormal state of displacement, and the warning message is issued; and when the absolute background difference is large, and the gray scale histogram of the plurality of pictures is in a specific interval The total number of pixels is not less than the total number of pixels of the gray-scale histogram of the absolute background, and when the total number of edge pixels of the plurality of pictures is not greater than the total number of edge pixels of the preset ratio of the absolute background, it is determined that the camera picture is suffering Mask the abnormal state and issue the alert message. [Item 5] The method for detecting an abnormality of a camera screen as described in claim 1 of the patent application further includes the following steps:
Determining whether the difference between the highest peak-to-peak value on the left side of the average brightness value of the plurality of pictures and the highest peak-to-peak value on the right side of the average brightness value is less than a preset gap threshold;
Determining whether the number of 250 to 255 grayscale values and 255 grayscale values of the edge image grayscale histogram of the plurality of pictures is less than a preset number threshold; and obtaining an edge image grayscale average by Sobel edge detection, and The warning value of the fog concentration is released according to the grayscale average value of the edge image of the plurality of pictures and the grayscale image standard deviation of the original picture to determine whether there is fog and fog concentration. [Item 6] The method for detecting an abnormality of a camera screen as described in claim 5 of the patent application further includes the following steps:
When the difference between the highest peak-to-peak value on the left side of the average brightness of the gray-scale histogram of the plurality of pictures and the highest peak-to-peak value on the right side of the average brightness is less than the preset gap threshold, the camera screen is initially judged to be in an abnormal state of fogging. . [Item 7] The method for detecting an abnormality of a camera picture as described in claim 6 of the patent application further includes the following steps:
When the number of the 250 to 255 grayscale value and the 255 grayscale value of the edge image histogram are both less than the preset number threshold, it is determined that the camera screen is in an abnormal state of fogging. [Item 8] The method for detecting an abnormality of a camera screen as described in claim 1 of the patent application further includes the following steps:
Setting a high frequency information amount reference value, and then setting a preset loss amount by using a high frequency information amount reference value;
Determining whether the high frequency information loss amount of the plurality of pictures is greater than the preset loss amount;
When it is greater than the preset loss amount, it is determined that the camera screen is in an abnormal state of out-of-focus, and the warning message is issued; and when it is not greater than the preset loss amount, the high-frequency information reference value is updated, and the next time Judging the ratio of the high frequency information loss of the plurality of pictures and the preset loss amount. [Item 9] The method for detecting an abnormality of a camera screen as described in claim 1 of the patent application further includes the following steps:
Determining whether a difference between a maximum average brightness of each color channel of the plurality of pictures and a minimum average brightness of each channel is greater than a predetermined average brightness difference;
Determining whether the difference between the average value of the bright portion of the color channel having the largest color shift and the average of the bright portion ranges of the other channels is greater than a predetermined average difference;
Determining whether the difference between the average value of the dark portion of the color channel having the largest color shift and the average of the dark portion of the other color channels is greater than the preset average difference;
When the difference between the maximum average brightness of each color channel and the minimum average brightness of each color channel is greater than the average brightness difference of the preset channel, and the average of the bright portion of the color channel with the largest color shift is brighter than other color channels The difference between the average value of the range and the difference between the average grayscale value of the dark portion of the color channel and the average of the dark portion of the other color channels is greater than the preset average difference, and the camera screen is judged. The color is in an abnormal state and the warning message is issued. [Item 10] The method for detecting an abnormality of a camera screen as described in claim 1 of the patent application further includes the following steps:
Determining a difference value of adjacent frames of the plurality of pictures;
When the difference value of the adjacent frames is greater than a second threshold, determining that the image state changes; and when the number of image state changes is greater than a preset accumulated number, determining that the camera screen is in an abnormal state of image flicker, and releasing The warning message.