Image sensor quality detection method, device and system

Abstract

The invention relates to the technical field of image sensor detection, solves the technical problem that a traditional method cannot perform multi-dimensional quality detection on an image sensor, and particularly relates to an image sensor quality detection method, device and system. The detection method comprises the following steps: constructing an image acquisition environment which keeps consistent illumination conditions when acquiring sequence images in a darkroom environment based on a sequence image acquisition device; setting exposure time, and acquiring a plurality of sequence images with different gray levels, which are output by the image sensor under different illumination intensities, in an acquisition environment; performing multi-dimensional detection on the sequence image and outputting a quality detection result; and determining whether the image sensor has defects according to the quality detection result. According to the invention, multi-dimensional detection can be carried out on whether the image sensor has defects or not, so that the image capturing performance of the image sensor under different conditions is accurately evaluated, the product performance is improved, and the image quality is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of image sensor detection, and in particular to an image sensor quality detection method, device and system. Background Art

[0002] Image sensor quality testing is mainly used to evaluate the performance of sensors in capturing images under different conditions, which is critical to ensuring image quality and product performance. The importance of image sensor quality testing. Currently, commonly used image sensor quality testing methods include a variety of technical means and processes. In actual testing, different experimental scenarios may need to be set up to evaluate the corresponding performance of image sensors.

[0003] Due to reasons such as manufacturing process, the original image output by the image sensor contains stripe noise and bad pixels, and the response linearity is poor. For example, there are an image quality detection method and system with publication number CN113628203A; a method and device for detecting image quality with publication number CN105357411A; a uniform light source platform, non-uniformity measurement, and correction method for camera calibration with publication number CN115037922 A; a method and device for detecting and correcting bad pixels of an image sensor with publication number CN116074495B; an image quality assessment method and system with publication number CN114419008A; and an image sensor quality detection method and system with publication number CN108737815B.

[0004] After searching by the applicant, it was found that the prior art does not provide a quality inspection method for different dimensions of image sensors, nor a supporting quality inspection device, which results in the inability to accurately reflect the quality of the image sensor, the inability to ensure the imaging quality under different light intensities, and difficulty in evaluating the defects of the image sensor and their impact on the imaging system. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides an image sensor quality detection method, device and system, which solve the technical problem that traditional methods cannot perform multi-dimensional quality detection on image sensors.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for detecting quality of an image sensor, the method comprising the following steps:

[0007] S1. Building an image acquisition environment based on a sequence image acquisition device to maintain consistent lighting conditions when acquiring sequence images in a darkroom environment;

[0008] S2, setting the exposure time, and collecting a plurality of sequential images with different gray levels output by the image sensor under different light intensities in the collection environment;

[0009] S3, performing multi-dimensional detection on the sequence images and outputting quality detection results, the multi-dimensional detection including stripe noise detection, linearity detection and bad pixel detection on the sequence images;

[0010] S4. Determine whether the image sensor has defects based on the quality inspection result.

[0011] Further, in step S2, the exposure time is set from the camera's minimum exposure to Exp, where Exp is the exposure time for 90% of the pixels of the image sensor to be overexposed, and P-1 time interval images are collected, where P ≥ 10 and P is an integer, and finally a total of P sequence images with different gray levels can be obtained.

[0012] Furthermore, in step S3, the specific process includes the following steps:

[0013] S31, performing stripe noise detection on the sequence images and outputting the stripe noise detection result;

[0014] S32, performing linearity detection on the sequence images and outputting the linearity detection result;

[0015] S33, performing bad pixel detection on the sequence image and outputting the bad pixel detection result;

[0016] S34, integrating the detection results of different dimensions to generate quality detection results, including stripe noise S, linearity deviation E, and bad pixel rate B1.

[0017] Furthermore, in step S31, the specific process includes the following steps:

[0018] S311, read any sequence image I to be detected, and subtract the image mean I mean The mean-removed image I1 is obtained, and the expression of the mean-removed image I1 is:

[0019] I1(x,y)=I(x,y)-I mean

[0020]

[0021] In the above formula, I1(x,y) is the pixel value with x as the horizontal coordinate and y as the vertical coordinate in the mean image I1; I(x,y represents the pixel value of the xth row and yth column in the sequence image I; M and N are the row height and column width of the sequence image I respectively;

[0022] S312, perform Fourier transform on each row and column of the mean-removed image I1 to obtain the frequency spectrum y[m][v] and y[n][v] of any row and column, where:

[0023]

[0024] In the above formula, y[m][v] represents the value of the Fourier transform of the mth row at the vth frequency point; y[n][v] represents the value of the Fourier transform of the nth column at the vth frequency point; M` and N` represent the row height and column width of the de-averaged image I1 respectively; y[m][n] represents the pixel value of the mth row and nth column of the de-averaged image I1; All represent the complex exponential kernel of Fourier transform, where i is the imaginary unit and v is the frequency variable, ranging from 0 to N-1;

[0025] S313, calculating the spectrum H(v) in the horizontal direction and the spectrum V(v) in the vertical direction of the de-averaged image I1 according to the spectrum y[m][v] and y[n][v], that is:

[0026]

[0027] In the above formula, H(v) represents the sum of the spectrum in the horizontal direction at the vth frequency point; V(v) represents the sum of the spectrum in the vertical direction at the vth frequency point; y * represents the complex conjugate;

[0028] S314, select the maximum spectrum peak H from the horizontal spectrum H(v) and the vertical spectrum V(v) respectively. max and V max As the stripe noise value S i , where S i =(H max ,V max );

[0029] S315, repeat steps S311-S314 to detect P sequential images in sequence, and take the average value to obtain the stripe noise value S of the image sensor. The calculation formula of the stripe noise value S is:

[0030]

[0031] In the above formula, P represents the number of sequential images collected; S i Represents the stripe noise value of the i-th image.

[0032] Furthermore, in step S32, the specific process includes the following steps:

[0033] S321, reading any sequence image I to be detected;

[0034] S322, determining whether a pixel point on the sequence image I is a central pixel;

[0035] If the pixel point belongs to the (MM*2*2%) row and (NN*2*2%) column, it is the center pixel and goes to step S323;

[0036] If not, return to step S321;

[0037] S323, setting the central pixel to include the linear grayscale range of the sequence to be detected Pix(a)-Pix(b) and the exposure time sequence Exp(a)-Exp(b);

[0038] S324, calculating the linearity deviation E of the central pixel sequence to be detected Pix(a) to Pix(b) i ;

[0039] S325, repeat steps S321-S324 to detect P sequential images in sequence, and take the average value to obtain the linearity deviation E of the image sensor, that is:

[0040]

[0041] In the above formula, E represents the average linearity deviation of the image sensor; E i Indicates the linear deviation of the central pixel; num indicates the number of pixels to be calculated.

[0042] Furthermore, in step S324, the specific process includes the following steps:

[0043] S3241. Use the least square method to perform linear fitting with the horizontal coordinate being the exposure time Exp(a) to Exp(b) and the vertical coordinate being Pix(a) to Pix(b) to obtain the fitting slope K and intercept B. The calculation formula is as follows:

[0044]

[0045]

[0046] In the above formula, K represents the slope of the fitted line; B represents the intercept of the fitted line; N = b-a+1 is the number of data points in the interval [a, b]; is the sum of all exposure times in the exposure time series; is the sum of all pixels of the pixel point in the sequence to be detected; It is the sum of the products of each pair of exposure time and pixel in the sequence; is the sum of the squares of all exposure times;

[0047] S3242, calculate the predicted value yi of the central pixel fitting line according to the slope K and the intercept B, that is:

[0048] yi=KPix(i)+B

[0049] In the above formula, Pix(i) represents the real pixel value at exposure time Exp(i);

[0050] S3243, calculate the difference ei between the real pixel value Pix(i) and the predicted value yi of the central pixel at the exposure time Exp(i), that is:

[0051] ei=Pix(i)-yi

[0052] S3244, calculate the linearity deviation E of the central pixel fitting line according to the difference ei i ,but:

[0053]

[0054] In the above formula, N represents the number of data points, that is, N = b-a+1; ei 2 It represents the square of the difference between the linear fitting prediction value of the central pixel at exposure time Exp(i) and the actual pixel value.

[0055] Furthermore, in step S33, the specific process includes the following steps:

[0056] S331, read any sequence image I to be detected, and draw a curve with the exposure time Exp(i) of the sequence image I as the horizontal coordinate and the pixel value of the pixel point as the vertical coordinate to obtain the light response curve Lp of each pixel point;

[0057] S332, calculate the image mean I of P sequence images I mean , and the exposure time Exp(i) is used as the horizontal coordinate, and the image mean I of different sequence images I mean The sequence is used as the ordinate to draw a curve to obtain the mean response curve Lmean;

[0058] S333, performing bad pixel detection according to the light response curve Lp of each pixel;

[0059] If Lp_min>150 and |Lmean_min-Lp_min|>50, the pixel is a bright spot;

[0060] If Lp_max≠255, the pixel is a dark point;

[0061] S334, performing response abnormal point detection according to the light response curve Lp of each pixel point;

[0062] S335, repeating steps S332-S334 to detect each pixel in the sequence image I and counting the number of bad pixels Num;

[0063] S336, calculating the bad pixel rate B1 of the image sensor according to the bad pixel number Num, that is:

[0064]

[0065] In the above formula, M represents the row height of the sequence image, and N represents the column width of the sequence image.

[0066] Furthermore, in step S334, the specific process includes the following steps:

[0067] S3341, using the least squares method to perform linear fitting on the light response curve Lp to obtain a slope K1, and to perform linear fitting on the mean response curve Lmean to obtain a slope K2, the calculation method is as follows:

[0068]

[0069] In the above formula, K1 and K2 are the linear fitting slopes of the pixel and mean image respectively; Exp(i) represents the i-th exposure time; Lp(i) represents the pixel value of the pixel at the exposure time Exp(i); P represents the total number of sequence images of different levels collected; Lmean(i) represents the image mean at the exposure time Exp(i);

[0070] S3342, calculate the response ratio K12 according to the slopes K1 and K2, that is:

[0071]

[0072] S3343, judging whether the pixel point is a response abnormal point according to the response ratio K12;

[0073] If K12 <K min or K12>K max , then the pixel is a response abnormal point;

[0074] If not, proceed to step S3344;

[0075] S3344, determining whether the light response curve Lp corresponding to the pixel point is an increasing sequence;

[0076] If so, then end;

[0077] If not, the pixel is a response abnormal point.

[0078] By means of the above technical solution, the present invention provides an image sensor quality detection method, device and system, which have at least the following beneficial effects:

[0079] 1. The present invention realizes multi-dimensional quality inspection of image sensors and has strong operability. It only needs to collect a set of images to output the quality inspection results of image sensors, which can quickly realize the screening of defective image sensors, ensure the imaging quality of sensors under different light intensities, and ensure the stability and reliability of the imaging system.

[0080] 2. The present invention can perform multi-dimensional detection on whether there are defects in image sensing, thereby accurately evaluating the performance of the image sensor in capturing images under different conditions, improving product performance and ensuring image quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0082] Figure 1 is a flow chart of the image sensor quality detection method of the present invention;

[0083] Figure 2 It is a structural diagram of the sequential image acquisition device of the present invention;

[0084] Figure 3 It is a flow chart of stripe noise detection of the present invention;

[0085] Figure 4 It is a flow chart of linearity detection of the present invention;

[0086] Figure 5 A flowchart of bad pixel detection of the present invention;

[0087] Figure 6 It is a principle block diagram of the image sensor quality detection system of the present invention.

[0088] In the figure: 1. Integrating sphere; 2. Light source controller; 3. Optical platform; 4. Collimation sleeve; 5. Manual lifting platform; 6. Darkroom. DETAILED DESCRIPTION

[0089] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific implementation methods, so that the implementation process of how the present application uses technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly.

[0090] Please refer to Figure 1-Figure 5This embodiment proposes an image sensor quality detection method, which uses a sequence image acquisition device to build an image acquisition environment and complete the acquisition of different exposure sequence images, implements stripe noise detection, linearity detection and bad pixel detection for the acquired sequence images, and finally outputs the quality detection result of the image sensor, thereby realizing multi-dimensional quality detection of the image sensor, and has strong operability, and only needs to acquire a set of images to output the quality detection result of the image sensor. The method includes the following steps:

[0091] S1. An image acquisition environment is built based on a sequence image acquisition device to maintain consistent lighting conditions when acquiring sequence images in a darkroom environment; the image acquisition environment built is as follows: Figure 2 As shown, in order to ensure consistent lighting conditions during detection and reduce interference from ambient light, all detection images need to be collected in a darkroom environment. The device proposed in the present invention can realize the collection of image sensor quality detection sequence images. The specific operation process is as follows: the integrating sphere light source controller is used to stabilize the integrating sphere brightness to a fixed value. The circular light outlet of the integrating sphere is wrapped by one side of the collimating sleeve entrance. Black matte material is used inside and outside the collimating sleeve to avoid interference from stray light on imaging and to ensure the directionality and stability of the light beam as much as possible. Then, the camera equipped with the image sensor to be tested is placed on the right manual lifting platform, and the height of the lifting platform is adjusted to ensure that the center of the image sensor is at the same height as the center of the sleeve light outlet.

[0092] S2. Set the exposure time to collect multiple sequential images with different grayscale levels output by the image sensor under different light intensities in the acquisition environment. Since when the light intensity of the integrating sphere is constant, adjusting the exposure time can change the amount of light radiation received by the image sensor, thereby affecting the pixel grayscale of the acquired image. Therefore, the exposure time is set from the camera's minimum exposure to Exp, where Exp is the exposure time when 90% of the image sensor's pixels are overexposed (the grayscale of the 8-bit image reaches 255). Collect P-1 time interval images, P ≥ 10 and P is an integer, and finally obtain a total of P sequential images with different grayscale levels, and store the images in the storage unit of this system.

[0093] S3, performing multi-dimensional detection on the sequence images and outputting quality detection results, the multi-dimensional detection includes performing stripe noise detection, linearity detection and bad pixel detection on the sequence images; in step S3, the specific process includes the following steps:

[0094] S31, performing stripe noise detection on the sequence image and outputting the stripe noise detection result; Figure 3 As shown, in step S31, the specific process includes the following steps:

[0095] S311, read any sequence image I to be detected, and subtract the image mean I meanThe de-averaged image I1 is obtained. The reason for using the de-averaged image in this embodiment is to remove the strong low-frequency components generated by the DC component in the image in the spectrum diagram and enhance the influence of the high-frequency components. The expression of the de-averaged image I1 is:

[0096] I1(x,y)=I(x,y)-I mean

[0097]

[0098] In the above formula, I1(x,y) is the pixel value with x as the horizontal coordinate and y as the vertical coordinate in the mean image I1; I(x,y) represents the pixel value of the xth row and yth column in the sequence image I; M and N are the row height and column width of the sequence image I respectively.

[0099] S312, perform Fourier transform on each row and column of the mean-removed image I1 to obtain the frequency spectrum y[m][v] and y[n][v] of any row and column, where:

[0100]

[0101] In the above formula, y[m][v] represents the value of the Fourier transform of the mth row at the vth frequency point; y[n][v] represents the value of the Fourier transform of the nth column at the vth frequency point; M` and N` represent the row height and column width of the de-averaged image I1 respectively; y[m][n] represents the pixel value of the mth row and nth column of the de-averaged image I1; Both represent the complex exponential kernel of Fourier transform, where i is the imaginary unit and v is the frequency variable, ranging from 0 to N-1.

[0102] S313, calculate the horizontal spectrum H(v) and the vertical spectrum V(v) of the mean-removed image I1 according to the spectrum y[m][v] and y[n][v], and sum the spectrum y[m][v] of all rows and the spectrum y[n][v] of all columns to obtain the horizontal spectrum H(v) and the vertical spectrum V(v), that is:

[0103]

[0104] In the above formula, H(v) represents the sum of the spectrum in the horizontal direction at the vth frequency point; V(v) represents the sum of the spectrum in the vertical direction at the vth frequency point; y * represents the complex conjugate;

[0105] S314, select the maximum spectrum peak H from the horizontal spectrum H(v) and the vertical spectrum V(v) respectively. max and V max As the stripe noise value S i , where Si =(H max ,V max ); specifically, H max Indicates the maximum peak value of the spectrum. It should be noted that since the power spectrum is even symmetrical, the peak value can be calculated for only half of the spatial frequency curve. max It indicates the vertical stripe noise level. The larger the value, the more serious the stripes. max Indicates the maximum peak value of the vertical spectrum, V max It can represent the horizontal stripe noise level. Therefore, H max =max(H(v)),V max =max(V(v)), finally take H max and V max The larger value S i Indicates the fringe noise level.

[0106] S315, repeat steps S311-S314 to detect P sequential images in sequence, and take the average value to obtain the stripe noise value S of the image sensor; by detecting all the sequential images in sequence, and finally taking the average value S of all the results to represent the stripe noise level of the image sensor to be detected, the higher S means the more serious the image stripe noise is, and the lower the image quality is. The calculation formula of the stripe noise value S is:

[0107]

[0108] In the above formula, P represents the number of sequential images collected; S i Represents the stripe noise value of the i-th image.

[0109] S32, perform linearity detection on the sequence images and output the linearity detection result; Figure 4 As shown, in step S32, the specific process includes the following steps:

[0110] S321, read any sequence image I to be detected; first, read a total of P sequence images with different uniform gray levels to be detected, and each pixel has a total of P pixel values ​​Pix(1) to Pix(p).

[0111] S322, determine whether the pixel point on the sequence image I is the central pixel; the illumination of the integrating sphere is generally the most uniform and symmetrical at the center position. The integrating sphere image collected by the image sensor may produce edge response due to reflection, occlusion, etc. Although the image acquisition device provided in the previous article can minimize this phenomenon, it is still unavoidable. Therefore, when using the collected sequence image to detect the linearity of the image sensor, the central pixel can be selected to avoid interference caused by the edge effect. This embodiment selects a ratio of 2%. If the pixel to be calculated does not belong to the (MM*2*2%) row or (NN*2*2%) column, it will not participate in the subsequent linearity detection calculation, and the number of pixels to be calculated is num=(MM*2*2%)*(NN*2*2%), where M and N are the row height and column width of the sequence image I, respectively.

[0112] If the pixel point belongs to the (MM*2*2%) row and (NN*2*2%) column, it is the center pixel and the process goes to step S323;

[0113] If not, return to step S321;

[0114] S323, set the linear grayscale range of the central pixel to include the sequence to be detected Pix(a)~Pix(b) and the exposure time sequence Exp(a)~Exp(b); assume that for a response analysis of a central pixel in the 8-bit image, when Pix(1)~Pix(p) belongs to the grayscale interval [255*10%,255*90%], it is retained, and the corresponding sequence of retained data points is [a,b]. Then the sequence of pixels to be detected is Pix(a)~Pix(b), and the exposure time sequence is Exp(a)~Exp(b).

[0115] S324, calculating the linearity deviation E of the central pixel sequence to be detected Pix(a) to Pix(b) i ; In step S324, the specific process includes the following steps:

[0116] S3241. Use the least square method to perform linear fitting with the horizontal coordinate being the exposure time Exp(a) to Exp(b) and the vertical coordinate being Pix(a) to Pix(b) to obtain the fitting slope K and intercept B. The calculation formula is as follows:

[0117]

[0118] In the above formula, K represents the slope of the fitted line; B represents the intercept of the fitted line; N = b-a+1 is the number of data points in the interval [a, b]; is the sum of all exposure times in the exposure time series; is the sum of all pixels of the pixel point in the sequence to be detected; It is the sum of the products of each pair of exposure time and pixel in the sequence; is the sum of the squares of all exposure times;

[0119] S3242, calculate the predicted value yi of the central pixel fitting line according to the slope K and the intercept B, that is:

[0120] yi=KPix(i)+B

[0121] In the above formula, Pix(i) represents the real pixel value at exposure time Exp(i);

[0122] S3243, calculate the difference ei between the real pixel value Pix(i) and the predicted value yi of the central pixel at the exposure time Exp(i), that is:

[0123] ei=Pix(i)-yi

[0124] S3244, calculate the linearity deviation E of the central pixel fitting line according to the difference ei i ,but:

[0125]

[0126] In the above formula, N represents the number of data points, that is, N = b-a+1; ei 2 It represents the square of the difference between the linear fitting prediction value of the central pixel at exposure time Exp(i) and the actual pixel value.

[0127] S325, repeat steps S321-S324 to detect P sequential images in sequence, and take the average value to obtain the linearity deviation E of the image sensor; in this embodiment, the detection steps S321-S324 are completed in sequence for the central pixels of all sequential images, and finally all linearity deviations E are taken. i The average value E is the linearity deviation result of the image sensor to be tested. The smaller the linearity deviation E, the better. The smaller the linearity deviation E, the better the response linearity of the image sensor, which is more conducive to the subsequent linear processing of the image.

[0128]

[0129] In the above formula, E represents the average linearity deviation of the image sensor; E i Indicates the linear deviation of the central pixel; num indicates the number of pixels to be calculated.

[0130] S33, performing bad pixel detection on the sequence image and outputting the bad pixel detection result; Figure 5 As shown, in step S33, the specific process includes the following steps:

[0131] S331, read any sequence image I to be detected, and draw a curve with the exposure time Exp(i) of the sequence image I as the horizontal coordinate and the pixel value of the pixel point as the vertical coordinate to obtain the light response curve Lp of each pixel point;

[0132] S332, calculate the image mean I of P sequence images I mean , and the exposure time Exp(i) is used as the horizontal coordinate, and the image mean I of different sequence images I mean The sequence is used as the ordinate to draw a curve to obtain the mean response curve Lmean;

[0133] S333, performing bad pixel detection according to the light response curve Lp of each pixel; specifically, first performing dark and bright point judgment, that is:

[0134] The pixel responds to a minimum value Lp_min=min(Lp), and a mean value responds to a minimum value Lmean_min=min(Lp). When Lp_min>150 and |Lmean_min-Lp_min|>50, the pixel is a bright spot. If it is not a bright spot, the pixel responds to a maximum value Lp_max=max(Lp). When Lp_max is not equal to 255 (8-bit image), the pixel is a dark spot. The above specific values ​​can be adjusted according to actual conditions, and this embodiment does not impose any limitation.

[0135] If Lp_min>150 and |Lmean_min-Lp_min|>50, the pixel is a bright spot;

[0136] If Lp_max≠255, the pixel is a dark point;

[0137] S334, performing response abnormal point detection according to the light response curve Lp of each pixel point; in step S334, the specific process includes the following steps:

[0138] S3341, using the least squares method to perform linear fitting on the light response curve Lp to obtain a slope K1, and to perform linear fitting on the mean response curve Lmean to obtain a slope K2, the calculation method is as follows:

[0139]

[0140] In the above formula, K1 and K2 are the linear fitting slopes of the pixel and mean image respectively; Exp(i) represents the i-th exposure time; Lp(i) represents the pixel value of the pixel at the exposure time Exp(i); P represents the total number of sequence images of different levels collected; Lmean(i) represents the image mean at the exposure time Exp(i);

[0141] S3342, calculate the response ratio K12 according to the slopes K1 and K2, that is:

[0142]

[0143] S3343. Determine whether the pixel is a response anomaly point according to the response ratio K12;

[0144] If K12 < K min or K12 > K max , then this pixel is a response anomaly point; when the response ratio K12 is too small or too large, it indicates that the difference between the pixel and the average response of the image sensor is too large, belonging to abnormal response. Therefore, if K12 < 0.1 or K12 > 4, it indicates that this pixel is a response anomaly point, and the specific value can be adjusted according to the actual situation, and no specific limitation is made here.

[0145] If the opposite is true, then go to step S3344;

[0146] S3344. Determine whether the light response curve Lp corresponding to this pixel is an increasing sequence; specifically, assume that the response pixel value sequence of this pixel is Lp = [Lp1, Lp1, …, Lpn], when Lpi < Lpi+1 holds for all i, indicating that the light response curve Lp is an increasing sequence. If there exists any i such that Lpi > Lpi+1, then the light response curve Lp is not increasing.

[0147] If so, end;

[0148] If not, then this pixel is a response anomaly point; if the light response curve Lp is not an increasing sequence, it indicates that the gray level of this pixel is abnormal under a specific light intensity, belonging to a response anomaly point.

[0149] S335. Repeat steps S332 - S334 to detect each pixel point in the sequence image I and count the number of bad points Num;

[0150] S336. If the pixel point neither belongs to a bright point nor a dark point, nor a response anomaly point, then it is a non-bad point. As long as it belongs to one of them, it is a bad point. Calculate the bad point rate B1 of the image sensor according to the number of bad points Num, that is:

[0151]

[0152] In the above formula, M represents the row height of the sequence image, and N represents the column width of the sequence image;

[0153] The bad point rate B1 is the bad point detection result of the image sensor to be detected. The smaller the value of B1, the better. The smaller the value, the fewer the number of bad points in the original image output by the image sensor, and the better the pixel quality of the image sensor.

[0154] S34, integrating the detection results of different dimensions to generate quality detection results, including stripe noise S, linearity deviation E, and bad pixel rate B1; by integrating the detection results of different dimensions, output the quality detection results of the final image sensor. Specifically, all the detection results are plotted in a table, including stripe noise S, linearity deviation E, and bad pixel rate B1. Among them:

[0155] The smaller the stripe noise S is, the better. When the stripe noise S is smaller, the stripe noise level of the image sensor is lower, the uniformity of the output original image is better, and the image quality is higher;

[0156] The smaller the linearity deviation E, the better. When the linearity deviation S is smaller, the linearity deviation of the image sensor is lower, the response linearity is higher, and it is more conducive to the subsequent linear processing of the image.

[0157] The smaller the bad pixel rate B1 is, the better. When the bad pixel rate B1 is smaller, the bad pixel rate of the image sensor is lower, the number of bad pixels is smaller, the number of abnormal pixels is smaller, and the quality of the original pixel is higher.

[0158] S4. Determine whether the image sensor has defects based on the quality inspection results. In this embodiment, the quality inspection results of the image sensor can be determined based on the generated quality inspection results, wherein the larger the stripe noise S, linearity deviation E and bad pixel rate B1, the more serious the quality defects of the image sensor. The values ​​of the three different dimensions can reflect whether the image sensor has quality defects in different dimensions. Of course, when all the quality inspection results are within the standard range, the image sensor does not have any quality defects.

[0159] The present invention realizes multi-dimensional quality inspection of image sensors and has strong operability. It only needs to collect a set of images to output the quality inspection results of the image sensors. It can quickly screen out defective image sensors, ensure the imaging quality of the sensors under different light intensities, and guarantee the stability and reliability of the imaging system.

[0160] like Figure 2 As shown, this embodiment further proposes a sequential image acquisition device, which includes a darkroom 6, an optical platform 3 is arranged at the bottom of the darkroom 6, and an integrating sphere 1 and a light source controller 2 for stabilizing the brightness of the integrating sphere 1 to a fixed value are respectively installed on one side of the optical platform 3; a collimating sleeve 4 with a circular light outlet wrapped therein is arranged on one side of the integrating sphere 1, and a manual lifting platform 5 located on the optical platform 3 is arranged on one side of the other end of the collimating sleeve 4, and a camera loaded with an image sensor to be tested is placed on the manual lifting platform 5, and the height of the manual lifting platform 5 is adjusted to ensure that the center of the image sensor is at the same height as the center of the light outlet of the collimating sleeve 4.

[0161] Specifically, in order to ensure consistent lighting conditions during detection and reduce interference from ambient light, all detection images need to be collected in a darkroom environment. The sequence image acquisition device proposed in the present invention can realize the acquisition of sequence images for image sensor quality detection. The specific operation process is as follows: the brightness of the integrating sphere 1 is stabilized at a fixed value using the light source controller 2. The circular light outlet of the integrating sphere 1 is wrapped by one side of the entrance of the collimating sleeve 4. The collimating sleeve 4 uses black matte material inside and outside to avoid interference from stray light on imaging and to ensure the directionality and stability of the light beam as much as possible. Then, the camera equipped with the image sensor to be tested is placed on the manual lifting platform 5 on the right side. The height of the manual lifting platform 5 is adjusted to ensure that the center of the image sensor is consistent with the height of the center of the light outlet of the collimating sleeve 4.

[0162] like Figure 6 As shown, this embodiment also proposes an image sensor quality detection system, which includes a sequence image acquisition unit for completing different exposure sequence image acquisition through a sequence image acquisition device; a storage unit for storing the acquired sequence images; a quality detection unit for performing multi-dimensional detection on the acquired sequence images and generating corresponding detection results, the quality detection unit including a stripe noise detection unit, a linearity detection unit and a bad pixel detection unit; a detection result output unit for summarizing the detection results of different dimensions and outputting the final image sensor quality detection result.

[0163] Specifically, the sequence image acquisition device provided by this patent is used to complete the acquisition of different exposure sequence images, and realize stripe noise detection, linearity detection, and bad pixel detection, and finally output the quality detection result of the image sensor, wherein the stripe detection method is to analyze the significant peaks of the row and column power spectra of the mean-removed image one by one, and use the peak values ​​to characterize the stripe noise level; the linearity detection method is to analyze the sensor pixel by pixel, and calculate the linearity deviation of the central pixel in combination with the least squares method; the bad pixel detection method is to determine whether it is a dark spot, a bright spot or a response abnormal point based on the light response curve of the pixel, and finally summarize the detection results of different dimensions to output the final image sensor quality detection result.

[0164] The image sensor quality detection method, device and system proposed in the present invention can perform multi-dimensional detection on whether there are defects in image sensing, so as to accurately evaluate the image sensor's performance in capturing images under different conditions, thereby improving product performance and ensuring image quality. The importance of image sensor quality detection is mainly reflected in the following aspects:

[0165] Ensure imaging quality under various lighting conditions; identify image sensor defects and their impact on the imaging system; use the test results to screen defective products, improve production processes, and enhance quality control levels; high-quality image sensors can support higher computing accuracy, more functional integration, and improve product versatility and user experience.

[0166] Those skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a program, so the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0167] The above implementation methods have been described in detail. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for detecting quality of an image sensor, characterized in that: The method comprises the following steps: S1. Building an image acquisition environment based on a sequence image acquisition device to maintain consistent lighting conditions when acquiring sequence images in a darkroom environment; S2, setting the exposure time, and collecting a plurality of sequential images with different gray levels output by the image sensor under different light intensities in the collection environment; S3, performing multi-dimensional detection on the sequence images and outputting quality detection results, the multi-dimensional detection including stripe noise detection, linearity detection and bad pixel detection on the sequence images; S4. Determine whether the image sensor has defects based on the quality inspection result.

2. The image sensor quality detection method according to claim 1, characterized in that: In step S2, the exposure time is set from the camera's minimum exposure to Exp, where Exp is the exposure time for 90% of the image sensor's pixels to be overexposed, and P-1 time interval images are collected, where P ≥ 10 and P is an integer, and finally a total of P sequence images with different gray levels can be obtained.

3. The image sensor quality detection method according to claim 1, characterized in that: In step S3, the specific process includes the following steps: S31, performing stripe noise detection on the sequence images and outputting the stripe noise detection result; S32, performing linearity detection on the sequence images and outputting the linearity detection result; S33, performing bad pixel detection on the sequence image and outputting the bad pixel detection result; S34, integrating the detection results of different dimensions to generate quality detection results, including stripe noise S, linearity deviation E, and bad pixel rate B1.

4. The image sensor quality detection method according to claim 3, characterized in that: In step S31, the specific process includes the following steps: S311, read any sequence image I to be detected, and subtract the image mean I mean The mean-removed image I1 is obtained, and the expression of the mean-removed image I1 is: I1(x,y)=I(x,y)-I mean In the above formula, I1(x,y) is the pixel value with x as the horizontal coordinate and y as the vertical coordinate in the mean image I1; I(x,y represents the pixel value of the xth row and yth column in the sequence image I; M and N are the row height and column width of the sequence image I respectively; S312, perform Fourier transform on each row and column of the mean-removed image I1 to obtain the frequency spectrum y[m][v] and y[n][v] of any row and column, where: In the above formula, y[m][v] represents the value of the Fourier transform of the mth row at the vth frequency point; y[n][v] represents the value of the Fourier transform of the nth column at the vth frequency point; M` and N` represent the row height and column width of the de-averaged image I1 respectively; y[m][n] represents the pixel value of the mth row and nth column of the de-averaged image I1; All represent the complex exponential kernel of Fourier transform, where i is the imaginary unit and v is the frequency variable, ranging from 0 to N-1; S313, calculating the spectrum H(v) in the horizontal direction and the spectrum V(v) in the vertical direction of the de-averaged image I1 according to the spectrum y[m][v] and y[n][v], that is: In the above formula, H(v) represents the sum of the spectrum in the horizontal direction at the vth frequency point; V(v) represents the sum of the spectrum in the vertical direction at the vth frequency point; y * represents the complex conjugate; S314, select the maximum spectrum peak H from the horizontal spectrum H(v) and the vertical spectrum V(v) respectively. max and V max As the stripe noise value S i , where S i =(H max ,V max ); S315, repeat steps S311-S314 to detect P sequential images in sequence, and take the average value to obtain the stripe noise value S of the image sensor. The calculation formula of the stripe noise value S is: In the above formula, P represents the number of sequential images collected; S i Represents the stripe noise value of the i-th image.

5. The image sensor quality detection method according to claim 3, characterized in that: In step S32, the specific process includes the following steps: S321, reading any sequence image I to be detected; S322, determining whether a pixel point on the sequence image I is a central pixel; If the pixel point belongs to the (MM*2*2%) row and (NN*2*2%) column, it is the center pixel and the process goes to step S323; If not, return to step S321; S323, setting the central pixel to include the linear grayscale range of the sequence to be detected Pix(a)-Pix(b) and the exposure time sequence Exp(a)-Exp(b); S324, calculating the linearity deviation E of the central pixel sequence to be detected Pix(a) to Pix(b) i ; S325, repeat steps S321-S324 to detect P sequential images in sequence, and take the average value to obtain the linearity deviation E of the image sensor, that is: In the above formula, E represents the average linearity deviation of the image sensor; E i Indicates the linear deviation of the central pixel; num indicates the number of pixels to be calculated.

6. The image sensor quality detection method according to claim 5, characterized in that: In step S324, the specific process includes the following steps: S3241. Use the least square method to perform linear fitting with the horizontal coordinate being the exposure time Exp(a) to Exp(b) and the vertical coordinate being Pix(a) to Pix(b) to obtain the fitting slope K and intercept B. The calculation formula is as follows: In the above formula, K represents the slope of the fitted line; B represents the intercept of the fitted line; N = b-a+1 is the number of data points in the interval [a, b]; is the sum of all exposure times in the exposure time series; is the sum of all pixels of the pixel point in the sequence to be detected; It is the sum of the products of each pair of exposure time and pixel in the sequence; is the sum of the squares of all exposure times; S3242, calculate the predicted value yi of the central pixel fitting line according to the slope K and the intercept B, that is: yi=KPix(i)+B In the above formula, Pix(i) represents the real pixel value at exposure time Exp(i); S3243, calculate the difference ei between the real pixel value Pix(i) and the predicted value yi of the central pixel at the exposure time Exp(i), that is: ei=Pix(i)-yi S3244, calculate the linearity deviation E of the central pixel fitting line according to the difference ei i ,but: In the above formula, N represents the number of data points, that is, N = b-a+1; ei 2 It represents the square of the difference between the linear fitting prediction value of the central pixel at exposure time Exp(i) and the actual pixel value.

7. The image sensor quality detection method according to claim 3, characterized in that: In step S33, the specific process includes the following steps: S331, read any sequence image I to be detected, and draw a curve with the exposure time Exp(i) of the sequence image I as the horizontal coordinate and the pixel value of the pixel point as the vertical coordinate to obtain the light response curve Lp of each pixel point; S332, calculate the image mean I of P sequence images I mean , and the exposure time Exp(i) is used as the horizontal coordinate, and the image mean I of different sequence images I mean The sequence is used as the ordinate to draw a curve to obtain the mean response curve Lmean; S333, performing bad pixel detection according to the light response curve Lp of each pixel; If Lp_min>150 and |Lmean_min-Lp_min|>50, the pixel is a bright spot; If Lp_max≠255, the pixel is a dark point; S334, performing response abnormal point detection according to the light response curve Lp of each pixel point; S335, repeating steps S332-S334 to detect each pixel in the sequence image I and counting the number of bad pixels Num; S336, calculating the bad pixel rate B1 of the image sensor according to the bad pixel number Num, that is: In the above formula, M represents the row height of the sequence image, and N represents the column width of the sequence image.

8. The image sensor quality detection method according to claim 7, characterized in that: In step S334, the specific process includes the following steps: S3341, using the least squares method to perform linear fitting on the light response curve Lp to obtain a slope K1, and to perform linear fitting on the mean response curve Lmean to obtain a slope K2, the calculation method is as follows: In the above formula, K1 and K2 are the linear fitting slopes of the pixel and mean image respectively; Exp(i) represents the i-th exposure time; Lp(i) represents the pixel value of the pixel at the exposure time Exp(i); P represents the total number of sequence images of different levels collected; Lmean(i) represents the image mean at the exposure time Exp(i); S3342, calculate the response ratio K12 according to the slopes K1 and K2, that is: S3343, judging whether the pixel point is a response abnormal point according to the response ratio K12; If K12 <K min or K12>K max , then the pixel is a response abnormal point; If not, proceed to step S3344; S3344, determining whether the light response curve Lp corresponding to the pixel point is an increasing sequence; If so, then end; If not, the pixel is a response abnormal point.

9. The sequential image acquisition device according to claim 1, characterized in that: The device comprises a darkroom (6), an optical platform (3) is arranged at the bottom of the darkroom (6), and an integrating sphere (1) and a light source controller (2) for stabilizing the brightness of the integrating sphere (1) to a fixed value are respectively installed on one side of the optical platform (3); A collimating sleeve (4) whose circular light outlet is wrapped is arranged on one side of the integrating sphere (1), and a manual lifting platform (5) located on an optical platform (3) is arranged on one side of the other end of the collimating sleeve (4). A camera loaded with an image sensor to be tested is placed on the manual lifting platform (5), and the height of the manual lifting platform (5) is adjusted to ensure that the center of the image sensor and the center of the light outlet of the collimating sleeve (4) are at the same height.

10. A system for implementing the image sensor quality detection method according to any one of claims 1 to 8, characterized in that: The system includes: A sequence image acquisition unit, used to complete the acquisition of different exposure sequence images through a sequence image acquisition device; A storage unit, used for storing the acquired sequence images; A quality detection unit, used to perform multi-dimensional detection on the acquired sequence images and generate corresponding detection results, the quality detection unit includes a stripe noise detection unit, a linearity detection unit and a bad pixel detection unit; The detection result output unit is used to summarize the detection results of different dimensions and output the final quality detection result of the image sensor.

Images