CN114554188A - Mobile phone camera detection method and device based on image sensor pixel array - Google Patents

Abstract

The invention discloses a mobile phone camera detection method based on an image sensor pixel array, which is characterized in that the characteristics of a target camera area or a gray level step in an image before and after the active irradiation of an external infrared light source are obtained to realize the target detection and obtain an input image, the input image is preprocessed to improve the signal-to-noise ratio of the image, a suspicious target seed point is obtained by using a target segmentation algorithm of a histogram and an image entropy, a complete highlight area is obtained, the discrimination of a suspicious target is screened by discriminating the target through a feature descriptor of the target camera, the repeated detection condition of the target is determined through a detection clustering algorithm, and the detection result is displayed in an RGB color image. Changing the shape of the target mask reduces the accuracy of later stage feature analysis and calculation, affecting the final target decision. The fact that the re-detection targets are necessarily close in physical distance is utilized, the judgment results are clustered, the situation of repeated detection of the targets is eliminated, and the precision and the efficiency of camera detection are improved to a certain extent.

Description

Mobile phone camera detection method and device based on image sensor pixel array

technical field

The invention belongs to the technical field of image processing, and in particular relates to a mobile phone camera detection method and device based on an image sensor pixel array.

Background technique

With the increasing demand for smart electronic product terminals and automotive electronics, the domestic market demand for camera modules has gradually increased. The mainstream configuration of the front camera has reached the level of 8 million pixels or even 10 million pixels. The rear cameras of major mobile phone manufacturers are equipped with autofocus function, and some manufacturers are already trying to apply the optical zoom function to smartphones. , The autofocus function of the micro camera module will greatly improve the focusing efficiency and reduce the space occupied. In order to improve the user's experience when taking pictures and solve the problem of blurred photos, the focus motor is gradually transformed from a driven shrapnel focus motor to a closed-loop focus motor. Optical image stabilization focusing motor. Since the camera recognition technology is mainly used to distinguish true and false targets, the variety of false alarms leads to the poor robustness of traditional target recognition methods. The camera recognition algorithm of deep learning needs to select an appropriate model and a reasonable training method, and also requires a large number of sample sets. As input, there is no camera public dataset trained model available, reducing the efficiency and accuracy of camera detection.

SUMMARY OF THE INVENTION

In view of this, the present invention provides a mobile phone camera detection method and device based on an image sensor pixel array, which solves the problem of effectively detecting and screening out abnormal products in batches, and ensures productivity and accuracy. The following technical solutions are specifically implemented.

In a first aspect, the present invention provides a mobile phone camera detection method based on an image sensor pixel array, comprising the following steps:

Obtain the characteristics of the target camera area or grayscale step in the image before and after the active illumination of the external infrared light source to achieve target detection and obtain the input image, in which the EL-NIR image and the NL-NIR image are used as input;

The input image is preprocessed to improve the image signal-to-noise ratio, and the target segmentation algorithm using histogram and image entropy is used to obtain suspicious target seed points;

Obtain the complete highlight area, and use the feature descriptor of the target camera to perform target discrimination to screen suspicious targets. The screening process includes performing image difference between the EL-NIR image and the NL-NIR image to obtain a differential image to achieve preliminary background suppression. Use morphological filtering to improve the saliency of the real target in the image, obtain target highlight area fragments through target segmentation, and use adaptive area growth to obtain the target complete highlight area;

The detection clustering algorithm is used to determine the repeated detection of the target, and the detection results are displayed in RGB color images.

As a further improvement of the above technical solution, the screening process includes performing image difference between the EL-NIR image and the NL-NIR image to obtain a differential image to achieve preliminary background suppression, including:

The similar parts of the images are weakened by correspondingly subtracting the pixel values of the two images, and the changed parts of the images are highlighted. The acquisition method of the difference image includes the difference between the current image and the fixed background, and the difference between two consecutive images;

The expression obtained after image difference is _Idif ( _x ,y)= _Ip ( _x ,y)-In(x,y), where _ip and In are EL-NIR image and NL-NIR image respectively , I _dif is the obtained differential image, (x, y) is the pixel coordinate of the image alignment for differential;

其中I_dst为经背景抑制算法处理后的背景抑制图像，I_dif为差分图像，

为膨胀运算，

为腐蚀运算，M_d为算法中的膨胀结构元素，M_e为算法中的腐蚀结构元素。The small-sized circular light spot formed by the target camera under the irradiation of infrared light can improve the signal-to-noise ratio of the image by analyzing the target shape characteristics and grayscale characteristics, and its expression is as follows:

where I _dst is the background suppression image processed by the background suppression algorithm, and I _dif is the difference image,

for the expansion operation,

is the erosion operation, M _d is the expansion structure element in the algorithm, and _Me is the erosion structure element in the algorithm.

As a further improvement of the above technical solution, the erosion eliminates relatively isolated pixels in the edge of the image, and erodes the outline of the particles according to the template defined by the structuring element, so that the area of pixels with high gray values is reduced. For any given pixel p ₀ , the structuring element is centered on p ₀ , the element masked by the structuring element is equal to 1, and denoted as p _i , then the value of pixel p _i is equal to 0, and p ₀ is set to 0, if p _i is 1, p i ₀ is set to 1;

Dilation eliminates the tiny holes between the image summary pixels and expands the outline of the pixels according to the template defined by the structuring elements. Dilation increases the brightness of the pixels around the pixels with high brightness, and makes the area of pixels with high gray values. big.

As a further improvement of the above technical solution, the characteristics of the target camera area or grayscale steps in the images before and after the active illumination of the external infrared light source are obtained to realize the target detection and obtain the input image, including:

Divide the image into multiple regions and process them to detect the noise in them. Different regions adapt to the size of the module according to their actual conditions, and process the checked modules to remove the noise;

Establish a reference coordinate system based on the standard image, and create one or more ROI search regions of interest, the ROI contains the stable features and detection content of the product image;

The coordinate system of the image to be detected is determined by the algorithm positioning function, edge detection or template matching based on the reference system, and the position and orientation of the object in the image are tracked based on the coordinate system.

As a further improvement of the above technical solution, a reference coordinate system is established according to the standard image, and one or more ROI search regions of interest are created, including:

Obtain the area to be processed and draw the ROI, multiply the feature to be processed by the area of interest or a preset mask, so that the gray value of the pixel in the area to be processed remains unchanged, and the gray value of other areas is zero;

Remove the interference to shield the interference features through the mask, so that they do not participate in the operation;

Obtain target features, and use similar shape or template matching algorithms to detect and obtain morphological features similar to masks in the image to be tested.

As a further improvement of the above technical solution, the target segmentation algorithm using histogram and image entropy is used to obtain suspicious target seed points, including:

其中H(x)越大，表示x的不确定性越大，将其累加，则代表整个系统的总体熵，表征信息源的总信息量，x为一个随机离散变量，满足x∈{x₁,x₂,x₃...}，其概率分布的表达式为p(X＝x_i)＝p_i，i＝1,2,3...n，其中p_i表示图像中的每一个灰度级所出现的概率值；The histogram threshold segmentation algorithm is used to extract a fixed proportion of high-gray areas to extract the real and target areas. Combined with the global two-dimensional information entropy of the image, the Shannon entropy is used to measure the uncertainty of random variables, and its expression is

The larger H(x) is, the greater the uncertainty of x is, and it is accumulated to represent the overall entropy of the entire system, representing the total amount of information of the information source, x is a random discrete variable, satisfying x∈{x ₁ ,x ₂ ,x ₃ ... }, the expression of its probability distribution is p(X=x _i )=pi , _i =1,2,3...n, where _pi represents each The probability value of gray level occurrence;

P₀(q)、P₁(q)分别表示I_dst经阈值q(0≤q≤k-1)分割后背景与前景的累计概率，两者志合为1，其中估算I₀、I₁的概率密度函数表示为I₀:

I₁:

其中

计算得到I₀、I₁的熵H₀(q)、H₁(q)后，得到使得I_dst经阈值分割后的熵H(q)最大的q，H(q)＝H₀(q)+H₁(q)。For the image I _dst to be processed, the entropy H ₀ (q) and H ₁ (q) of the background and foreground I ₀ and I ₁ obtained by dividing the threshold q are calculated respectively, and their expressions are respectively:

P ₀ (q) and P ₁ (q) respectively represent the cumulative probability of background and foreground after I _dst is segmented by the threshold q (0≤q≤k-1), and the two are equal to 1, where I ₀ and I ₁ are estimated. The probability density function of is expressed as I ₀ :

I ₁ :

in

After calculating the entropy H ₀ (q) and H ₁ (q) of I ₀ and I ₁ , obtain the q that maximizes the entropy H(q) after I _dst is divided by the threshold, H(q)=H ₀ (q) +H ₁ (q).

As a further improvement of the above technical solution, according to the gray value from high to low, select the gray value corresponding to the number of pixels in a fixed proportion from the image histogram to obtain the target segmentation threshold th based on the histogram, th=[p( gt>th)≤10%], where _gt is the gray value of each pixel in the image, _t represents each pixel in the image, and p() represents the probability value of the pixel satisfying the condition in the image I _dst .

As a further improvement of the above technical solution, the complete highlight area is obtained, and the target discrimination is performed through the feature descriptor of the target camera to screen for the discrimination of suspicious targets, including:

From the seed point, the adjacent pixels with the same attributes as the seed point are merged into the same area to realize area amplification;

The size of the growth area and the contrast between the central grayscale and the peripheral grayscale are respectively defined, and any point in R _i (R _i ∈ φ, i=1, 2, 3...t) is taken as the seed point, and the seed point is similar to Neighboring pixels are discriminated to make the stop region grow.

As a further improvement of the above technical solution, obtaining a complete highlight area includes:

Use the threshold of the column gray value to convert the corresponding gray value into a logical value of 0 or 1;

Down-sampling the obtained logic value into bit information according to the data bit resolution;

Decode the detected data frame header to obtain the carried data information to demodulate the signal.

In a second aspect, the present invention also provides a mobile phone camera detection device based on an image sensor pixel array, including:

The acquisition module is used to acquire the characteristics of the target camera area or grayscale steps in the images before and after the active irradiation of the external infrared light source to realize the target detection and obtain the input image, wherein the EL-NIR image and the NL-NIR image are used as input;

The segmentation module is used to preprocess the input image to improve the signal-to-noise ratio of the image, and use the target segmentation algorithm of histogram and image entropy to obtain suspicious target seed points;

The screening module is used to obtain the complete highlight area, and perform target discrimination through the feature descriptor of the target camera to screen suspicious targets. The screening process includes performing image difference between the EL-NIR image and the NL-NIR image to obtain a differential image. Achieve preliminary background suppression, use morphological filtering to improve the saliency of the real target in the image, obtain target highlight area fragments through target segmentation, and use adaptive area growth to obtain the target complete highlight area;

The detection module is used to determine the repeated detection of the target through the detection clustering algorithm, and display the detection result as an RGB color image.

The present invention provides a mobile phone camera detection method and device based on an image sensor pixel array, which has the following beneficial effects compared to the prior art:

By obtaining the characteristics of the target camera area or grayscale steps in the image before and after the active illumination of the external infrared light source, the target detection is realized and the input image is obtained, the input image is preprocessed to improve the image signal-to-noise ratio, and the target segmentation using histogram and image entropy is used. The algorithm obtains the seed point of the suspicious target, obtains the complete highlight area, and uses the feature descriptor of the target camera to perform target discrimination to screen the suspicious target. exhibit. The target can be accurately detected in a smooth background, the real target can be detected in a complex background, and there are fewer false alarms, the discrimination conditions are simple, and the detection performance is guaranteed with good running performance, changing the shape of the target mask It will reduce the accuracy of later feature analysis and calculation, and affect the final target determination effect. Using the fact that the re-inspection targets must be close in physical distance, the discrimination results are clustered to eliminate the situation of repeated target detection, and to a certain extent, the accuracy and efficiency of camera detection are also improved.

Description of drawings

In order to illustrate the technical solutions of the embodiments of the present invention more clearly, the following briefly introduces the accompanying drawings used in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, and therefore do not It should be regarded as a limitation of the scope, and for those of ordinary skill in the art, other related drawings can also be obtained according to these drawings without any creative effort.

1 is a flowchart of a mobile phone camera detection method based on an image sensor pixel array of the present invention;

Fig. 2 is the flow chart of the noise detection of the present invention;

Fig. 3 is the flowchart of the noise removal of the present invention;

Fig. 4 is the flow chart of the acquisition complete highlight area of the present invention;

FIG. 5 is a structural block diagram of a mobile phone camera detection device based on an image sensor pixel array of the present invention.

Detailed ways

The following describes in detail the embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals refer to the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary, only used to explain the present invention, and should not be construed as a limitation of the present invention.

Referring to FIG. 1, the present invention provides a mobile phone camera detection method based on an image sensor pixel array, including the following steps:

S10: Obtain the characteristics of the target camera area or grayscale steps in the images before and after the active illumination of the external infrared light source to achieve target detection and obtain an input image, in which the EL-NIR image and the NL-NIR image are used as input;

S11: Preprocess the input image to improve the signal-to-noise ratio of the image, and use the target segmentation algorithm of histogram and image entropy to obtain suspicious target seed points;

S12: Obtain the complete highlight area, and perform target discrimination through the feature descriptor of the target camera to screen the suspicious target, wherein the screening process includes performing image difference between the EL-NIR image and the NL-NIR image to obtain a difference image to achieve a preliminary background Suppression, use morphological filtering to improve the saliency of the real target in the image, obtain target highlight area fragments through target segmentation, and use adaptive area growth to obtain the target complete highlight area;

S13: Determine the repeated detection situation of the target through the detection clustering algorithm, and display the detection result as an RGB color image.

其中I_dst为经背景抑制算法处理后的背景抑制图像，I_dif为差分图像，

为膨胀运算，

为腐蚀运算，M_d为算法中的膨胀结构元素，M_e为算法中的腐蚀结构元素。In this embodiment, the screening process includes performing image difference between the EL-NIR image and the NL-NIR image to obtain a difference image to achieve preliminary background suppression, and by correspondingly subtracting the pixel values of the two images to weaken the similar part of the image, and highlight the image For the changed part, the acquisition method of the difference image includes the difference between the current image and the fixed background, and the difference between two consecutive images; the expression obtained after the image difference is I _dif (x, y)=I _p (x, y) -In ( _x , y), wherein _{i p} and In are the EL-NIR image and the NL-NIR image respectively, I _dif is the difference image obtained, (x, y) is the pixel coordinate of the image alignment for difference; The small-sized circular light spot formed by the target camera under the irradiation of infrared light can improve the signal-to-noise ratio of the image by analyzing the target shape characteristics and grayscale characteristics, and its expression is as follows:

where I _dst is the background suppression image processed by the background suppression algorithm, and I _dif is the difference image,

for the expansion operation,

is the erosion operation, M _d is the expansion structure element in the algorithm, and _Me is the erosion structure element in the algorithm.

It should be noted that the erosion eliminates relatively isolated pixels in the edge of the image, and erodes the outline of the particles according to the template defined by the structuring element, so that the area of pixels with high gray value is reduced. For any given pixel p ₀ , the structure The element is centered on p ₀ , the element masked by the structuring element is equal to 1, and denoted as p _i , then the value of pixel p _i is equal to 0, set p ₀ to 0, and if p _i is 1 set p ₀ to 1; Expansion eliminates the tiny holes between the image summary pixels, expands the outline of the pixel points according to the template defined by the structural element, and the expansion increases the brightness of the pixels around the pixels with high brightness, so that the pixels with high gray value area becomes larger.

It should be understood that the saliency of the real object in the image is further improved after most of the background is eliminated by the background suppression algorithm. In order to detect possible target cameras in the image, it is necessary to extract candidate target regions through target segmentation. In order to extract the real target and avoid the target over-segmented image caused by the excessive value of the histogram segmentation algorithm, the smooth threshold is achieved by calculating the geometric weighted average. The actual results obtained using the histogram-based thresholding style separated the real target, but caused the phenomenon of over-segmentation and brought in a lot of noise. The over-segmentation phenomenon not only introduced a lot of meaningless workload for subsequent target discrimination mitigation, It will also reduce the overall detection effect of the algorithm. In a smooth background, the adaptive region growing algorithm can better expand the highlight area of the target camera based on the target segmentation results. In a complex background, the adaptive region growing algorithm can not only optimize the target segmentation results of the real target, but also have the The ability to extract false targets such as lines and blocks helps to filter out false alarms.

Referring to FIG. 2, optionally, acquiring the characteristics of the target camera area or grayscale steps in the images before and after the active illumination of the external infrared light source to achieve target detection and obtain the input image, including:

S20: Divide the image into multiple regions and process them to detect noise in them. Different regions adapt to the size of the modules according to their actual conditions, and process the checked modules to remove the noise;

S21: Establish a reference coordinate system according to the standard image, and create one or more ROI search regions of interest, where the ROI contains the stable features and detection content of the product image;

S22: Determine the coordinate system of the image to be detected by using the algorithm positioning function, edge detection or template matching based on the reference system, and track the position and direction of the object in the image based on the coordinate system.

In this embodiment, in most cases, the position of the object to be measured in the field of view of the camera cannot be completely fixed, and when the product position shifts, the ROI search area also moves relative to the coordinate system. The glue in the four corners is used for detection, that is, the ROI area is the four corners of the camera. The image mask is a binary image. The size of the mask itself is required to be less than or equal to the image to be detected. The image mask module will detect the corresponding module size in the processed image. If a pixel is in the image mask If there is a non-zero value in the image, the corresponding pixel will be processed in the process of detecting the image. On the contrary, if the gray value of a pixel in the image mask is zero, the corresponding pixel will be ignored in the image processing. , the UV glue part of the four corners of the object to be tested is extracted separately, which provides efficiency and stability for subsequent algorithm processing and analysis.

Referring to Figure 3, optionally, a reference coordinate system is established according to the standard image, and one or more ROI search regions of interest are created, including:

S30: Obtain the region to be processed and draw a ROI, multiply the feature to be processed by the region of interest or a preset mask, so that the gray value of the pixel in the region to be processed remains unchanged, and the gray value of other regions is zero;

S31: remove the interference and shield the interference feature through the mask, so that it does not participate in the operation;

S32: Obtain target features, and use a similar shape or template matching algorithm to detect and obtain morphological features similar to the mask in the image to be tested.

其中H(x)越大，表示x的不确定性越大，将其累加，则代表整个系统的总体熵，表征信息源的总信息量，x为一个随机离散变量，满足x∈{x₁,x₂,x₃...}，其概率分布的表达式为p(X＝x_i)＝p_i，i＝1,2,3...n，其中p_i表示图像中的每一个灰度级所出现的概率值；对于待处理图像I_dst，分别计算阈值q所分割得到的背景与前景I₀、I₁的熵H₀(q)、H₁(q)，其表达式分别为：

P₀(q)、P₁(q)分别表示I_dst经阈值q(0≤q≤k-1)分割后背景与前景的累计概率，两者志合为1，其中估算I₀、I₁的概率密度函数表示为I₀:

I₁:

其中

计算得到I₀、I₁的熵H₀(q)、H₁(q)后，得到使得I_dst经阈值分割后的熵H(q)最大的q，H(q)＝H₀(q)+H₁(q)。In this embodiment, the target segmentation algorithm of histogram and image entropy is used to obtain suspicious target seed points, and the histogram threshold segmentation algorithm is used to extract a fixed-proportion high-gray area to extract the real, target area, combined with the global two-dimensional information entropy of the image , the Shannon entropy is used to measure the uncertainty of random variables, and its expression is

The larger H(x) is, the greater the uncertainty of x is, and it is accumulated to represent the overall entropy of the entire system, representing the total amount of information of the information source, x is a random discrete variable, satisfying x∈{x ₁ ,x ₂ ,x ₃ ... }, the expression of its probability distribution is p(X=x _i )=pi , _i =1,2,3...n, where _pi represents each The probability value of gray level occurrence; for the image I _dst to be processed, the entropy H ₀ (q) and H ₁ (q) of the background and foreground I ₀ and I ₁ obtained by dividing the threshold q are calculated respectively, and their expressions are respectively for:

P ₀ (q) and P ₁ (q) respectively represent the cumulative probability of background and foreground after I _dst is segmented by the threshold q (0≤q≤k-1), and the two are equal to 1, where I ₀ and I ₁ are estimated. The probability density function of is expressed as I ₀ :

I ₁ :

in

After calculating the entropy H ₀ (q) and H ₁ (q) of I ₀ and I ₁ , obtain the q that maximizes the entropy H(q) after I _dst is divided by the threshold, H(q)=H ₀ (q) +H ₁ (q).

It should be noted that, according to the gray value from high to low, select the gray value corresponding to the number of pixels in a fixed proportion from the image histogram to obtain the target segmentation threshold th based on the histogram, th=[p(g _t > th)≤10%], where g _t is the gray value of each pixel in the image, t represents each pixel in the image, and p( ) represents the probability value of the pixel that satisfies the condition in the image I _dst .

Referring to Figure 4, optionally, obtaining the complete highlight area includes:

S40: using the threshold value of the column gray value to convert the corresponding gray value into a logical value of 0 or 1;

S41: down-sampling the obtained logic value into bit information according to the data bit resolution;

S42: Decode the detected data frame header to obtain the carried data information to demodulate the signal.

In this embodiment, the complete highlight area is obtained, and the target discrimination is performed by the feature descriptor of the target camera to screen the suspicious target. Realize area amplification; the size of the growing area and the contrast between the central grayscale and the peripheral grayscale are respectively defined, and any point in R _i (R _i ∈ φ, i=1, 2, 3...t) is taken as the seed Point, the adjacent pixels of the seed point are discriminated, so that the stop area grows.

Referring to FIG. 5, the present invention also provides a mobile phone camera detection device based on an image sensor pixel array, including:

The acquisition module is used to acquire the characteristics of the target camera area or grayscale steps in the images before and after the active irradiation of the external infrared light source to realize the target detection and obtain the input image, wherein the EL-NIR image and the NL-NIR image are used as input;

The segmentation module is used to preprocess the input image to improve the signal-to-noise ratio of the image, and use the target segmentation algorithm of histogram and image entropy to obtain suspicious target seed points;

The screening module is used to obtain the complete highlight area, and perform target discrimination through the feature descriptor of the target camera to screen suspicious targets. The screening process includes performing image difference between the EL-NIR image and the NL-NIR image to obtain a differential image. Achieve preliminary background suppression, use morphological filtering to improve the saliency of the real target in the image, obtain target highlight area fragments through target segmentation, and use adaptive area growth to obtain the target complete highlight area;

The detection module is used to determine the repeated detection of the target through the detection clustering algorithm, and display the detection result as an RGB color image.

In this embodiment, since the target segmentation algorithm splits the connected domain of the same target and makes it recalled repeatedly, repeated detection of the same target will cause the system to generate unnecessary trigger alarms, resulting in repeated detection and waste of manpower. In the suspicious target extraction stage, the morphological opening and closing operation is used to connect the split regions, and then the target is determined. After the target is determined, the clustering algorithm is used to treat the detection results with similar clusters as one target. Changing the shape of the target mask will reduce the accuracy of later feature analysis and calculation, and affect the final target determination effect. Using the fact that the re-examination targets must be close in physical distance, the discriminant results are clustered to eliminate the situation of repeated detection of targets.

In all examples shown and described herein, any specific value should be construed as merely exemplary and not as limiting, as other examples of exemplary embodiments may have different values.

It should be noted that like numerals and letters refer to like items in the following figures, so once an item is defined in one figure, it does not require further definition and explanation in subsequent figures.

The above-mentioned embodiments only represent several embodiments of the present invention, and the descriptions thereof are specific and detailed, but should not be construed as limiting the scope of the present invention. It should be pointed out that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can also be made, which all belong to the protection scope of the present invention.

Claims (10)

1. a mobile phone camera detection method based on an image sensor pixel array, is characterized in that, comprises the following steps: Obtain the characteristics of the target camera area or grayscale step in the image before and after the active illumination of the external infrared light source to achieve target detection and obtain the input image, in which the EL-NIR image and the NL-NIR image are used as input; The input image is preprocessed to improve the image signal-to-noise ratio, and the target segmentation algorithm using histogram and image entropy is used to obtain suspicious target seed points; Obtain the complete highlight area, and use the feature descriptor of the target camera to perform target discrimination to screen suspicious targets. The screening process includes performing image difference between the EL-NIR image and the NL-NIR image to obtain a differential image to achieve preliminary background suppression. Use morphological filtering to improve the saliency of the real target in the image, obtain target highlight area fragments through target segmentation, and use adaptive area growth to obtain the target complete highlight area; The detection clustering algorithm is used to determine the repeated detection of the target, and the detection results are displayed in RGB color images. 2. the mobile phone camera detection method based on image sensor pixel array according to claim 1, is characterized in that, screening process comprises that EL-NIR image and NL-NIR image are carried out image difference to obtain difference image and realize preliminary background suppression, comprising: The similar parts of the images are weakened by correspondingly subtracting the pixel values of the two images, and the changed parts of the images are highlighted. The acquisition method of the difference image includes the difference between the current image and the fixed background, and the difference between two consecutive images; The expression obtained after image difference is _Idif ( _x ,y)= _Ip ( _x ,y)-In(x,y), where _ip and In are EL-NIR image and NL-NIR image respectively , I _dif is the obtained differential image, (x, y) is the pixel coordinate of the image alignment for differential; The small-sized circular light spot formed by the target camera under the irradiation of infrared light can improve the signal-to-noise ratio of the image by analyzing the target shape characteristics and grayscale characteristics, and its expression is as follows:

where I _dst is the background suppression image processed by the background suppression algorithm, and I _dif is the difference image,

for the expansion operation,

is the erosion operation, M _d is the expansion structure element in the algorithm, and _Me is the erosion structure element in the algorithm. 3. The method for detecting a mobile phone camera based on an image sensor pixel array according to claim 2, wherein the corrosion is performed by eliminating relatively isolated pixels in the edge of the image, and according to the template defined by the structuring element to corrode the outline of the particles, so that the gray The area of a pixel with a high degree value decreases. For any given pixel p ₀ , the structuring element is centered at p ₀ , the element masked by the structuring element is equal to 1, and denoted by p _i , then the value of pixel p _i is equal to 0, set p ₀ to 0, if p _i is 1, set p ₀ to 1; Dilation eliminates the tiny holes between the image summary pixels and expands the outline of the pixels according to the template defined by the structuring elements. Dilation increases the brightness of the pixels around the pixels with high brightness, and makes the area of pixels with high gray values. big. 4. The mobile phone camera detection method based on an image sensor pixel array according to claim 1, characterized in that, acquiring the characteristics of the target camera area or grayscale steps in the images before and after the active irradiation of the external infrared light source to achieve target detection and obtain the input image ,include: Divide the image into multiple regions and process them to detect the noise in them. Different regions adapt to the size of the module according to their actual conditions, and process the checked modules to remove the noise; Establish a reference coordinate system based on the standard image, and create one or more ROI search regions of interest, the ROI contains the stable features and detection content of the product image; The coordinate system of the image to be detected is determined by the algorithm positioning function, edge detection or template matching based on the reference system, and the position and orientation of the object in the image are tracked based on the coordinate system. 5. the mobile phone camera detection method based on image sensor pixel array according to claim 4, is characterized in that, establishes a reference coordinate system according to standard image, and creates one or more ROI interest search area, comprising: Obtain the area to be processed and draw the ROI, multiply the feature to be processed by the area of interest or a preset mask, so that the gray value of the pixel in the area to be processed remains unchanged, and the gray value of other areas is zero; Remove the interference to shield the interference features through the mask, so that they do not participate in the operation; Obtain target features, and use similar shape or template matching algorithms to detect and obtain morphological features similar to masks in the image to be tested. 6. the mobile phone camera detection method based on image sensor pixel array according to claim 1, is characterized in that, uses the target segmentation algorithm of histogram and image entropy to obtain suspicious target seed point, comprising: The histogram threshold segmentation algorithm is used to extract a fixed proportion of high-gray areas to extract the real and target areas. Combined with the global two-dimensional information entropy of the image, the Shannon entropy is used to measure the uncertainty of random variables, and its expression is

The larger H(x) is, the greater the uncertainty of x is, and it is accumulated to represent the overall entropy of the entire system, representing the total amount of information of the information source, x is a random discrete variable, satisfying x∈{x ₁ ,x ₂ ,x ₃ ... }, the expression of its probability distribution is p(X=x _i )=pi , _i =1,2,3...n, where _pi represents each The probability value of gray level occurrence; For the image I _dst to be processed, the entropy H ₀ (q) and H ₁ (q) of the background and foreground I ₀ and I ₁ obtained by dividing the threshold q are calculated respectively, and their expressions are respectively:

P ₀ (q) and P ₁ (q) respectively represent the cumulative probability of background and foreground after I _dst is segmented by the threshold q (0≤q≤k-1), and the two are equal to 1, where I ₀ and I ₁ are estimated. The probability density function of is expressed as

in

After calculating the entropy H ₀ (q) and H ₁ (q) of I ₀ and I ₁ , obtain the q that maximizes the entropy H(q) after I _dst is divided by the threshold, H(q)=H ₀ (q) +H ₁ (q). 7. The method for detecting a mobile phone camera based on an image sensor pixel array according to claim 6, further comprising: According to the gray value from high to low, select the gray value corresponding to the number of pixels in a fixed proportion from the image histogram to obtain the target segmentation threshold th based on the histogram, th=[p(g _t >th)≤10% ], where gt is the gray value of each pixel in the image, _t represents each pixel in the image, and p() represents the probability value of the pixel that satisfies the condition in the image I _dst . 8. the mobile phone camera detection method based on image sensor pixel array according to claim 1, is characterized in that, obtain complete highlight area, and carry out target discrimination by the feature descriptor of target camera and screen the discrimination of suspicious target, comprising: From the seed point, the adjacent pixels with the same attributes as the seed point are merged into the same area to realize area amplification; The size of the growth area and the contrast between the central grayscale and the peripheral grayscale are respectively defined, and any point in R _i (R _i ∈ φ, i=1, 2, 3...t) is taken as the seed point, and the seed point is similar to Neighboring pixels are discriminated to make the stop region grow. 9. The mobile phone camera detection method based on an image sensor pixel array according to claim 8, wherein obtaining a complete highlight area comprises: Use the threshold of the column gray value to convert the corresponding gray value into a logical value of 0 or 1; Down-sampling the obtained logic value into bit information according to the data bit resolution; Decode the detected data frame header to obtain the carried data information to demodulate the signal. 10. A mobile phone camera detection device based on an image sensor pixel array according to the mobile phone camera detection method based on an image sensor pixel array according to any one of claims 1-9, characterized in that, comprising: The acquisition module is used to acquire the characteristics of the target camera area or grayscale steps in the images before and after the active irradiation of the external infrared light source to realize the target detection and obtain the input image, wherein the EL-NIR image and the NL-NIR image are used as input; The segmentation module is used to preprocess the input image to improve the signal-to-noise ratio of the image, and use the target segmentation algorithm of histogram and image entropy to obtain suspicious target seed points; The screening module is used to obtain the complete highlight area, and perform target discrimination through the feature descriptor of the target camera to screen suspicious targets. The screening process includes performing image difference between the EL-NIR image and the NL-NIR image to obtain a differential image. Achieve preliminary background suppression, use morphological filtering to improve the saliency of the real target in the image, obtain target highlight area fragments through target segmentation, and use adaptive area growth to obtain the target complete highlight area; The detection module is used to determine the repeated detection of the target through the detection clustering algorithm, and display the detection result as an RGB color image.

Images