A Mini LED backlight function testing method and device

Abstract

The present invention proposes a Mini LED backlight function testing method and device, including: responding to acquiring a set of images of Mini LED backlight panels with different luminous areas; detecting bright and dark areas based on the Hough circle line method and batch-producing mask images; obtaining a U-Net model; using images in a dataset to perform multiple rounds of iterative training on the obtained U-Net model; segmenting the processed image to identify the luminous area; performing dilated convolution and eroded convolution operations on the processed image to identify the luminous and dark areas; and calculating the area of ​​dark pixels to identify the number of normal and abnormal lamp beads. The present invention utilizes a computer vision supervised learning algorithm to solve the problem of rapid detection of Mini LED backlight panels. Using a U-shaped network to train a large amount of simulated light and dark data, the trained model has a good ability to identify the areas of normal and abnormal lamp beads. The model can replace manual visual inspection, improving work efficiency and the degree of control over product quality.

Description

Technical Field

[0001] The present invention belongs to the field of industrial detection technology, and specifically relates to a Mini LED backlight function testing method and device. Background Art

[0002] In the field of modern display technology, MiniLED backlighting is gaining increasing attention. MiniLED backlighting utilizes matrix-based zone-based light control, allowing the brightness of each MiniLED zone to be controlled freely. This allows for improved color contrast by turning off or reducing the brightness of corresponding MiniLED zones when displaying dark images. However, because MiniLEDs far outnumber traditional edge-lit backlights, traditional detection methods are inefficient and inaccurate. Therefore, new detection methods are needed to determine whether each MiniLED is functioning properly.

[0003] In recent years, the rapid development of MiniLED technology has brought enormous business opportunities to the display industry. As technology flourishes, product quality inspection and control solutions must also keep pace. Currently, detecting abnormalities in MiniLED backlight panels primarily relies on visual inspection. However, due to the large number of MiniLED lamps, visual inspection is subject to significant subjectivity and false positives. Therefore, it is necessary to introduce image processing and supervised learning methods to help humans quickly and efficiently detect the areas and number of abnormally functioning MiniLED lamps.

[0004] In view of this, it is very meaningful to propose a Mini LED backlight function testing method and device. Summary of the Invention

[0005] In order to solve the problem of rapid detection of existing MiniLED backlight panels, the present invention provides a MiniLED backlight function testing method and device to solve the above-mentioned technical defects.

[0006] In a first aspect, the present invention provides a Mini LED backlight function testing method, which includes the following steps:

[0007] In response to acquiring a set of images of MiniLED backlight panels having different light-emitting areas;

[0008] Perform threshold processing and Sobel operator processing on the acquired image;

[0009] Detect bright and dark areas based on the Hough circle line method and generate mask images in batches;

[0010] The acquired original image and the corresponding mask image are formed into a data set, which is divided into a training set and a validation set;

[0011] Based on the PyTorch deep learning framework, a U-shaped encoder-decoder network structure was built to obtain the U-Net model.

[0012] Use the images in the dataset to obtain the U-Net model for multiple rounds of iterative training, and record the model evaluation indicators IOU and Dice coefficient at the end of each epoch;

[0013] The trained U-Net model is further used to segment the image to be processed to identify the luminous area;

[0014] Perform dilation convolution and erosion convolution operations on the processed image to find the luminous and dark areas;

[0015] By calculating the area of ​​dark pixels, the number of normal and abnormal lamp beads can be identified.

[0016] Preferably, the method further includes: lighting the MiniLED backlight panel, and using a shading plate to simulate the LED chip backlight panel images under different brightness conditions to obtain a group of images of the MiniLED backlight panel with different light-emitting areas.

[0017] Further preferably, performing threshold processing on the acquired image includes: converting the acquired image into a grayscale image, and then performing T processing using a global threshold method, setting the threshold to T, where I' represents the processed image, and the formula can be expressed as:

[0018] I'(x,y)={255,if I(x,y)>T;0,otherwise}.

[0019] Further preferably, performing Sobel operator processing on the acquired image includes: applying the Sobel operator to the binarized image to obtain the horizontal gradient Gx and the vertical gradient Gy of the image. The calculation formula of the Sobel operator is as follows:

[0020] Gx=[[-1,0,1],[-2,0,2],[-1,0,1]]*I'(x,y)

[0021] Gy=[[-1,-2,-1],[0,0,0],[1,2,1]]*I'(x,y).

[0022] Further preferably, the gradient magnitude image is processed using a non-maximum suppression algorithm to obtain a sparse and clear edge image, and the gradient magnitude image I_magnitude and the gradient direction image I_direction are traversed. For each pixel point (x, y), its direction θ in the gradient direction image is first determined, which can be expressed as: θ(x, y) = arctan(I_direction(x, y)).

[0023] Further preferably, the method further includes: for each interval, checking neighboring pixel points of the current pixel point in the direction, such as checking the pixel points on the left and right in the 0° direction;

[0024] If the gradient magnitude of the current pixel is greater than the gradient magnitudes of its two neighbors, the pixel is retained; otherwise, its gradient magnitude is set to 0.

[0025] Further preferably, the U-Net model specifically includes:

[0026] The encoder consists of multiple convolutional layers Conv2D, batch normalization BatchNorm2D, and maximum pooling layers MaxPooling2D. The number of output channels of the convolutional layer increases layer by layer. The formula of the convolution operation is: F(x) = W*x+b, where F(x) represents the output feature map, W represents the convolution kernel weight, x represents the input feature map, and b represents the bias term.

[0027] The decoder consists of multiple upsampling layers, convolutional layers, and batch normalization. The upsampling layers amplify the feature maps through interpolation and then concatenate them with the feature maps from the corresponding encoder layers. The number of output channels in the convolutional layers decreases with each layer; in this network, the number is 512, 256, 128, and 64.

[0028] Further preferably, the method further comprises: using an Adam optimizer during the training of the U-Net model, setting a learning rate of 0.001, a training batch of 50, and a training step size of up to 4500;

[0029] At the same time, the cross entropy loss function is used to minimize the difference between the model prediction and the true segmentation mask; the formula of the cross entropy loss function is:

[0030] L(y,t)=-∑t_i*log(y_i)

[0031] Among them, L(y,t) represents the loss value, y is the segmentation image predicted by the model, t is the true segmentation mask, and i represents the pixel index.

[0032] Further preferably, the calculation formulas for the model evaluation indicators IOU and Dice coefficient are:

[0033] IOU=|A∩B| / |A∪B|

[0034] Dice=2*|A∩B| / (|A|+|B|)

[0035] Among them, A represents the predicted segmentation result, B represents the final detection result of the true segmentation, the Dice coefficient result tends to be stable with the training, and the optimal value is 92.3%, and the IOU coefficient result tends to be stable with the training, and the optimal value is 91.5%.

[0036] In a second aspect, an embodiment of the present invention further provides a Mini LED backlight function testing device, the device comprising:

[0037] an acquisition module, configured to acquire, in response, a set of images of MiniLED backlight panels having different light-emitting areas;

[0038] The threshold processing module is used to convert the acquired image into a grayscale image and then perform T processing using the global threshold method;

[0039] The Sobel operator processing module is used to apply the Sobel operator to the binarized image to obtain the horizontal gradient Gx and vertical gradient Gy of the image;

[0040] Hough circle and line method detection module, used to detect bright and dark areas based on the Hough circle and line method, and batch generate mask images;

[0041] A data set module is used to form a data set by combining the acquired original image and the corresponding mask image into a training set and a validation set;

[0042] The U-Net model module is used to build a U-shaped encoder-decoder network structure based on the PyTorch deep learning framework to obtain the U-Net model;

[0043] The training module is used to perform multiple rounds of iterative training on the U-Net model using the images in the dataset, and record the model evaluation indicators IOU and Dice coefficient at the end of each epoch;

[0044] The image segmentation module is used to segment the image to be processed using the trained U-Net model to identify the luminous area;

[0045] The dilation and erosion module is used to perform dilation and erosion convolution operations on the processed image to find the luminous and dark areas;

[0046] The recognition module is used to identify the number of normal lamp beads and abnormal lamp beads by calculating the area of ​​dark pixels.

[0047] In a third aspect, an embodiment of the present invention provides an electronic device comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by one or more processors, the one or more processors implement the method described in any implementation manner in the first aspect.

[0048] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any implementation manner in the first aspect.

[0049] Compared with the prior art, the present invention has the following advantages:

[0050] This invention uses computer vision supervised learning algorithms to solve the problem of rapid detection of MiniLED backlight panels. A U-shaped network is used to train a large amount of simulated light and dark data. The trained model has a good ability to identify the areas of normal lamp beads and abnormal lamp beads, which not only demonstrates the rationality of the preprocessing method, but also can expand the application of deep learning. Finally, combined with image processing, the number of normal chips and abnormal chips is detected. The model can replace manual visual inspection and improve work efficiency and product quality control. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The accompanying drawings are included to provide a further understanding of the embodiments and are incorporated into and constitute a part of this specification. The accompanying drawings illustrate the embodiments and, together with the description, serve to explain the principles of the present invention. Other embodiments and many of the expected advantages of the embodiments will be readily apparent as they become better understood by reference to the following detailed description. The elements of the drawings are not necessarily to scale with respect to each other. Like reference numerals designate corresponding similar parts.

[0052] Figure 1 is a diagram of an exemplary device architecture in which an embodiment of the present invention may be applied;

[0053] Figure 2 Schematic diagram of the process of a Mini LED backlight function test method according to an embodiment of the present invention;

[0054] Figure 3 1 is a schematic diagram of a method flow for testing the Mini LED backlight function according to an embodiment of the present invention;

[0055] Figure 4 Schematic diagram of a MiniLED backlight panel in a MiniLED backlight function testing method according to an embodiment of the present invention;

[0056] Figure 5 Schematic diagram of a threshold segmentation image processing method in a Mini LED backlight function testing method according to an embodiment of the present invention;

[0057] Figure 6 This is a schematic diagram of the Hough circle transform principle in the Mini LED backlight function test method according to an embodiment of the present invention;

[0058] Figure 7 This is a batch mask detection image of the light-emitting area in the Mini LED backlight function test method of an embodiment of the present invention;

[0059] Figure 8 This is a diagram of the U-shaped encoder-decoder network structure in the Mini LED backlight function test method according to an embodiment of the present invention;

[0060] Figure 9 This is a graph showing the IOU coefficient results in the Mini LED backlight function test method according to an embodiment of the present invention;

[0061] Figure 10 This is a diagram showing the Dice coefficient results in the Mini LED backlight function testing method according to an embodiment of the present invention;

[0062] Figure 11 This is the image after dilation and convolution in the Mini LED backlight function testing method according to an embodiment of the present invention;

[0063] Figure 12 This is the image after subtraction in the Mini LED backlight function testing method according to an embodiment of the present invention;

[0064] Figure 13 This is an image after corrosion treatment in the Mini LED backlight function testing method according to an embodiment of the present invention;

[0065] Figure 14 This is a diagram showing the final test results of the Mini LED backlight function test method according to an embodiment of the present invention;

[0066] Figure 15 Schematic diagram of a Mini LED backlight function test system according to an embodiment of the present invention;

[0067] Figure 16 It is a schematic structural diagram of a computer device suitable for implementing an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0068] In the following detailed description, reference is made to the accompanying drawings, which form a part of the detailed description and are illustrated by illustrative specific embodiments in which the present invention may be practiced. To this end, directional terms, such as "top," "bottom," "left," "right," "up," "down," etc., are used with reference to the orientation of the figures being described. Because the components of the embodiments may be positioned in several different orientations, directional terms are used for illustrative purposes and are in no way limiting. It should be understood that other embodiments may be utilized or logical changes may be made without departing from the scope of the present invention. Therefore, the following detailed description should not be adopted in a limiting sense, and the scope of the present invention is defined by the appended claims.

[0069] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is merely illustrative. Any number of terminal devices, networks and servers may be provided as required.

[0070] Figure 1 An exemplary system architecture 100 is shown to which a method for processing information or an apparatus for processing information according to an embodiment of the present invention can be applied.

[0071] like Figure 1 As shown, system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. Network 104 is a medium for providing communication links between terminal devices 101, 102, 103 and server 105. Network 104 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0072] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, and 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0073] The terminal devices 101 , 102 , and 103 may be various electronic devices with communication functions, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0074] The server 105 may be a server that provides various services, such as a background information processing server that processes verification request information sent by the terminal devices 101, 102, and 103. The background information processing server may analyze and perform other processing on the received verification request information and obtain a processing result (e.g., verification success information indicating that the verification request is a legitimate request).

[0075] It should be noted that the method for processing information provided in the embodiment of the present invention is generally executed by the server 105, and accordingly, the device for processing information is generally set in the server 105. In addition, the method for sending information provided in the embodiment of the present invention is generally executed by the terminal devices 101, 102, and 103, and accordingly, the device for sending information is generally set in the terminal devices 101, 102, and 103.

[0076] It should be noted that the server can be either hardware or software. When the server is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. When the server is software, it can be implemented as multiple software programs or software modules (for example, to provide distributed services), or as a single software program or multiple software modules, without specific limitation herein.

[0077] To address the rapid detection of MiniLED backlight panels, this paper proposes a detection method based on computer vision and deep learning that effectively identifies abnormal areas and counts abnormal LEDs. A light shield is used to block the mini LEDs in different areas, simulating the brightness and darkness of the LEDs on the panel surface. Based on this method, a mask image is generated. A U-shaped encoding and decoding network is used for deep learning training and evaluation of the effectiveness of identifying bright and dark areas of the LEDs. Finally, corrosion and dilation are used to perform the difference between the normal and abnormal LED counts.

[0078] Figure 2 The embodiment of the present invention discloses a method for testing the function of Mini LED backlight. Figure 2 As shown, the method includes the following steps:

[0079] S1, in response to acquiring a set of images of MiniLED backlight panels having different light-emitting areas;

[0080] S2, performing threshold processing and Sobel operator processing on the acquired image;

[0081] S3, detect bright and dark areas based on the Hough circle line method, and batch produce mask images;

[0082] S4, forming a data set from the acquired original image and the corresponding mask image into a training set and a validation set;

[0083] S5. Build a U-shaped encoder-decoder network structure based on the PyTorch deep learning framework to obtain the U-Net model;

[0084] S6. Use the images in the dataset to perform multiple rounds of iterative training on the obtained U-Net model, and record the model evaluation indicators IOU and Dice coefficient at the end of each epoch;

[0085] S7, further using the trained U-Net model to segment the image to be processed to identify the luminous area;

[0086] S8, performing dilation convolution and erosion convolution operations on the image to be processed to find the luminous and dark areas;

[0087] S9. By calculating the area of ​​dark pixels, the number of normal lamp beads and abnormal lamp beads can be identified.

[0088] Specifically, such as Figure 3 As shown, to address the aforementioned technical issues in the prior art, the present invention first creates a mask image and then uses a U-shaped encoder-decoder network to perform image segmentation to identify luminous areas. By subtracting the image after dilated convolution from the image after threshold segmentation, we obtain a binary image with relatively clear outlines. Next, an eroded convolution operation is performed on this image to connect closely spaced black dots in the image, thereby identifying the number of luminous chips and the number of abnormal chips.

[0089] First, light up the MiniLED backlight panel and use a shading plate to simulate the LED chip backlight panel images under different brightness conditions, so as to obtain a set of Mini backlight panel images with different light-emitting areas, such as Figure 4 shown.

[0090] These images are converted into grayscale images, and then the global threshold method is used for T processing, and the threshold is set to T, where I' represents the processed image, such as Figure 5 As shown, the formula can be expressed as:

[0091] I'(x,y)={255,if I(x,y)>T;0,otherwise}

[0092] Apply the Sobel operator to the binarized image to obtain the horizontal gradient Gx and vertical gradient Gy of the image. The calculation formula of the Sobel operator is as follows:

[0093] Gx=[[-1,0,1],[-2,0,2],[-1,0,1]]*I'(x,y)

[0094] Gy=[[-1,-2,-1],[0,0,0],[1,2,1]]*I'(x,y)

[0095] The gradient magnitude image is processed using the non-maximum suppression algorithm to produce a sparse and clear edge image. Non-maximum suppression (NMS) is an edge thinning technique whose goal is to find local maxima in an image and eliminate other non-maximum points. The gradient magnitude image I_magnitude and the gradient direction image I_direction are traversed. For each pixel (x, y), its direction θ in the gradient direction image is first determined, which can be expressed as:

[0096] θ(x,y)=arctan(I_direction(x,y))

[0097] For each interval, check the neighboring pixels of the current pixel in that direction. For example, in the 0° direction, check the pixels on the left and right. If the gradient magnitude of the current pixel is greater than the gradient magnitudes of its two neighbors, then keep the pixel; otherwise, set its gradient magnitude to 0.

[0098] The Hough circle straight line method is used to detect the luminous area. The detection principle diagram is as follows: Figure 6 As shown. Hough transform is a method for detecting specific shapes from an image, such as a line or a circle. Traverse the edge image I_nms after non-maximum suppression. For each pixel point (x, y), if its gradient amplitude is greater than the set threshold, it is regarded as a possible center of a circle. According to the center of the circle and the pre-set radius range, the accumulator array H(x, y, r) of the Hough space is calculated for each possible center of the circle. Each array element represents the accumulated evidence that a circle exists at that point and radius. The point with the maximum value in the accumulator array is selected as the optimal center and radius. The image result of batch detection of luminous areas is shown in the figure below. Figure 7 As shown, the circle corresponding to these points is the detected luminous area. The equation of the circle in the Cartesian coordinate system is: (xa) 2 +(yb) 2 =r 2 , where (a,b) is the center of the circle and r is the radius, which can also be expressed as:

[0099] x=a+r cos θ

[0100] y=b+r sin θ

[0101] The original images and the corresponding mask images are combined into a dataset, which is divided into a training set and a validation set. The ratio of the training set to the validation set is 7:3.

[0102] The U-shaped encoder-decoder network structure is built using the PyTorch deep learning framework in the Python programming language. The network structure consists of an encoder and a decoder, where the encoder is used to extract image features and the decoder is used to map these features back to the segmentation mask of the original image. The network structure diagram is shown below. Figure 8 As shown. Encoder: Contains multiple convolutional layers (Conv2D), batch normalization (BatchNorm2D) and maximum pooling layers (MaxPooling2D). The number of output channels of the convolutional layer increases layer by layer, for example: 64, 128, 256, 512, etc. The formula for the convolution operation is:

[0103] F(x)=W*x+b

[0104] Where F(x) represents the output feature map, W is the convolution kernel weight, x is the input feature map, and b is the bias term. The decoder consists of multiple upsampling layers, convolutional layers, and batch normalization. The upsampling layers amplify the feature map through interpolation and then concatenate it with the feature map from the corresponding encoder layer. The number of output channels in the convolutional layers decreases with each layer; in this network, the number is 512, 256, 128, and 64.

[0105] The U-Net model is trained using the images in the dataset. During training, the Adam optimizer is used with a learning rate of 0.001, a training batch size of 50, and a training step size of up to 4500. The cross-entropy loss function is also used to minimize the difference between the model prediction and the true segmentation mask. The formula for the cross-entropy loss function is:

[0106] L(y,t)=-∑t_i*log(y_i)

[0107] Among them, L(y,t) represents the loss value, y is the segmentation image predicted by the model, t is the true segmentation mask, and i represents the pixel index.

[0108] Perform multiple rounds of iterative training and record the model evaluation indicators (such as IOU and Dice coefficient) at the end of each epoch. IOU and Dice coefficient are commonly used segmentation task evaluation indicators, which are used to measure the similarity between the predicted segmentation results and the actual segmentation results. The IOU and Dice coefficient result graphs are respectively Figure 9 and Figure 10 Their calculation formulas are:

[0109] IOU=|A∩B| / |A∪B|

[0110] Dice=2*|A∩B| / (|A|+|B|)

[0111] Among them, A represents the predicted segmentation result, B represents the final detection result of the true segmentation, the Dice coefficient result tends to be stable with the training, and the optimal value is 92.3%, and the IOU coefficient result tends to be stable with the training, and the optimal value is 91.5%.

[0112] The trained U-Net encoder-decoder model is used to segment the image and identify luminous areas. The input image is fed into the trained U-Net model, which predicts a segmentation mask. The pixel values ​​in this mask indicate whether the corresponding location is a luminous area. The predicted mask is binarized, so that areas with a pixel value of 1 represent luminous areas, and areas with a pixel value of 0 represent non-luminous areas.

[0113] Subtract the image after dilated convolution and the image after threshold segmentation to obtain a binary image with relatively clear contours. The image after dilated convolution can highlight the edge of the luminous area, and the image after threshold segmentation can eliminate some noise. The image obtained after dilated convolution is as follows: Figure 11 By subtracting these two images, a clearer binary image containing only the edges of the luminous area can be obtained, as shown in Figure 12 shown.

[0114] Perform an eroded convolution operation on the binary image to connect the black dots that are close to each other. Eroded convolution is a morphological operation that can reduce the white area in the image. After applying the eroded convolution, the black dots near the edge of the luminous area (i.e., the non-luminous area) will be connected together, making the boundary between the luminous areas more obvious, such as Figure 13 shown.

[0115] By calculating the area of ​​dark pixels, the number of light-emitting chips and the number of abnormal chips can be identified. In the binary image after the erosion convolution operation, the connected black dots represent the non-luminous area. The number of light-emitting chips can be identified by calculating the number of connected domains (for example, using the connected component labeling algorithm). Then, the brightness, shape and other features of the light-emitting chip can be used to determine whether it is an abnormal chip. For example, if the brightness of a light-emitting chip is significantly lower than that of other light-emitting chips, or its shape is obviously irregular, it can be judged as an abnormal chip. In this way, the number of abnormal chips can be identified, such as Figure 14 shown.

[0116] This invention uses computer vision image processing algorithms to solve the problem of rapid detection of MiniLED backlight panels. A U-shaped network is used to train a large amount of simulated brightness and darkness data. The trained model has a good ability to identify the areas of normal and abnormal lamp beads, which not only demonstrates the rationality of the preprocessing method, but also can expand the application of deep learning. Finally, combined with image processing, the number of normal and abnormal chips is detected. The model can replace manual visual inspection and improve work efficiency and product quality control.

[0117] In a second aspect, an embodiment of the present invention further discloses a Mini LED backlight function testing device, such as Figure 15 As shown, the device includes: an acquisition module 151, a threshold processing module 152, a Sobel operator processing module 153, a Hough circle and line method detection module 154, a data set module 155, a U-Net model module 156, a training module 157, an image segmentation module 158, an expansion and corrosion module 159 and a recognition module 160.

[0118] In a specific embodiment, an acquisition module 151 is used to respond to acquiring a group of images of MiniLED backlight panels with different light-emitting areas; a threshold processing module 152 is used to convert the acquired image into a grayscale image, and then perform T processing using a global threshold method; a Sobel operator processing module 153 is used to apply the Sobel operator to the binarized image to obtain the horizontal gradient Gx and vertical gradient Gy of the image; a Hough circle straight line method detection module 154 is used to detect bright and dark areas based on the Hough circle straight line method, and batch produce mask images; a data set module 155 is used to form a data set with the acquired original image and the corresponding mask image, which is divided into a training set and a validation set.

[0119] The U-Net model module 156 is used to build a U-shaped encoder-decoder network structure based on the PyTorch deep learning framework to obtain a U-Net model; the training module 157 is used to use the images in the data set to obtain the U-Net model for multiple rounds of iterative training, and record the model evaluation indicators IOU and Dice coefficient at the end of each epoch; the image segmentation module 158 is used to use the trained U-Net model to segment the image to be processed to identify the luminous area; the expansion and corrosion module 159 is used to perform expansion convolution and corrosion convolution operations on the image to be processed to find the luminous and dark areas; the identification module 160 is used to identify the number of normal lamp beads and abnormal lamp beads by calculating the area of ​​dark pixels.

[0120] The present invention proposes a unique preprocessing method for MiniLED backlight panels. Since the lamp beads are independent of each other, ordinary contour extraction can only detect the outer contours of all single lamp beads, but cannot locate the contours surrounded by dimmed lamp beads, because the contour information of that area needs to be deeply mined. Through this process, the contours of abnormal areas can be quickly and accurately located. Later, a shading plate is used to simulate the brightness and darkness of different areas of the backlight panel. The above method is used for preprocessing batch operations to produce mask data. The method is used to perform deep learning training to obtain good Dice and IOU coefficients, which proves the rationality of the labeling method. It also expands the application of deep learning to the normal working area detection of mini backlight panels. Finally, combined with image processing, the number of normal chips and abnormal chips is detected.

[0121] Reference below Figure 16 , which shows an electronic device (eg Figure 1 Schematic diagram of the structure of a computer device 1600 (server or terminal device shown). Figure 16 The electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present invention.

[0122] like Figure 16 As shown, the computer device 1600 includes a central processing unit (CPU) 1601 and a graphics processing unit (GPU) 1602, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 1603 or the program loaded from the storage part 1609 to the random access memory (RAM) 1606. Various programs and data required for the operation of the device 1600 are also stored in the RAM 1604. The CPU 1601, GPU 1602, ROM 1603 and RAM 1604 are connected to each other via a bus 1605. An input / output (I / O) interface 1606 is also connected to the bus 1605.

[0123] The following components are connected to the I / O interface 1606: an input section 1607 including a keyboard, a mouse, and the like; an output section 1608 including, for example, a liquid crystal display (LCD), and a speaker; a storage section 1609 including, for example, a hard disk; and a communication section 1610 including, for example, a network interface card such as a LAN card or a modem. The communication section 1610 performs communication processing via a network such as the Internet. A drive 1611 may also be connected to the I / O interface 1606 as needed. Removable media 1612, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 1611 as needed, so that computer programs read therefrom can be installed into the storage section 1609 as needed.

[0124] In particular, according to an embodiment disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment disclosed in the present invention includes a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via the communication portion 1610, and / or installed from a removable medium 1612. When the computer program is executed by a central processing unit (CPU) 1601 and a graphics processing unit (GPU) 1602, the above-mentioned functions defined in the method of the present invention are performed.

[0125] It should be noted that the computer-readable medium described in the present invention may be a computer-readable signal medium or a computer-readable medium, or any combination thereof. Computer-readable media may be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor devices, apparatuses, or components, or any combination thereof. More specific examples of computer-readable media may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution device, apparatus, or component. In the present invention, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution apparatus, device, or device. Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wireline, optical cable, RF, or any suitable combination thereof.

[0126] Computer program code for performing the operations of the present invention may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0127] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the devices, methods and computer program products according to various embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the module, program segment, or a part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented with a dedicated hardware-based device that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0128] The modules involved in the embodiments of the present invention may be implemented in software or hardware, and may also be set in a processor.

[0129] As another aspect, the present invention further provides a computer-readable medium, which may be included in the electronic device described in the above embodiment; or may exist independently without being assembled into the electronic device. The above-mentioned computer-readable medium carries one or more programs. When the above-mentioned one or more programs are executed by the electronic device, the electronic device: responds to obtaining a group of images of MiniLED backlight panels with different luminous areas; performs threshold processing and Sobel operator processing on the obtained images; detects bright and dark areas based on the Hough circle straight line method, and batch produces mask images; forms a data set with the obtained original images and the corresponding mask images, which is divided into a training set and a verification set; builds a U-shaped encoder-decoder network structure based on the PyTorch deep learning framework to obtain a U-Net model; uses the images in the data set to obtain the U-Net model for multiple rounds of iterative training, and records the model evaluation indicators IOU and Dice coefficient at the end of each epoch; further uses the trained U-Net model to segment the image to be processed to identify the luminous area; performs dilated convolution and eroded convolution operations on the image to be processed to find the luminous and dark areas; and calculates the area of ​​dark pixels to identify the number of normal lamp beads and abnormal lamp beads.

[0130] The above description is merely an illustration of the preferred embodiments of the present invention and the technical principles employed. Those skilled in the art should understand that the scope of the invention is not limited to the technical solutions formed by the specific combination of the above-mentioned technical features, but also encompasses other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the above-mentioned inventive concept. For example, a technical solution formed by replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in the present invention.

Claims

1. A Mini LED backlight function test method, characterized in that: The method comprises the following steps: In response to acquiring a set of images of a MiniLED backlight panel having different light-emitting areas, the method includes lighting up the MiniLED backlight panel and using a light shield to simulate images of the LED chip backlight panel under different brightness conditions; Perform threshold processing and Sobel operator processing on the acquired image; The Sobel operator processing includes: using the non-maximum suppression algorithm to process the gradient magnitude image to obtain a sparse and clear edge image, traversing the gradient magnitude image I_magnitude and the gradient direction image I_direction, and for each pixel point (x, y), first determining its direction θ in the gradient direction image, which can be expressed as: ; Based on the Hough circle line method, bright and dark areas are detected and mask images are batch generated. Specifically, for each pixel point (x, y), if its gradient amplitude is greater than a set threshold, it is regarded as a possible circle center. Based on the circle center and a pre-set radius range, an accumulator array H(x, y, r) in Hough space is calculated for each possible circle center. Each array element represents the accumulated evidence of the existence of a circle at that point and radius. The point with the maximum value in the accumulator array is selected as the optimal circle center and radius. The acquired original image and the corresponding mask image are formed into a data set, which is divided into a training set and a validation set; A U-shaped encoder-decoder network structure is built based on the PyTorch deep learning framework to obtain a U-Net model. The U-Net model specifically includes an encoder, which contains multiple convolutional layers Conv2D, batch normalization BatchNorm2D, and maximum pooling layer MaxPooling2D. The number of output channels of the convolutional layer increases layer by layer. The formula for the convolution operation is: , where F(x) represents the output feature map, W represents the convolution kernel weight, x represents the input feature map, and b represents the bias term; the decoder contains multiple upsampling layers, convolution layers, and batch normalization. The upsampling layer amplifies the feature map by interpolation and then concatenates it with the feature map of the corresponding layer of the encoder; the number of output channels of the convolution layer decreases layer by layer, and in this network, it is 512, 256, 128, and 64; Use the images in the dataset to obtain the U-Net model for multiple rounds of iterative training, and record the model evaluation indicators IOU and Dice coefficient at the end of each epoch; The trained U-Net model is further used to segment the image to be processed to identify the luminous area; Performing dilated convolution and eroded convolution operations on the image to be processed to identify luminous and dark areas; including: subtracting the image after dilated convolution from the image after threshold segmentation to obtain a binary image with relatively clear contours; performing an eroded convolution operation on the binary image to connect closely spaced black dots in the image; By calculating the area of ​​dark pixels, the number of normal and abnormal lamp beads can be identified.

2. The Mini LED backlight function testing method according to claim 1, characterized in that: Threshold processing of the acquired image includes: The acquired image is converted into a grayscale image, and then the global threshold method is used for T processing, and the threshold is set to T, where Represents the processed image, and its formula can be expressed as: 。 3. The Mini LED backlight function testing method according to claim 2, characterized in that: The Sobel operator processing of the acquired image includes: Apply the Sobel operator to the binarized image to obtain the horizontal gradient Gx and vertical gradient Gy of the image. The calculation formula of the Sobel operator is as follows: 。 4. The Mini LED backlight function testing method according to claim 3, wherein: Also includes: For each interval, check the neighboring pixels of the current pixel in that direction. For example, in the 0° direction, check the pixels on the left and right. If the gradient magnitude of the current pixel is greater than the gradient magnitudes of its two neighbors, the pixel is retained; otherwise, its gradient magnitude is set to 0.

5. The Mini LED backlight function testing method according to claim 4, characterized in that: Also includes: During the training of the U-Net model, the Adam optimizer was used, and the learning rate was set to 0.001, the training batch size was 50, and the training step size was up to 4500. At the same time, the cross entropy loss function is used to minimize the difference between the model prediction and the true segmentation mask; the formula of the cross entropy loss function is: ； in, represents the loss value, is the segmentation image predicted by the model, is the true segmentation mask, Represents a pixel index.

6. The Mini LED backlight function testing method according to claim 5, characterized in that: The calculation formulas for the model evaluation indicators IOU and Dice coefficient are: ； Among them, A represents the predicted segmentation result, B represents the final detection result of the true segmentation, the Dice coefficient result tends to be stable with the training, and the optimal value is 92.3%, and the IOU coefficient result tends to be stable with the training, and the optimal value is 91.5%.

7. A Mini LED backlight function test device, characterized in that: The method according to any one of claims 1 to 6, wherein the device comprises: an acquisition module, configured to acquire, in response, a set of images of MiniLED backlight panels having different light-emitting areas; The threshold processing module is used to convert the acquired image into a grayscale image and then perform T processing using the global threshold method; The Sobel operator processing module is used to apply the Sobel operator to the binarized image to obtain the horizontal gradient Gx and vertical gradient Gy of the image; Hough circle and line method detection module, used to detect bright and dark areas based on the Hough circle and line method, and batch generate mask images; A data set module is used to form a data set by combining the acquired original image and the corresponding mask image into a training set and a validation set; The U-Net model module is used to build a U-shaped encoder-decoder network structure based on the PyTorch deep learning framework to obtain the U-Net model; The training module is used to perform multiple rounds of iterative training on the U-Net model using the images in the dataset, and record the model evaluation indicators IOU and Dice coefficient at the end of each epoch; The image segmentation module is used to segment the image to be processed using the trained U-Net model to identify the luminous area; The dilation and erosion module is used to perform dilation and erosion convolution operations on the processed image to find the luminous and dark areas; The recognition module is used to identify the number of normal lamp beads and abnormal lamp beads by calculating the area of ​​dark pixels.

8. An electronic device comprising: one or more processors; a storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images