Method of installing and implementing a luminaire failure automated identification system

Abstract

The application relates to a method for installing and implementing a lamp failure automatic identification system. The method comprises the following steps: S01, installing a building lamp, installing the lamp on the building; S02, installing a master controller, a sub-controller and a distribution box, sequentially connecting the distribution box, the master controller, the sub-controller and the building lamp; S03, performing lamp debugging; S04, monitoring installation, installing a camera for monitoring, the camera is arranged to face the building, and all the building lamps can be seen in the camera picture used for monitoring; S05, determining the specific position of the lamp; S06, regularly checking lamp failure, analyzing and identifying the lamp failure through the video / picture captured by the camera; and S07, confirming the lamp with failure by adopting an optimization learning algorithm, and feeding back the position of the lamp with failure to a user. The lamp position is automatically analyzed accurately through the combination of software and hardware, optimization learning and other modes, the scheme is resistant to outdoor light interference, and the error rate of judging the lamp failure position is low.

Description

Technical Field

[0001] This invention relates to the field of image recognition and processing, and more specifically, to a method for installing and implementing an automated lighting fault identification system. Background Technology

[0002] With the rapid development of China's economy and science and technology, night lighting projects have flourished in major cities, greatly promoting the development of China's night economy. While lighting projects have developed rapidly, corresponding maintenance often lags behind. When appreciating the lighting effects, we often find that some lights are faulty, but the maintenance company has failed to detect it. This is because maintenance companies need to spend a lot of manpower and resources to inspect various light fixtures for malfunctions, and also need to record the specific locations of the faults. This identification and recording of fault locations is a very tedious task.

[0003] With the rapid development of artificial intelligence, the basic principle of using surveillance cameras to inspect lighting fixtures is to compare the grayscale values ​​of faulty and normal lighting fixtures to determine the location of the fault. The core of this process is determining the specific location of the lighting fixture and analyzing the location of the fault. However, outdoor light (such as indoor lighting, streetlights, vehicle lights, and natural ambient light) significantly interferes with the location of lighting fixtures and fault diagnosis. Furthermore, the irregular shapes of buildings make it virtually impossible to directly analyze and calculate the location of lighting fixtures from captured images.

[0004] To maximize noise reduction, we need to address not only the algorithm but also the controllability of the lighting fixtures on-site. This requires a combination of hardware and software to minimize noise and accurately pinpoint the location of the lighting fixtures. Therefore, there is an urgent need to develop a method that combines hardware and software to improve the identification of lighting fixture faults. Summary of the Invention

[0005] The technical problem to be solved by this invention is the automatic monitoring and reporting of lamp malfunctions in night scene lighting projects. It proposes an AI-based method for handling lamp malfunctions in night scene lighting, enabling users to monitor the malfunction status of building facade lighting fixtures in real time.

[0006] To solve the above-mentioned technical problems, according to one aspect of the present invention, a method for automated identification of lighting fixture faults is provided, comprising the following steps: S1, video monitoring: using a camera to target the lighting fixture to be evaluated for faults, lighting the fixture in red, green, and blue primary colors respectively, and maintaining the lighting for a specific time, recording the lighting process through video monitoring; S2, taking screenshots: receiving and recording the video monitoring, saving screenshots from the video clips, and marking the grayscale values ​​of red, green, and blue at corresponding positions in the images based on the grayscale values ​​of red, green, and blue in the images; S3, analyzing the obtained specific lighting fixture positions, determining the common area in the images of the three lighting stages (red, green, and blue). As a base point for determining the specific location of the light fixture; S4, image recognition of light fixture faults. The light fixtures are periodically identified for faults, such as once a quarter, and the identification is staggered from the lighting time during major holidays, and multiple tests are conducted (early morning, evening, late night, etc.); The light fixture location map obtained in step S3 and the saved monitoring map are set as layer 1 and layer 2 respectively. Using layer 1 as the main reference position, multiple grayscale values ​​are taken from layer 2, and the average value of the grayscale values ​​obtained in each area is calculated. If the difference between the average value and the actual grayscale value is within a certain range, the light fixture is evaluated as normal; if the difference between the average value and the actual grayscale value exceeds a certain range, the light fixture at that point is determined to be faulty.

[0007] According to an embodiment of the present invention, the method for automated identification of lighting fixture faults may further include the following steps: S5, optimization learning, wherein after step S4, personnel check the lighting fixture fault image obtained by the processor on the monitoring camera to confirm whether the lighting fixture fault judgment is correct; if all judgment results are true, the algorithm is considered correct; if it is found that some points are incorrectly judged, the processor needs to perform optimization learning again. If it is determined that there is a problem with the point judgment, the lighting fixture fault identification through the image fails, and at this time, step S5 is entered: optimization learning is performed, and steps S1, S2, S3, and S4 are repeated until the lighting fixture fault is successfully identified through the image. The optimization learning process is as follows: different collection points are changed according to different algorithms, and grayscale values ​​are re-acquired until the judgment is correct. Multiple algorithms are machine-learned. Before the next judgment of the lighting fixture fault point is reported, multiple algorithms are used to collect points and grayscale values ​​multiple times, and the judgment and analysis are performed again to determine whether it is a fault. If the judgment results are all judged as faulty, then the fault is considered to be present and the fault point is reported.

[0008] According to an embodiment of the present invention, the optimization learning algorithm can adopt a convolutional neural network (CNN). A LeNet-5 model is selected, and data is collected and labeled on-site before training. The optimization learning includes the following steps: S51, collecting an image dataset on-site and dividing the dataset into two parts: 80% for training and 20% for testing; S52, convolution calculation, randomly dropping some neurons in each layer of the convolutional neural network using Dropout, and performing forward and backward propagation using Dropout regularization. The training and testing data output from S51 are used for training and learning, and feature maps are learned from the convolutional layers through convolution calculation; Convolutional Neural Network: The purpose of convolution is to extract different features from the input. Some convolutional layers may only extract low-level features such as edges and lines. More layers can iteratively extract more complex features from low-level features. We will find that the image becomes smaller after convolution. Assuming N is the image size and F is the convolution kernel size, it is equivalent to N-F+1=5-3+1=3. S53. Activation Calculation: Activation calculation is performed using activation layers to activate the feature maps learned by the convolutional layers. S54. Pooling Calculation: Pooling calculation is performed using pooling layers to subsample the feature maps learned by the convolutional layers. Pooling subsampling includes max pooling and average pooling. Max pooling takes the maximum value within the window as the output, while average pooling takes the average of all values ​​within the window as the output. This reduces the input dimension of subsequent network layers, shrinks the model size, improves computation speed, enhances the robustness of the feature map, and prevents overfitting. S55. Algorithm Determination: Fully connected layers, including convolutional, activation, and pooling layers, are used as the feature learning and extraction layers of the convolutional neural network. The feature maps learned from the image data are ultimately applied to model tasks including classification and regression. The model task first flattens the feature maps (reshapes them into 1xN vectors) and then connects them to one or more fully connected layers for model learning. In a convolutional neural network, the receptive field is formed by the size of the region mapped onto the original image by the pixels in the feature map output by each layer of the neural network. The formula for calculating the receptive field is:

[0009] r = r + (k-1) * s

[0010] Where r is the size of the receptive field of this layer, r is the size of the receptive field of the previous layer, s is the product of the convolutions or strides before the (i-1)th layer, and k is the size of the convolution kernel.

[0011] According to an embodiment of the present invention, in step S52, in the neural network, forward propagation can be the process of information flowing directly from the previous neuron to the next neuron until the output. Based on the input of each neuron and according to the corresponding rules, the output can be calculated, and finally the loss function under the current parameters can be obtained.

[0012] The output of the hidden layer is as follows:

[0013] a11=x1·W1,1(1)+x2·W2,1(1)

[0014] a12=x1·W1,2(1)+x2·W2,2(1)

[0015] a13=x1·W1,3(1)+x2·W2,3(1)

[0016] The final output layer output is as follows:

[0017] y^=a11·W1,1(2)+a12·W1,2(2)+a13·W1,3(2)

[0018] The final loss function is calculated as follows (taking the mean squared error loss function as an example):

[0019]

[0020] Gradient descent algorithm:

[0021] The update formulas for parameters w and b are:

[0022]

[0023] According to an embodiment of the present invention, in step S52, in the neural network,

[0024] Backpropagation builds upon the forward propagation algorithm by calculating the gradient of the loss function with respect to each parameter sequentially from the output layer to the input layer of the neural network, and updating the relevant parameters given a learning rate.

[0025] The specific formula is as follows:

[0026] According to an embodiment of the present invention, the method for automated identification of lighting fixture faults may further include the following steps:

[0027] S6. The application terminal reports the lighting fixture malfunction, marks the location of the malfunctioning fixture, and sends it to the web terminal to be displayed to the user.

[0028] According to an embodiment of the present invention, the specific lighting time can be 1 to 10 seconds, preferably 3 to 7 seconds, and more preferably 5 seconds.

[0029] According to a second aspect of the present invention, a system for automated identification of lighting fixture faults is provided, comprising: a video monitoring module, the video monitoring module including a camera, the camera being aimed at the lighting fixture to be evaluated for faults, the lighting fixture being illuminated with red, green, and blue primary colors respectively, and kept illuminated for a specific time, the video monitoring module recording the illumination process via video monitoring; a screenshot module, the video monitoring module receiving and recording the video monitoring, the screenshot module being used to save screenshots of the video clips, and marking the grayscale values ​​of red, green, and blue at corresponding positions in the image according to the grayscale values ​​of red, green, and blue in the image; and a positioning module, using... Based on the specific locations of the lamps obtained from the analysis, the common area in the images of the red, green, and blue lights is determined as the base point for determining the specific location of the lamps. The image recognition module for lamp faults periodically identifies faults in the lamps and performs multiple tests. The lamp location map obtained from the positioning module and the saved image to be monitored are placed as Layer 1 and Layer 2, respectively. Using Layer 1 as the main reference position, multiple grayscale values ​​are taken from multiple points in Layer 2. The average value of the grayscale values ​​obtained for each area is then calculated. If the difference between the average value and the actual grayscale value is within a certain range, the lamp is considered normal; if the difference exceeds a certain range, the image... The image recognition lighting fault module determines that a lighting fixture at a given location is faulty. The optimization learning module then uses this fault image obtained by the processor to verify the fault diagnosis on a surveillance camera. If all the fault diagnoses from the image recognition module are accurate, the algorithm is considered correct. If any location diagnoses are found to be incorrect, the optimization learning module relearns and performs the re-diagnostic. The optimization learning module uses a convolutional neural network (CNN) algorithm, specifically the LeNet-5 model, which is trained on-site after data collection and labeling. The optimization learning process... To achieve this, different sampling points are changed according to different algorithms, and grayscale values ​​are re-acquired until the judgment is correct. Multiple machine learning algorithms are used to collect grayscale values ​​multiple times before the next identified lamp fault location is reported. The algorithm is then used to analyze whether it is a fault. If multiple judgments result in a fault, the location is considered faulty and the fault location is reported. The optimization learning process includes convolution calculation, activation calculation, pooling calculation, and algorithm determination. The application-side feedback module is used to identify the lamp fault status and the location of the faulty lamp and send it to the web interface for presentation to the user.

[0030] According to a third aspect of the present invention, an electronic device is provided, comprising: a memory, a processor, and a lighting fixture fault automatic identification program stored in the memory and executable on the processor, wherein the lighting fixture fault automatic identification program, when executed by the processor, implements the steps of the above-described lighting fixture fault automatic identification method.

[0031] According to a fourth aspect of the present invention, a computer storage medium is provided, wherein the computer storage medium stores an automatic lamp fault identification program, which, when executed by a processor, implements the steps of the above-described automatic lamp fault identification method.

[0032] According to another aspect of the present invention, a method for installing and implementing an automated lighting fixture fault identification system is provided, comprising the following steps: S01, installing building lighting fixtures on buildings; S02, installing a main controller, sub-controllers, and a distribution box, and connecting the distribution box, main controller, sub-controllers, and building lighting fixtures sequentially; S03, debugging the lighting fixtures; S04, installing a monitoring system, installing a camera for monitoring, setting the camera to face the building, and ensuring that all building lighting fixtures are visible in the camera's view; S05, determining the specific location of the lighting fixtures; S06, periodically checking for lighting fixture faults, analyzing and identifying lighting fixture faults through video / images captured by the camera; S07, if step S06 finds that the judgment of individual points is incorrect, then entering optimization learning, using an optimization learning algorithm to confirm the faulty lighting fixtures, and feeding back the location of the faulty lighting fixtures to the user.

[0033] According to an embodiment of the present invention, the luminaire may be an LED luminaire that supports the DMX512 protocol.

[0034] According to an embodiment of the present invention, step S03 may include the following steps: S031, supplying power to the lamps, main controller, and sub-controller through the distribution box; S032, drawing a lighting layout diagram based on the relationship between the lamps connected to the sub-controller's port and the actual building lighting fixtures; S033, adjusting the lighting status of all lamps and troubleshooting.

[0035] According to an embodiment of the present invention, the control camera can be a Hikvision camera, model iDS-2DF8C435MHS-AFW / SP(T5).

[0036] According to an embodiment of the present invention, in step S05, the lamps can be lit up with the three primary colors of red, green and blue respectively, and kept lit for a specific time. The camera records the lighting process through video monitoring; the image screenshots in the video clips are saved, and the gray values ​​of red, green and blue at the corresponding positions in the images are marked according to the gray values ​​of red, green and blue in the images; these are used to analyze the specific lamp positions.

[0037] According to an embodiment of the present invention, after step S05, the camera can be rotated to analyze the specific positions of other building lights, and the camera will maintain its position at the watch point.

[0038] According to an embodiment of the present invention, in step S06, all building lights can be turned on and illuminated in a single color, including red, green, and blue; video is recorded by a camera and screenshots are saved; the light fixture location map obtained in step S05 and the saved monitoring map are set as layer 1 and layer 2, respectively; grayscale values ​​are taken from multiple points in layer 2 with layer 1 as the main reference position; the average value of the grayscale values ​​obtained in each area is calculated, wherein if the average value and the actual grayscale value are within a certain range, the light fixture is considered normal; otherwise, it is considered faulty; and if it is determined to be faulty in multiple tests, the light fixture at that point is determined to be faulty, the location of the faulty light fixture is marked, and sent to the web terminal to be presented to the user.

[0039] Furthermore, the regular inspection of lighting fixtures may include a periodic inspection, conducted quarterly, with testing times including early morning, evening, and late at night. Preferably, testing times should avoid major holidays and be staggered from the lighting times during major holidays.

[0040] According to an embodiment of the present invention, in step S07, the lamp fault diagram obtained by the processor can be checked on the monitoring camera to confirm whether the lamp fault judgment is correct. If all the judgment results are true, the algorithm is considered correct. If some points are found to be incorrectly judged, the processor needs to perform optimization learning. Different sampling points are changed according to different algorithms, and gray values ​​are re-acquired until the judgment is correct. Multiple machine learning algorithms are used for optimization learning. Before the next lamp fault point is reported, multiple algorithms are used to take points and collect gray values ​​multiple times, and the judgment and analysis are performed again to determine whether it is a fault. If multiple judgment results are all judged as faults, the fault is considered to be here, and the fault point is reported. Personnel check the lamp fault diagram obtained by the processor on the monitoring camera to confirm whether the lamp fault judgment is correct. If all the judgment results are true, the algorithm is considered correct. If some points are found to be incorrectly judged, the processor needs to perform optimization learning again. This cycle is repeated to make the algorithm more and more accurate.

[0041] Furthermore, machine learning algorithms used for optimization learning may include convolution computation, activation computation, and pooling computation.

[0042] Compared with the prior art, the technical solution provided by the embodiments of the present invention can achieve at least the following beneficial effects:

[0043] This invention proposes an AI-based method for handling faults in nighttime lighting fixtures. It employs a combination of hardware and software, along with optimization learning, to accurately and automatically analyze the location of the fixtures. Compared to traditional inspection methods, efficiency is significantly improved; compared to purely software analysis methods, this method has a shorter development cycle, better resistance to interference from outdoor light (such as indoor lighting, streetlights, vehicle lights, and natural ambient light), and a much lower error rate in determining the location of faulty fixtures. Attached Figure Description

[0044] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments will be briefly described below. Obviously, the drawings described below only relate to some embodiments of the present invention and are not intended to limit the present invention.

[0045] Figure 1 This is a flowchart illustrating the automated identification of lighting fixture faults according to an embodiment of the present invention;

[0046] Figure 2 yes Figure 1 The flowchart for optimization learning. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the described embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains. The terms “first,” “second,” and similar terms used in the specification and claims of this patent application do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms “an” or “a” and similar terms do not indicate a limitation of quantity, but rather indicate the presence of at least one.

[0049] like Figure 1 and Figure 2 As shown, the method for automated identification of lighting fixture faults includes the following steps:

[0050] S1. Video monitoring: A camera is used to point at the light fixture to be evaluated for malfunction. The light fixture is turned on in red, green and blue primary colors respectively, and kept on for a specific time. The lighting process is recorded by video monitoring.

[0051] S2. Take screenshot: Receive the video surveillance footage and record it. Save the screenshot from the video clip. Mark the corresponding red, green, and blue grayscale values ​​at the image locations based on the red, green, and blue grayscale values ​​in the image.

[0052] S3. Analyze the specific locations of the lamps and determine the common area in the images of the red, green, and blue lights, using it as the base point for determining the specific locations of the lamps.

[0053] S4. Image recognition of lighting fixture malfunctions: Lighting fixtures are periodically identified for malfunctions, such as once a quarter, with the identification time staggered from major holidays, and multiple tests conducted (early morning, evening, late night, etc.). The lighting fixture location map obtained in step S3 and the saved monitoring image are set as Layer 1 and Layer 2, respectively. Using Layer 1 as the main reference position, multiple grayscale values ​​are taken from Layer 2. The average value of the grayscale values ​​obtained in each area is then calculated. If the difference between the average value and the actual grayscale value is within a certain range, the lighting fixture is considered normal; if the difference between the average value and the actual grayscale value exceeds a certain range, the lighting fixture at that location is determined to be faulty.

[0054] This invention proposes an AI-based method for handling faults in nighttime lighting fixtures. It employs a combination of hardware and software, along with optimization learning, to accurately and automatically analyze the location of the fixtures. Compared to traditional inspection methods, efficiency is significantly improved; compared to purely software analysis methods, this method has a shorter development cycle, better resistance to interference from outdoor light (such as indoor lighting, streetlights, vehicle lights, and natural ambient light), and a much lower error rate in determining the location of faulty fixtures.

[0055] According to one or more embodiments of the present invention, the method for automated identification of lighting fixture faults further includes the step of: S5, optimization learning, wherein, after step S4, personnel check the lighting fixture fault image obtained by the processor on the monitoring camera to confirm whether the lighting fixture fault judgment is correct; if all the judgment results are true, the algorithm is considered correct; if it is found that there are incorrect point judgments, the processor needs to perform optimization learning again. If it is determined that there are incorrect point judgments, the lighting fixture fault identification through the image fails, and at this time, step S5 is entered: optimization learning is performed, and steps S1, S2, S3, and S4 are repeated until the lighting fixture fault is successfully identified through the image. The optimization learning process is as follows: different collection points are changed according to different algorithms, and grayscale values ​​are re-acquired until the judgment is correct. Multiple algorithms are machine-learned. Before the next lighting fixture fault point is reported, multiple algorithms are used to collect grayscale values ​​multiple times, and the judgment and analysis are performed again to determine whether it is a fault. If the judgment results are all judged as faults, then the fault is considered to be present, and the fault point is reported.

[0056] According to one or more embodiments of the present invention, the optimization learning algorithm adopts a convolutional neural network (CNN), selects the LeNet-5 model, collects and labels data on-site, and then trains it. The optimization learning includes the following steps: S51, collects an image dataset on-site, divides the dataset into two parts, with 80% as the training set and 20% as the test set; S52, performs convolution calculation, randomly discarding some neurons in each layer of the convolutional neural network using Dropout, and performs forward and backward propagation using Dropout regularization, training and learning on the training set and test set data output from S51, and obtaining feature maps from the convolutional layers through convolution calculation; Convolutional Neural Network: The purpose of convolution operation is to extract different features from the input. Some convolutional layers may only extract some low-level features such as edges, lines, etc. More layers of the network can iteratively extract more complex features from low-level features. We will find that the image becomes smaller after convolution. Assuming N is the image size and F is the convolution kernel size, it is equivalent to N-F+1=5-3+1=3. S53. Activation Calculation: Activation calculation is performed using activation layers to activate the feature maps learned by the convolutional layers. S54. Pooling Calculation: Pooling calculation is performed using pooling layers to subsample the feature maps learned by the convolutional layers. Pooling subsampling includes max pooling and average pooling. Max pooling takes the maximum value within the window as the output, while average pooling takes the average of all values ​​within the window as the output. This reduces the input dimension of subsequent network layers, shrinks the model size, improves computation speed, enhances the robustness of the feature map, and prevents overfitting. S55. Algorithm Determination: Fully connected layers, including convolutional, activation, and pooling layers, are used as the feature learning and extraction layers of the convolutional neural network. The feature maps learned from the image data are ultimately applied to model tasks including classification and regression. The model task first flattens the feature maps (i.e., reshapes them into 1xN vectors) and then connects them to one or more fully connected layers for model learning. In a convolutional neural network, the receptive field is formed by the size of the region mapped onto the original image by the pixels in the feature map output by each layer of the neural network. The formula for calculating the receptive field is:

[0057] r = r + (k-1) * s

[0058] Where r is the size of the receptive field of this layer, r is the size of the receptive field of the previous layer, s is the product of the convolutions or strides before the (i-1)th layer, and k is the size of the convolution kernel.

[0059] According to one or more embodiments of the present invention, in step S52, in the neural network, forward propagation is the process by which information flows directly from the previous neuron to the next neuron until the output. Based on the input of each neuron and according to the corresponding rules, the output can be calculated, and finally the loss function under the current parameters is obtained.

[0060] The output of the hidden layer is as follows:

[0061] a11=x1·W1,1(1)+x2·W2,1(1)

[0062] a12=x1·W1,2(1)+x2·W2,2(1)

[0063] a13=x1·W1,3(1)+x2·W2,3(1)

[0064] The final output layer output is as follows:

[0065] y^=a11·W1,1(2)+a12·W1,2(2)+a13·W1,3(2)

[0066] The final loss function is calculated as follows (taking the mean squared error loss function as an example):

[0067]

[0068] Gradient descent algorithm:

[0069] The update formulas for parameters w and b are:

[0070]

[0071] According to an embodiment of the present invention, in step S52, in the neural network,

[0072] Backpropagation builds upon the forward propagation algorithm by calculating the gradient of the loss function with respect to each parameter sequentially from the output layer to the input layer of the neural network, and updating the relevant parameters given a learning rate.

[0073] The specific formula is as follows:

[0074] According to one or more embodiments of the present invention, the method for automated identification of lighting fixture faults further includes the step of:

[0075] S6. The application terminal reports the lighting fixture malfunction, marks the location of the malfunctioning fixture, and sends it to the web terminal to be displayed to the user.

[0076] According to one or more embodiments of the present invention, the specific lighting time is 1 to 10 seconds, preferably 3 to 7 seconds, and more preferably 5 seconds.

[0077] In use, the system that automates lighting malfunctions identifies lighting faults through images. If the system fails to identify a lighting fault, it performs optimization learning and then makes a new judgment. The application layer provides feedback on the lighting fault situation, thereby analyzing and determining the specific location of the lighting fixture.

[0078] After the building lights are installed, a nighttime light location analysis is performed. Surveillance cameras are installed and pointed at the building facade. All building lights are activated with a single color, such as red, for 5 seconds. The surveillance video is recorded, and screenshots are saved to the processor. The processor then marks all red grayscale values ​​based on their grayscale values. Next, all building lights are activated with another single color, such as green, for 5 seconds. The surveillance video is recorded, and screenshots are saved to the processor. The processor then marks all red grayscale values ​​based on their grayscale values. Finally, all building lights are activated with a third single color, such as blue, for 5 seconds. The surveillance video is recorded, and screenshots are saved to the processor. The processor then marks all blue grayscale values ​​based on their grayscale values. The processor identifies the common areas in the images from these three lighting cycles as the specific locations of the lights. The cameras rotate to analyze the specific locations of other building lights, and the cameras save these locations as watchpoints.

[0079] The lighting fixtures undergo regular fault identification, such as once a quarter, staggered from the lighting times during major holidays, and multiple tests are conducted (early morning, evening, late night, etc.). All building lights are turned on, illuminating a single color, such as red. Video recordings are captured and screenshots saved to the processor. The processor places the obtained lighting fixture location map and the saved monitoring image as Layer 1 and Layer 2, respectively. Using Layer 1 as the primary reference position, multiple grayscale values ​​are taken from multiple points in Layer 2. The average of the grayscale values ​​obtained from each area is calculated. If the average value and the actual grayscale value are within a certain range, the lighting fixture is considered normal; otherwise, it is considered faulty. Only when a lighting fixture is repeatedly identified as faulty during multiple tests can it be definitively diagnosed as faulty, and the fault is identified through image recognition. The location of the faulty lighting fixture is marked and sent to the web interface for display to the user.

[0080] If a lighting fixture malfunction is identified but not correctly, the algorithm is optimized and relearned before making a judgment. Personnel check the malfunction image obtained by the processor against the surveillance camera to confirm the accuracy of the malfunction assessment. If all the judgments are correct, the algorithm is considered correct. If some points are incorrectly identified, the processor needs to be optimized and relearned. Different sampling points are used with different algorithms to re-acquire grayscale values ​​until the judgment is correct. Multiple algorithms are used in machine learning. Before reporting the next identified malfunction point, multiple points are sampled using various algorithms, grayscale values ​​are collected, and the analysis is repeated to determine if it is a malfunction. If multiple judgments confirm a malfunction, the location is considered faulty and reported. Personnel check the malfunction image obtained by the processor against the surveillance camera to confirm the accuracy of the malfunction assessment. If all the judgments are correct, the algorithm is considered correct. If some points are incorrectly identified, the processor needs to be optimized and relearned. This process is repeated to make the algorithm increasingly accurate.

[0081] According to a second aspect of the present invention, a system for automated identification of lighting fixture faults is provided, comprising: a video monitoring module, a screenshot module, a positioning module, an image recognition module for lighting fixture faults, and an optimization learning module.

[0082] The video monitoring module includes a camera. The camera is pointed at the light fixture to be evaluated for malfunction. The light fixture is turned on with red, green and blue primary colors respectively, and kept on for a specific time. The video monitoring module records the lighting process through video surveillance.

[0083] The video monitoring module receives and records the video, while the screenshot module saves screenshots of the video clips and marks the corresponding red, green, and blue grayscale values ​​at the image locations based on the red, green, and blue grayscale values ​​in the images.

[0084] The positioning module is used to analyze the specific locations of the lamps and determine the common area in the images of the red, green, and blue lights, which serves as the base point for determining the specific location of the lamps.

[0085] The image recognition lighting fixture fault module periodically identifies faults in the lighting fixtures and performs multiple tests. The lighting fixture location map obtained by the positioning module and the saved monitoring image are set as Layer 1 and Layer 2, respectively. Using Layer 1 as the main reference position, multiple grayscale values ​​are taken from multiple points in Layer 2. The average value of the grayscale values ​​obtained in each area is then calculated. If the difference between the average value and the actual grayscale value is within a certain range, the lighting fixture is considered to be normal. If the difference between the average value and the actual grayscale value exceeds a certain range, the image recognition lighting fixture fault module determines that the lighting fixture at that point is faulty.

[0086] Personnel verify the faulty lighting images obtained by the processor against the monitoring camera to confirm the accuracy of the fault assessment. If all the fault assessments by the image recognition module are accurate, the algorithm is considered correct. If the module makes incorrect point identifications, the optimization learning module relearns and re-learns. This optimization learning module uses a convolutional neural network (CNN) with a LeNet-5 model, trained on-site with collected and labeled data. The optimization learning process involves changing the sampling points according to different algorithms, re-acquiring grayscale values, and repeating this process until a correct assessment is obtained. Multiple algorithms are used to acquire grayscale values ​​multiple times before reporting the next identified faulty lighting point. If multiple assessments consistently result in a fault, the fault is confirmed and reported. The optimization learning process includes convolution calculations, activation calculations, pooling calculations, and algorithm determination. The application-side feedback module identifies the faulty lighting condition and its location and sends this information to the web interface for user feedback.

[0087] According to another aspect of the present invention, an apparatus for automatic identification of lighting fixture faults is provided, comprising: a memory, a processor, and an automatic identification program for lighting fixture faults stored in the memory and executable on the processor, wherein the automatic identification program for lighting fixture faults, when executed by the processor, implements the steps of the above-described automatic identification method for lighting fixture faults.

[0088] The present invention also provides a computer storage medium.

[0089] The computer storage medium stores an automatic lamp fault identification program, which, when executed by the processor, implements the steps of the above-described automatic lamp fault identification method.

[0090] The method implemented when the automatic lamp fault identification program running on the processor is executed can be referred to in various embodiments of the automatic lamp fault identification method of the present invention, and will not be repeated here.

[0091] The present invention also provides a computer program product.

[0092] The computer program product of the present invention includes an automatic lamp fault identification program, which, when executed by a processor, implements the steps of the automatic lamp fault identification method as described above.

[0093] The method implemented when the automatic lamp fault identification program running on the processor is executed can be referred to in various embodiments of the automatic lamp fault identification method of the present invention, and will not be repeated here.

[0094] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0095] According to another aspect of the present invention, a method for installing and implementing an automated lamp fault identification system is provided:

[0096] 1. For building lighting installation, select LED lights that support the DMX512 protocol.

[0097] 2. Install the main controller, sub-controllers, and distribution box, and connect them to the building lighting fixtures.

[0098] 3. Lighting fixture adjustment:

[0099] 3.1 Power is supplied to the lighting fixtures, main controller, and sub-controllers through the distribution box;

[0100] 3.2 Draw a lighting layout diagram based on the relationship between the lighting fixtures connected to the sub-controller's ports and the actual building lighting fixtures;

[0101] 3.3 Test the lighting status of all lights and troubleshoot any faults.

[0102] 4. Monitoring Installation: This invention selects Hikvision surveillance cameras, preferably the iDS-2DF8C435MHS-AFW / SP(T5) model, but specific models are not limited to this. The camera should be positioned facing directly into the building, and the camera's view should show all the lights inside the building.

[0103] 5. Turn on all building lights in a single color, such as red, for 5 seconds;

[0104] 6. Monitor and record video, and save screenshots to the processor. The processor marks all red grayscale values ​​based on the red grayscale values ​​(including a certain deviation range).

[0105] 7. Turn on all the building lights with a different single color, such as green, for 5 seconds;

[0106] 8. Monitor and record video, and save screenshots to the processor. The processor marks all red grayscale values ​​based on the green grayscale values ​​(within a certain deviation range).

[0107] 9. Turn on all building lights using the third single color, such as blue, for 5 seconds;

[0108] 10. Monitor and record video, and save screenshots to the processor. The processor marks all blue grayscale values ​​based on the blue grayscale values ​​(within a certain deviation range).

[0109] 11. The processor identifies the common area in the images of the three lighting events and defines it as the specific location of the light fixture.

[0110] 12. The camera rotates to analyze the specific locations of other building lights, and the camera saves the locations as watchpoints.

[0111] 13. Regularly identify faults in the lighting fixtures, such as once a quarter, and stagger the lighting time with major holidays, and conduct multiple tests (early morning, evening, late at night, etc.);

[0112] Turn on all the building lights, and illuminate them in a single color, such as red.

[0113] The surveillance video is recorded and screenshots are saved to the processor;

[0114] The processor sets the lamp location map obtained in step 2 and the saved monitoring map as layer 1 and layer 2, respectively;

[0115] Using layer 1 as the primary reference position, perform multi-point grayscale value extraction in layer 2;

[0116] The average value of the grayscale values ​​obtained from each region is calculated.

[0117] If the average value and the actual grayscale value are within a certain range, the lamp is considered normal; otherwise, it is considered faulty.

[0118] The light fixture at that location can only be determined to be faulty if it is identified as faulty in multiple tests.

[0119] The location of the faulty light fixture is marked and sent to the web interface for display to the user.

[0120] 14. Personnel verify the lighting fault diagram obtained by the processor on the monitoring camera to confirm whether the lighting fault diagnosis is correct.

[0121] If all the judgment results are true, the algorithm is considered correct. If some points are found to be incorrectly judged, the processor needs to be optimized and learned.

[0122] Different sampling points are changed according to different algorithms, and grayscale values ​​are re-acquired until the judgment is correct;

[0123] Machine learning uses multiple algorithms to take multiple points and collect grayscale values ​​before reporting the next identified faulty lighting location, and then performs another judgment and analysis to determine whether it is a fault.

[0124] If multiple assessments all indicate a fault, then the fault is considered to be present, and the fault location is reported.

[0125] Personnel check the fault diagram of the lighting fixtures obtained by the processor on the monitoring camera to confirm whether the fault diagnosis of the lighting fixtures is correct.

[0126] If all the judgment results are true, the algorithm is considered correct. If some points are found to be incorrectly judged, the processor needs to be optimized and learned again.

[0127] This cycle continues, making the algorithm increasingly accurate.

[0128] The above description is merely an exemplary embodiment of the present invention and is not intended to limit the scope of protection of the present invention, which is determined by the appended claims.

Claims

1. A method for installing and implementing an automated lighting fixture fault identification system, comprising the following steps: S01. Install building lighting fixtures on the building; S02. Install the main controller, sub-controller, and distribution box, and connect the distribution box, main controller, sub-controller, and building lighting fixtures in sequence; S03. Perform lighting fixture adjustments; S04. Monitoring installation: Install a camera for monitoring. The camera is set to face the building so that all building lights can be seen in the camera's view. S05. Determine the specific location of the light fixtures; S06. Regularly check for lighting fixture malfunctions and analyze and identify malfunctions using videos / images captured by cameras; S07. If step S06 finds that the judgment of some points is incorrect, then the optimization learning process begins. By using the optimization learning algorithm, the faulty light fixtures are identified, and the location of the faulty light fixtures is fed back to the user. In step S06, all building lights are turned on, illuminating a single color, including red, green, and blue; the camera records video and saves screenshots; the light location map obtained in step S05 and the saved monitoring map are set as layer 1 and layer 2 respectively; grayscale values ​​are taken from multiple points in layer 2, using layer 1 as the main reference position; the average value of the grayscale values ​​obtained in each area is calculated. If the average value and the actual grayscale value are within a certain range, the light fixture is considered normal; otherwise, it is considered faulty. If the light fixture is determined to be faulty in multiple tests, then the location of the faulty light fixture is marked and sent to the web interface to be presented to the user.

2. The method as described in claim 1, wherein, The lighting fixture is an LED fixture that supports the DMX512 protocol.

3. The method as described in claim 1, wherein, Step S03 includes the following steps: S031. Power is supplied to the lamps, main controller, and sub-controller through the distribution box; S032. Draw a lighting layout diagram based on the relationship between the lighting fixtures connected to the sub-controller's ports and the actual building lighting fixtures. S033. Check the lighting status of all lights and troubleshoot any faults.

4. The method of claim 1, wherein, The surveillance camera is a Hikvision camera, model iDS-2DF8C435MHS-AFW / SP(T5).

5. The method of claim 1, wherein, In step S05, the lamps are lit up with the three primary colors of red, green, and blue respectively, and kept lit for a specific time. The camera records the lighting process through video monitoring. The image screenshots in the video clips are saved, and the gray values ​​of red, green, and blue at the corresponding positions in the images are marked according to the gray values ​​of red, green, and blue in the images. These are used to analyze the specific lamp positions.

6. The method of claim 1, wherein, After step S05, the camera is rotated to analyze the specific locations of other building lights, while the camera remains at the watch point.

7. The method as described in claim 1, wherein the "periodic" in the periodic inspection of lighting fixture malfunctions includes: The testing period lasts for one quarter, including early morning, evening, and late at night.

8. In the method described in claim 1, in step S07, the lamp fault diagram obtained by the processor is checked on the monitoring camera to confirm whether the lamp fault judgment is correct; if all the judgment results are true, the algorithm is considered to be correct; if some points are found to be incorrectly judged, the processor needs to be optimized and learned; different collection points are changed according to different algorithms, and grayscale values ​​are re-acquired until the judgment is correct. Multiple machine learning algorithms are used for optimization learning. Before the next identified faulty lighting location is reported, multiple algorithms are used to take multiple points, collect grayscale values, and perform another judgment and analysis to determine whether it is a fault. If the results of multiple judgments all indicate a fault, then the fault is considered to be present and the fault location is reported. Personnel check the fault diagram of the lighting fixtures obtained by the processor on the monitoring camera to confirm whether the fault diagnosis is correct. If all the diagnoses are true, the algorithm is considered correct. If some points are found to be incorrectly diagnosed, the processor needs to be optimized and learned again. This process is repeated to make the algorithm more and more accurate.

9. The method of claim 8, wherein the machine learning algorithm for optimization learning includes convolution computation, activation computation, and pooling computation.

Images