Method and system for detecting influence degree of circadian rhythm

Abstract

The invention discloses a circadian rhythm influence degree detection method and system, and belongs to the technical field of intelligent detection. The method comprises the following steps: acquiring outdoor lighting environment image data to construct a data set, and performing physical quantity calibration on the image data to obtain related physical quantities; calculating a CS value based on the related physical quantity, and performing classification labeling on the data set according to the CS value to obtain a training set; building a deep convolutional network model, and training the deep convolutional network model based on the training set to obtain an optimized classification model; and acquiring outdoor lighting environment image data in real time, inputting the outdoor lighting environment image data into the optimization classification model, and outputting a circadian rhythm influence result. According to the method, melatonin inhibition condition information is migrated to the visible light image, whether the outdoor lighting environment affects the human body circadian rhythm or not is detected by using the image data through deep learning, the operation complexity is greatly reduced, the detection speed is increased, the detection cost is reduced, meanwhile, the good detection precision is kept, and the detection efficiency is improved. And efficient intelligent detection is realized.

Description

Technical Field

[0001] The present invention relates to the field of intelligent detection technology, and more specifically to a method and system for detecting the degree of influence of circadian rhythm. Background Art

[0002] Circadian rhythm refers to the regular changes in physiological and behavioral activities in an organism with a cycle of approximately 24 hours. In outdoor environments, light is one of the key factors affecting circadian rhythm. With the acceleration of urbanization, the complexity of the outdoor lighting environment is increasing, which has an unignorable impact on people's circadian rhythm. Circadian rhythm disorders are associated with a variety of health problems, including sleep disorders, metabolic disorders, cardiovascular diseases, etc. Therefore, it is particularly important to detect the impact of outdoor lighting environment on circadian rhythm.

[0003] At present, the detection of the impact of circadian rhythms in outdoor lighting environments mainly relies on methods such as questionnaires and biomarker analysis, which often have defects such as strong subjectivity, time-consuming and labor-intensive, and it is difficult to accurately reflect the impact of light on circadian rhythms in actual environments; and traditional detection methods lack in-depth analysis of lighting data and cannot accurately identify the changing patterns and trends of circadian rhythms under different lighting conditions.

[0004] Therefore, it is an urgent problem to be solved by those skilled in the art to propose a circadian rhythm influence detection method and system that can improve detection efficiency and accuracy. Summary of the invention

[0005] In view of this, the present invention provides a method and system for detecting the degree of influence of circadian rhythm, which can quickly, accurately, conveniently and at low cost detect whether the outdoor lighting environment has an influence on the circadian rhythm of the human body.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] In one aspect, the present invention provides a method for detecting the degree of influence of circadian rhythm, comprising the following steps:

[0008] Collecting outdoor lighting environment image data to construct a data set, and performing physical quantity calibration on the image data to obtain relevant physical quantities;

[0009] Calculating a CS value based on the relevant physical quantity, and classifying and labeling the data set according to the CS value to obtain a training set;

[0010] Building a deep convolutional network model, and training the deep convolutional network model based on the training set to obtain an optimized classification model;

[0011] The outdoor lighting environment image data is acquired in real time and input into the optimized classification model to output the circadian rhythm impact result.

[0012] Preferably, the relevant physical quantities include corneal illuminance and spectral power distribution.

[0013] Preferably, the calculation formula of the CS value is as follows:

[0014]

[0015] In the formula, CL A is the circadian light-normalized value.

[0016] Preferably, the circadian light normalized value CL A The calculation formula is as follows:

[0017]

[0018] In the formula, E λ is the spectral irradiance distribution of the light source and the related physical quantities; Mc λ Is a melanin protease; S λ It is cone cell base; mp λ V is the macular pigment transmittance; λ is the photoluminescence efficiency function; V′ λ is the dark vision luminous efficiency function; RodSat is the light intensity required to reduce the light sensitivity of rod cells by half; k, a b-y 、a rod are all parameters, k = 0.2616; a b-y =0.7000; a rod =3.3000.

[0019] Preferably, classifying and labeling the data set according to the CS value includes:

[0020] A working threshold of the circadian rhythm stimulation factor is set. When the CS value of the image data in the data set is greater than or equal to the working threshold, the image data is marked as causing non-visual negative impact on pedestrians; otherwise, the image data is marked as not causing non-visual negative impact on pedestrians.

[0021] Preferably, the deep convolutional network model includes two convolution layers with a convolution kernel of 64, one maximum pooling layer, two convolution layers with a convolution kernel of 128, one maximum pooling layer, three convolution layers with a convolution kernel of 256, one maximum pooling layer, three convolution layers with a convolution kernel of 512, one maximum pooling layer, three convolution layers with a convolution kernel of 512, one maximum pooling layer, two fully connected layers with a parameter amount of 4096, one fully connected layer with a parameter amount of 1000 and a Softmax output layer, which are connected in sequence.

[0022] Preferably, the activation function formula of the deep convolutional network model uses the Leaky_ReLU activation function.

[0023] Preferably, the calculation formula of the Softmax output layer is as follows:

[0024]

[0025] Where h is the feature map sequence number output by the last layer FC, h = 1, 2; G1 and G2 are the confidences of classification 1 and 2 respectively; S1 is the output classification probability of the first feature map; I and J are the number of rows and columns of the feature map; Xij is the value of the i-th row and j-th column of the feature map.

[0026] On the other hand, the present invention also provides a circadian rhythm influence degree detection system, which is used to implement the above circadian rhythm influence detection method, comprising:

[0027] A data acquisition and processing module is used to collect image data of outdoor lighting environment to construct a data set, and to perform physical quantity calibration on the image data to obtain relevant physical quantities;

[0028] A data set labeling module, used to calculate the CS value based on the relevant physical quantity, and classify and label the data set according to the CS value to obtain a training set;

[0029] A model optimization module, used to build a deep convolutional network model, train the deep convolutional network model based on the training set, and obtain an optimized classification model;

[0030] The detection module is used to obtain outdoor lighting environment image data in real time and input the optimized classification model to output circadian rhythm impact results.

[0031] Through the above technical solutions, it can be known that compared with the prior art, the present invention discloses a method and system for detecting the degree of circadian rhythm influence, collects outdoor lighting environment image data to construct a data set, and performs physical quantity calibration on the image data to obtain relevant physical quantities; calculates CS values ​​based on relevant physical quantities, and classifies and annotates the data set according to the CS values ​​to obtain a training set; builds a deep convolutional network model, trains the deep convolutional network model based on the training set, and obtains an optimized classification model; obtains outdoor lighting environment image data in real time and inputs the optimized classification model, and outputs circadian rhythm influence results. The present invention transfers melatonin suppression information to visible light images, and uses image data to detect whether the outdoor lighting environment affects the human circadian rhythm through deep learning, which greatly reduces the complexity of operation, improves the detection speed, and maintains good detection accuracy while reducing the detection cost, realizing efficient intelligent detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0033] Figure 1 A flow chart of the method provided by the present invention;

[0034] Figure 2 A structural diagram of the deep convolutional network model provided by the present invention;

[0035] Figure 3 This is a system architecture diagram provided by the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] On the one hand, the embodiment of the present invention discloses a method for detecting the degree of influence of circadian rhythm, referring to Figure 1 , the method comprises the following steps:

[0038] S1. Collect image data of outdoor lighting environment to construct a data set, and perform physical quantity calibration on the image data to obtain relevant physical quantities.

[0039] A camera is used to collect image data of the city's nighttime outdoor lighting environment. At the same time, the illuminance meter data is used to perform a secondary measurement of the scene to be measured at the same angle as the camera to obtain the corneal illuminance and spectral power distribution (SPD) of the measurement scene.

[0040] Furthermore, the image data acquisition angle in this embodiment simulates the perspective of a ground observer, with the camera placed vertically at an eye height of 5.6 feet from the ground, and the number of acquired images is greater than 500.

[0041] S2. Calculate the CS (circadian rhythm stimulus) value based on relevant physical quantities, and classify and label the data set according to the CS value to obtain a training set, including:

[0042] S21. Calculate the CS value according to the following formula:

[0043]

[0044] In the formula, CLA is the circadian light-normalized value.

[0045] Circadian light normalized value CL A The calculation formula is as follows:

[0046]

[0047] In the formula, E λ is the spectral irradiance distribution of the light source; Mc λ Is a melanin protease; S λ It is cone cell base; mp λ V is the macular pigment transmittance; λ is the photoluminescence efficiency function; V′ λ RodSat is the light intensity required to reduce the photosensitivity of rod cells by half, and its value in this embodiment is 6.5W / m 2 ; k, a b-y 、a rod are all parameters, k = 0.2616; a b-y =0.7000; a rod =3.3000.

[0048] Among them, the relevant physical quantities corneal illumination and SPD and E λ correspond.

[0049] S22. Classify and label the data set according to the CS value, including:

[0050] The working threshold of the circadian rhythm stimulation factor (i.e., the working threshold of acute nocturnal melatonin suppression) is set. When the CS value of the image data in the dataset is greater than or equal to the working threshold, the image data is marked as "causing non-visual negative impact on pedestrians"; otherwise, the image data is marked as "not causing non-visual negative impact on pedestrians".

[0051] In this embodiment, the working threshold is set to 0.05.

[0052] S23. Divide the classified and labeled dataset into a training set and a validation set.

[0053] S3. Build a deep convolutional network model and train it based on the training set to obtain an optimized classification model, including:

[0054] S31. Build a deep convolutional network model based on the VGG-16 network structure, such as Figure 2 shown.

[0055] The structure of the deep convolutional network model is two layers of convolutional layers with a convolution kernel of 64 (Conv3-64), one layer of maximum pooling layer (Maxpool), two layers of convolutional layers with a convolution kernel of 128 (Conv3-128), one layer of Maxpool, three layers of convolutional layers with a convolution kernel of 256 (Conv3-256), one layer of Maxpool, three layers of convolutional layers with a convolution kernel of 512 (Conv3-512), one layer of Maxpool, three layers of Conv3-512, one layer of Maxpool, two layers of fully connected layers with a parameter size of 4096 (FC-4096), one layer of fully connected layer with a parameter size of 1000 (FC-1000) and a Softmax output layer.

[0056] Among them, all convolutional layers use 3*3 convolution kernels with a step size of 1 and padding=same.

[0057] The activation function uses Leaky_ReLU. The formula is as follows:

[0058]

[0059] x i Calculate the input for the neural network, y i is the output after the activation function.

[0060] The maximum pooling layer uses a pooling kernel of size 2*2 and a stride of 2.

[0061] The calculation formula of the Softmax output layer is as follows:

[0062]

[0063] Where h is the feature map sequence number output by the last layer FC, h = 1, 2; G1 and G2 are the confidences of classification 1 and 2 respectively; S1 is the output classification probability of the first feature map, ranging from (0, 1); I and J are the number of rows and columns of the feature map; Xij is the value of the i-th row and j-th column of the feature map.

[0064] S32. The deep convolutional network model is trained based on the training set, and further adjusted based on the validation set to obtain an optimized classification model.

[0065] In this example, the training batch size is set to 256. Training is regularized by weight decay, and the L2 penalty multiplier is set to 0.0005. The first two fully connected layers are dropout regularized, and the dropout ratio is set to 0.5. The learning rate is initially set to 0.01, and then when the validation set accuracy stops improving, the learning rate is reduced by a factor of 10, and the number of iterations is set to 100.

[0066] According to the above process, the deep convolutional network model is trained and adjusted to obtain an optimized classification model.

[0067] S4. Acquire the outdoor lighting environment image data in real time and input it into the optimized classification model to output the circadian rhythm impact results.

[0068] In this embodiment, the image data input to the optimized classification model is a 224*244 RGB color image with 3 channels, that is, 224*224*3.

[0069] The optimized classification model outputs the classification probability, and then the circadian rhythm impact results are determined based on the classification probability. The circadian rhythm impact results include two categories: "causing non-visual negative impact on pedestrians" and "not causing non-visual negative impact on pedestrians."

[0070] On the other hand, the present invention also proposes a circadian rhythm influence degree detection system, which is used to implement the above circadian rhythm influence detection method, such as Figure 3 As shown, the system includes:

[0071] The data acquisition and processing module is used to collect image data of outdoor lighting environment to construct a data set, and to perform physical quantity calibration on the image data to obtain relevant physical quantities;

[0072] The data set annotation module is used to calculate the CS value based on the relevant physical quantities, and classify and annotate the data set according to the CS value to obtain the training set;

[0073] The model optimization module is used to build a deep convolutional network model and train the deep convolutional network model based on the training set to obtain an optimized classification model;

[0074] The detection module is used to obtain outdoor lighting environment image data in real time and input it into the optimized classification model to output the circadian rhythm impact results.

[0075] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0076] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting the degree of influence of circadian rhythm, characterized in that: The following steps are involved: Collecting outdoor lighting environment image data to construct a data set, and performing physical quantity calibration on the image data to obtain relevant physical quantities; Calculating a CS value based on the relevant physical quantity, and classifying and labeling the data set according to the CS value to obtain a training set; Building a deep convolutional network model, and training the deep convolutional network model based on the training set to obtain an optimized classification model; The outdoor lighting environment image data is acquired in real time and input into the optimized classification model to output the circadian rhythm impact result.

2. A circadian rhythm influence degree detection method according to claim 1, characterized in that: The relevant physical quantities include corneal illuminance and spectral power distribution.

3. A circadian rhythm influence degree detection method according to claim 2, characterized in that: The calculation formula of the CS value is as follows: In the formula, CL A is the circadian light-normalized value.

4. A circadian rhythm influence degree detection method according to claim 3, characterized in that: Circadian light normalized value CL A The calculation formula is as follows: In the formula, E λ is the spectral irradiance distribution of the light source and the related physical quantities; Mc λ Is a melanin protease; S λ It is cone cell base; mp λ V is the macular pigment transmittance; λ is the photoluminescence efficiency function; V′ λ is the dark vision luminous efficiency function; RodSat is the light intensity required to reduce the light sensitivity of rod cells by half; k, a b-y 、a rod are all parameters, k = 0.2616; a b-y =0.7000; a rod =3.3000.

5. The method for detecting the degree of influence of circadian rhythm according to claim 1, characterized in that: Classifying and labeling the data set according to the CS value includes: A working threshold of the circadian rhythm stimulation factor is set. When the CS value of the image data in the data set is greater than or equal to the working threshold, the image data is marked as causing non-visual negative impact on pedestrians; otherwise, the image data is marked as not causing non-visual negative impact on pedestrians.

6. A circadian rhythm influence degree detection method according to claim 1, characterized in that: The deep convolutional network model includes two convolutional layers with a convolution kernel of 64, one maximum pooling layer, two convolutional layers with a convolution kernel of 128, one maximum pooling layer, three convolutional layers with a convolution kernel of 256, one maximum pooling layer, three convolutional layers with a convolution kernel of 512, one maximum pooling layer, three convolutional layers with a convolution kernel of 512, one maximum pooling layer, two fully connected layers with a parameter amount of 4096, one fully connected layer with a parameter amount of 1000 and a Softmax output layer, which are connected in sequence.

7. A circadian rhythm influence degree detection method according to claim 6, characterized in that: The activation function formula of the deep convolutional network model uses the Leaky_ReLU activation function.

8. The method for detecting the degree of influence of circadian rhythm according to claim 6, characterized in that: The calculation formula of the Softmax output layer is as follows: Where h is the feature map sequence number output by the last layer FC, h = 1, 2; G1 and G2 are the confidences of classification 1 and 2 respectively; S1 is the output classification probability of the first feature map; I and J are the number of rows and columns of the feature map; Xij is the value of the i-th row and j-th column of the feature map.

9. A circadian rhythm influence degree detection system, characterized in that: include: A data acquisition and processing module is used to collect image data of outdoor lighting environment to construct a data set, and to perform physical quantity calibration on the image data to obtain relevant physical quantities; A data set labeling module, used to calculate the CS value based on the relevant physical quantity, and classify and label the data set according to the CS value to obtain a training set; A model optimization module, used to build a deep convolutional network model, train the deep convolutional network model based on the training set, and obtain an optimized classification model; The detection module is used to obtain outdoor lighting environment image data in real time and input the optimized classification model to output circadian rhythm impact results.

Images