Light source color coordinate estimation system and deep learning method thereof

Abstract

A light source color coordinate estimation system and a deep learning method thereof are disclosed. The light source color coordinate estimation system includes a plurality of light detectors, a normalization calculation circuit, and a neural network. The light detectors each have a detection wavelength band for detecting an energy integral value corresponding to the detection wavelength band in a light source. The normalization calculation circuit is electrically connected to the light detectors and divides each energy integral value by a maximum value among all energy integral values to calculate a plurality of normalized energy integral values. The neural network is electrically connected to the normalization calculation circuit to obtain all normalized energy integral values and output an estimated color coordinate.

Description

Technical Field

[0001] This invention relates to a color coordinate estimation system for a light source and a deep learning method thereof, and particularly to a color coordinate estimation system for a light source based on a neural network architecture and a deep learning method thereof. Background Technology

[0002] Different color coordinates represent different color temperatures and their color characteristics. Currently, the color coordinates of a light source are obtained by using a sensor to acquire the RGB values ​​of the light source, and then multiplying the RGB values ​​by a transformation matrix to calculate the color coordinate values.

[0003] However, converting RGB values ​​to color coordinates introduces errors. Furthermore, some light sources have brighter surfaces with higher reflectivity, while others have rougher surfaces with lower reflectivity. These differences in surface reflectivity and the intensity of the emitted light beam can also lead to errors in color coordinate calculation. Summary of the Invention

[0004] The technical problem to be solved by this invention is to provide a light source color coordinate estimation system that addresses the shortcomings of existing technologies. The system comprises: multiple photodetectors, a normalization calculation circuit, and a neural network. Each photodetector, upon receiving a light beam emitted by a light source, has its own spectral response, wherein the spectral response includes one or more detection bands and multiple energy integral values ​​corresponding to the detection bands. The normalization calculation circuit is electrically connected to the photodetectors and calculates multiple normalized energy integral values ​​by dividing each energy integral value by the largest of the energy integral values. The input of the neural network is electrically connected to the normalization calculation circuit. The neural network contains multiple neurons connected by multiple synapses, each synapse having a weight value. At least some neurons contain an activation function. The input of the neural network receives the normalized energy integral values, and the normalized energy integral values ​​are transformed into estimated color coordinates through calculations using the activation function and the weight values. The output of the neural network outputs the estimated color coordinates.

[0005] Preferably, the total number of detection bands covers 380nm to 700nm.

[0006] Preferably, the neuron comprises multiple input layer neurons, which are located in an input layer of the neural network. The number of input layer neurons is the same as the number of photodetectors, and each input layer neuron obtains the normalized energy integral value.

[0007] Preferably, the neural network includes an input layer, multiple hidden layers, and an output layer. The input layer and the output layer are respectively connected to two of the hidden layers. Any two adjacent hidden layers are connected to each other. The neurons include multiple hidden layer neurons, and each hidden layer includes a portion of the hidden layer neurons.

[0008] Preferably, the activation function used by the hidden layer neurons is the sigmoid function.

[0009] Preferably, the detection bands of the optical detector are all different.

[0010] Preferably, the photodetector is located in a user device, while the normalization calculation circuit and the neural network are located in a remote host.

[0011] Preferably, the photodetector, the normalization calculation circuit, and the neural network are housed within a system-on-a-chip.

[0012] This invention also discloses a deep learning method for estimating the color coordinates of a light source, comprising: providing multiple different light sources and sequentially executing a deep learning program on the light sources, wherein the deep learning program includes: obtaining multiple energy integral values ​​of multiple different detection bands in the light source through multiple photodetectors; calculating multiple normalized energy integral values ​​by dividing the energy integral values ​​by the largest of the energy integral values ​​through a normalization calculation circuit; receiving the normalized energy integral values ​​through a neural network, wherein the neural network includes multiple activation functions and multiple weight values; converting the normalized energy integral values ​​into estimated color coordinates according to the calculation of the activation functions and weight values ​​of the neural network; calculating a color coordinate error between the estimated color coordinates and a preset color coordinate of the light source; and adjusting at least one weight value of the neural network according to a backpropagation algorithm of the neural network and the coordinate error.

[0013] Preferably, the total number of detection bands covers 380nm to 700nm.

[0014] Preferably, after each light source completes the deep learning program, it further includes determining whether the color coordinate error is within an error convergence interval. If one of the color coordinate errors exceeds the error convergence interval, multiple other different light sources are provided to execute the deep learning program.

[0015] Preferably, the deep learning program is completed when the color coordinate error is within the error convergence interval; after the deep learning program is completed, a test light source is provided and the test light source emits a test beam, and the light source color coordinate estimation system obtains the test beam and generates an estimated color coordinate of the test beam.

[0016] One of the beneficial effects of this invention is that the light color coordinate estimation system and its deep learning method provided by this invention, through deep learning of the sensing differences of different photodetectors by the neural network, obtains the normalized energy integral values ​​of different photodetectors, wherein the sum of the spectral responses of all photodetectors preferably covers the entire visible light region (380nm~700nm). Therefore, the neural network has high anti-interference capability, resulting in high accuracy of the estimated color coordinates. Furthermore, the normalization calculation circuit normalizes the values ​​of multiple photodetectors every time the light source emits a beam. In this way, the energy integral values ​​detected by each photodetector are not weakened by the intensity of the light beam emitted by the light source or the reflectivity of the light source surface, which also results in high accuracy of the estimated color coordinates.

[0017] To further understand the features and technical content of the present invention, please refer to the following detailed description and drawings of the present invention. However, the drawings provided are for reference and illustration only and are not intended to limit the present invention. Attached Figure Description

[0018] Figure 1 This is a functional block diagram of the light source color coordinate estimation system according to the first embodiment of the present invention.

[0019] Figure 2 for Figure 1 The graph shows the relationship between the spectral response and wavelength of the six photodetectors.

[0020] Figure 3 This is a functional block diagram of the light source color coordinate estimation system according to the second embodiment of the present invention.

[0021] Figure 4 This is a functional block diagram of the light source color coordinate estimation system according to the third embodiment of the present invention.

[0022] Figure 5 This is a functional block diagram of the light source color coordinate estimation system according to the fourth embodiment of the present invention.

[0023] Figure 6 This is a deep learning method for the light source color coordinate estimation system in the first embodiment of the present invention.

[0024] Figure 7 This is a deep learning method for estimating the color coordinates of a light source according to the second embodiment of the present invention.

[0025] Figures 8A-8B The images show the x-axis and y-axis color coordinate error distributions of the light source color coordinate estimation system, respectively.

[0026] Figure 9This is a schematic diagram of the first usage state of the light source color coordinate estimation system of the present invention.

[0027] Figure 10 This is a schematic diagram of the second usage state of the light source color coordinate estimation system of the present invention. Detailed Implementation

[0028] The following specific embodiments illustrate the implementation of the "Light Source Color Coordinate Estimation System and its Deep Learning Method" provided by this invention. Those skilled in the art can understand the advantages and effects of this invention from the content provided in this specification. This invention can be implemented or applied through other different specific embodiments, and various details in this specification can also be modified and changed based on different viewpoints and applications without departing from the concept of this invention. Furthermore, the accompanying drawings of this invention are for simple illustrative purposes only and are not depictions of actual dimensions; this is stated beforehand. The following embodiments will further describe the relevant technical content of this invention in detail, but the content provided is not intended to limit the scope of protection of this invention.

[0029] It should be understood that while terms such as "first," "second," and "third" may be used in this document to describe various components or signals, these components or signals should not be limited by these terms. These terms are primarily used to distinguish one component from another, or one signal from another. Furthermore, the term "or" as used herein should be interpreted to include, as appropriate, any combination of one or more of the related listed items.

[0030] [First Embodiment]:

[0031] Figure 1 This is a functional block diagram of the light source color coordinate estimation system according to the first embodiment of the present invention. (See also...) Figure 1 As shown, the light source color coordinate estimation system 100 includes multiple photodetectors 1A-1F, a normalization calculation circuit 2, and a neural network 3. Each photodetector 1A-1F has one or more detection bands to detect the energy integral value of a light beam emitted by a light source S corresponding to different detection bands. Preferably, the sum of the detection bands covers the entire visible light region (380nm~700nm). The photodetectors 1A-1F are electrically connected to the normalization calculation circuit 2, which is further electrically connected to the neural network 3. The neural network 3 outputs a first estimated color coordinate value x and a second estimated color coordinate y. The first estimated color coordinate value x and the second estimated color coordinate value y form an estimated color coordinate (x,y). The color of the light emitted by the light source can be determined from the estimated color coordinate (x,y).

[0032] Figure 2 for Figure 1The spectral response diagrams of the aforementioned photodetectors 1A to 1F are shown. Figure 2 As shown, the horizontal axis represents wavelength (nm), and the vertical axis represents light energy. Curves SPR1 to SPR6 represent the six spectral responses of the photodetectors 1A-1F, respectively. Each spectral response reflects the detection band and energy intensity of the photodetector under light illumination. The detection bands of the photodetectors are all different; that is, the detection bands of any two photodetectors can partially overlap but cannot be completely identical, in order to maximize the diversity of the sensing range. Each photodetector 1A-1F sums the light energy within its corresponding detection band to obtain the corresponding energy integral value. For example... Figure 2 As shown, the sum of the detection bands is 380nm to 730nm, covering the visible light region (380nm to 700nm), avoiding the presence of depletion areas in the visible light band that could lead to inaccurate detection results.

[0033] Regarding the implementation of the photodetector 1A-1F, for example, it can be implemented using a hardware device, which includes any combination of circuit boards, diodes, metal-semiconductor field-effect transistors (MOS), bijunction transistors (BJTs), resistors, inductors, and capacitors. Alternatively, it can be implemented via firmware, by editing a hardware language (e.g., VHDL) and programming the hardware language into a microcontroller (MCU) or field-programmable gate array (FPGA). Alternatively, it can be implemented via software, for example, by editing C language.

[0034] See also Figure 1 The normalization calculation circuit 2 includes six input terminals 21A-21F, a central processing circuit 23, a non-volatile memory 25, and six output terminals 27A-27F. The six input terminals 21A-21F are electrically connected to the photodetectors 1A-1F, respectively. The central processing circuit 23 is electrically connected to the six input terminals 21A-21F, the non-volatile memory 25, and the six output terminals 27A-27F. The normalization calculation circuit 2 obtains six different energy integral values ​​through the six input terminals 21A-21F. The non-volatile memory 25 is a read-only memory (ROM), flash memory, or non-volatile random access memory (RVRAM) and stores the normalization algorithm. The normalization algorithm involves finding the maximum value among N different values, and then dividing each of the N values ​​by the maximum value to calculate N normalized values.

[0035] In this embodiment, the central processing circuit 23 reads and executes the normalization algorithm stored in the non-volatile memory 25 to normalize the six energy integral values. It first finds the maximum value among the six energy integral values, and then divides each of the six energy integral values ​​by the maximum value to calculate the six normalized energy integral values.

[0036] Regarding the implementation of the normalized computing circuit 2, for example, it can be implemented using a hardware device, which includes any combination of circuit boards, diodes, metal-semiconductor field-effect transistors (MOS), bijunction transistors (BJTs), resistors, inductors, and capacitors. Alternatively, it can be implemented via firmware, by editing a hardware language (e.g., VHDL) and burning the hardware language into a microcontroller (MCU) or field-programmable gate array (FPGA). Alternatively, it can be implemented via software, for example, by editing C language.

[0037] The normalization process of the normalization calculation circuit 2 has at least the following advantages: The normalization calculation circuit 2 normalizes the energy integral values ​​detected by multiple photodetectors each time the light source S emits a beam. In this way, the energy integral values ​​of the photodetectors are not weakened by differences in the intensity of the beam emitted by the light source or the reflectivity of the light source surface.

[0038] See also Figure 1 The neural network 3 includes an input layer 31, a first hidden layer 33, a second hidden layer 35, and an output layer 37. In this embodiment, the input layer 31 includes six input layer neurons. The six output terminals 27A to 27F of the normalization calculation circuit 2 are respectively connected to the six input layer neurons of the input layer 31, so that the six input layer neurons of the input layer 31 respectively obtain six normalized energy integral values.

[0039] In this embodiment, the first hidden layer 33 includes eight hidden layer neurons. Each input layer neuron of the input layer 31 is connected to the eight hidden layer neurons of the first hidden layer 33 via eight synapses. Therefore, the input layer 31 and the first hidden layer 33 are connected by a total of forty-eight synapses, and each of the forty-eight synapses has a specific weight value. Thus, the output value of each input layer neuron of the input layer 31 is multiplied by eight weight values ​​and then transmitted to the eight hidden layer neurons of the first hidden layer 33. Each hidden layer neuron of the first hidden layer 33 uses a preset activation function to operate on the input value to generate an output value. The activation function is, for example, a sigmoid function or a ReLU function, but is not limited to these.

[0040] Regarding the computation between input layer 31 and the first hidden layer 33, for example, the six input layer neurons of input layer 31 are connected to the first hidden layer neuron of the first hidden layer 33 via six synapses, and the weights of these six synapses are w1 to w6, respectively. The output values ​​of the six input layer neurons of input layer 31 are x1 to x6, respectively. At this time, the input value y1 received by the input terminal of the first hidden layer neuron of the first hidden layer 33 is y1 = x1*w1 + x2*w2 + x3*w3 + x4*w4 + x5*w5 + x6*w6. The activation function used by the first hidden layer neuron of the first hidden layer 33 is σ(x), and...

[0041] Therefore, the output value of the first hidden layer neuron in the first hidden layer 33 is σ(y1). Based on the calculation formula in the example above, the output values ​​of the other seven hidden layer neurons in the first hidden layer 33 can be deduced similarly.

[0042] In this embodiment, the second hidden layer 35 includes four hidden layer neurons. Each hidden layer neuron in the first hidden layer 33 is connected to the four hidden layer neurons in the second hidden layer 35 via four synapses. Therefore, the first hidden layer 33 and the second hidden layer 35 are connected by a total of thirty-two synapses, and each of the thirty-two synapses has a specific weight value. Thus, the output value of each hidden layer neuron in the first hidden layer 33 is first multiplied by four weight values ​​and then transmitted to the four hidden layer neurons in the second hidden layer 35. Each hidden layer neuron in the second hidden layer 35 uses a preset activation function to process the input value to generate an output value.

[0043] Regarding the calculations between the first hidden layer 33 and the second hidden layer 35, for example, the eight hidden layer neurons of the first hidden layer 33 are connected to the first hidden layer neuron of the second hidden layer 35 via eight synapses, and the weight values ​​of these eight synapses are w7 to w14 respectively. The output values ​​of the eight hidden layer neurons of the first hidden layer 33 are x7 to x14 respectively. At this time, the input value y2 received by the input terminal of the first hidden layer neuron of the second hidden layer 35 is y2 = x7*w7 + x8*w8 + x9*w9 + x10*w10 + x11*w11 + x12*w12 + x13*w13 + x14*w14.

[0044] The activation function used by the first hidden neuron in the second hidden layer 35 is σ(x), while

[0045] Therefore, the output value of the first hidden layer neuron in the second hidden layer 35 is σ(y2). Based on the calculation formula in the example above, the output values ​​of the other three hidden layer neurons in the second hidden layer 35 can be deduced similarly.

[0046] In this embodiment, output layer 37 includes two output layer neurons. Each hidden layer neuron of the second hidden layer 35 is connected to the two output layer neurons of output layer 37 via two synapses. Therefore, the second hidden layer 35 and output layer 37 are connected by a total of eight synapses, and each of the eight synapses has a specific weight value. Thus, the output value of each hidden layer neuron of the second hidden layer 35 is multiplied by two weight values ​​before being transmitted to the two output layer neurons of output layer 37. Each output layer neuron of output layer 37 uses an activation function to process the input value to generate an output value. The two output layer neurons of output layer 37 output a first estimated color coordinate value x and a second estimated color coordinate value y, respectively.

[0047] Regarding the computation between the second hidden layer 35 and the output layer 37, for example, the four hidden layer neurons of the second hidden layer 35 are connected to the first output layer neuron of the output layer 37 via four synapses, and the weights of the four synapses are w15 to w18, respectively. The output values ​​of the four hidden layer neurons of the second hidden layer 35 are x15 to x18, respectively. At this time, the input value y3 received by the input terminal of the first output layer neuron of the output layer 37 is y3 = x15*w15 + x16*w16 + x17*w17 + x18*w18. The activation function used by the first output layer neuron of the output layer 37 is σ(x), and... Therefore, the output value of the first output layer neuron in output layer 37 is σ(y3). Based on the calculation formula in the example above, the output value of the other output layer neuron in output layer 37 can be deduced similarly.

[0048] Regarding the implementation of neural network 3, for example, it can be implemented using hardware devices, which include any combination of circuit boards, diodes, metal-semiconductor field-effect transistors (MOS), bijunction transistors (BJTs), resistors, inductors, and capacitors. Alternatively, it can be implemented via firmware, by editing a hardware language (such as VHDL) and burning the hardware language into a microcontroller (MCU) or field-programmable gate array (FPGA). Alternatively, it can be implemented via software, for example, by editing C language.

[0049] [Second Embodiment]:

[0050] Figure 3 This is a functional block diagram of the light source color coordinate estimation system according to the second embodiment of the present invention. Figure 3 The second embodiment of the light source color coordinate estimation system 200 and Figure 1 The difference between the light source color coordinate estimation system 100 in the first embodiment and the light source color coordinate estimation system 200 is that the light source color coordinate estimation system 200 includes eight photodetectors 1A-1H, the normalization calculation circuit 2 includes eight input terminals 21A-21H, the eight input terminals 21A-21H are electrically connected to the photodetectors 1A-1H respectively, and the central processing circuit 23 is electrically connected to the eight input terminals 21A-21H. The central processing circuit 23 reads and executes the normalization algorithm stored in the non-volatile memory 25 to normalize the eight energy integral values ​​detected by the eight photodetectors. It first finds the maximum value among the eight energy integral values, and then divides each of the eight energy integral values ​​by the maximum value to calculate the eight normalized energy integral values. The input layer 31 of the neural network 3 includes eight neurons, and the eight output terminals 27A-27H of the normalization calculation circuit 2 are respectively connected to the eight neurons of the input layer 31, so that the eight input layer neurons of the input layer 31 respectively obtain eight normalized energy integral values.

[0051] [Third Embodiment]:

[0052] Figure 4 This is a functional block diagram of the light source color coordinate estimation system according to the third embodiment of the present invention. Figure 4 The light source color coordinate estimation system 300 of the third embodiment and Figure 1 The difference between the light source color coordinate estimation system 100 in the first embodiment and the light source color coordinate estimation system 300 is that the number of hidden layer neurons in the first hidden layer 33 is changed to five and the number of hidden layer neurons in the second hidden layer 35 is changed to three. Each hidden layer neuron in the first hidden layer 33 is connected to six hidden layer neurons in the input layer 31 via six synapses, and there are a total of thirty synapses connecting the first hidden layer 33 and the input layer 31, with each synapse having a specific weight value. Each hidden layer neuron in the second hidden layer 35 is connected to five hidden layer neurons in the first hidden layer 33 via five synapses, and there are a total of fifteen synapses connecting the first hidden layer 33 and the second hidden layer 35, with each synapse having a specific weight value.

[0053] [Fourth Embodiment]:

[0054] Figure 5 This is a functional block diagram of the light source color coordinate estimation system according to the fourth embodiment of the present invention. Figure 5 The light source color coordinate estimation system 400 of the fourth embodiment and Figure 1The difference between the light source color coordinate estimation system 100 in the first embodiment and the neural network 3 is that, in addition to the input layer 31, the first hidden layer 33, the second hidden layer 35, and the output layer 37, the neural network 3 also includes a third hidden layer 39. The third hidden layer 39 is located between the second hidden layer 35 and the output layer 37. The third hidden layer 39 contains four hidden layer neurons. Each hidden layer neuron in the second hidden layer 35 is connected to the four hidden layer neurons in the third hidden layer 39 via four synapses. A total of sixteen synapses connect the second hidden layer 35 and the third hidden layer 39, and each synapse has a specific weight value. Each hidden layer neuron in the third hidden layer 39 is connected to the two output layer neurons in the output layer 37 via two synapses. A total of eight synapses connect the third hidden layer 39 and the output layer 37, and each synapse has a specific weight value.

[0055] via Figure 1 , Figure 3 , Figure 4 and Figure 5 As can be seen from the embodiments of the light source color coordinate estimation system, the number of detectors, the number of neurons in the hidden layer, the number of hidden layers, and the activation function used by the neurons can be adjusted appropriately according to the usage requirements and the accuracy of the estimated color coordinates, and are not limited to the above embodiments.

[0056] Figure 6 The deep learning method for the light source color coordinate estimation system in the first embodiment of the present invention is as follows: Figure 6 As shown, in step S601, multiple different light sources are provided, and a deep learning program is executed on the light sources sequentially. For example, when there are ten light sources, each light source emits a beam to the light source color coordinate estimation system, allowing the light source color coordinate estimation system to obtain a training data point. When there are ten light sources in total, the light source color coordinate system obtains a total of ten training data points, and a total of ten deep learning operations are performed. The deep learning program includes at least the following steps:

[0057] In step S603, multiple photodetectors, each with a different detection band, detect multiple energy integral values ​​in the light beam emitted by the light source corresponding to the detection bands. The sum of these detection bands preferably covers the entire visible light region (380nm–700nm). For example, there are five photodetectors, each corresponding to a detection band for red, green, blue, yellow, and violet light. These detection bands partially overlap and, when summed, can cover the entire visible light region (380nm–700nm).

[0058] In step S605, the normalization calculation circuit 2 divides each energy integral value by the largest of all energy integral values ​​to calculate multiple normalized energy integral values. For example, among photodetectors that detect red, green, blue, yellow, and violet light, the energy integral value corresponding to violet light is the largest. The normalization calculation circuit divides the energy integral values ​​corresponding to red, green, blue, yellow, and violet light by the energy integral value of violet light to calculate five normalized energy integral values ​​respectively.

[0059] In step S607, all normalized energy integral values ​​are received through neural network 3, wherein neural network 3 contains multiple activation functions and multiple weight values.

[0060] In step S609, the normalized energy integral value is transformed into estimated color coordinates based on the activation function of the neural network 3 and the calculation of the weight values.

[0061] In step S611, the color coordinate error between the estimated color coordinates and the preset color coordinates of the light source is calculated. For example, the estimated color coordinates x and y calculated by the light source color coordinate estimation system are 0.3 and 0.2, respectively, while the preset color coordinates x and y of the light source are 0.25 and 0.15, respectively. The x-axis color coordinate error is -0.5, and the y-axis color coordinate error is -0.05.

[0062] In step S613, the weights of neural network 3 are adjusted based on the back propagation algorithm and color coordinate error. For example, the input layer, hidden layer, and output layer of the neural network have i neurons, j neurons, and k neurons, respectively. k T is the estimated value of the output layer. k Let E be the target value, and E be the error function, where The backpropagation algorithm adjusts the weights w of the neural chains connecting the input and hidden layers. ij and the weights W of each neural chain connecting the hidden layer and the output layer. jk To minimize the error function E, the weights of the neural chains connecting the input and hidden layers are adjusted using the following formula: The formula for adjusting the weights of each neural chain connecting the hidden layer and the output layer is as follows: Where δ is the learning rate. The number of neurons in the input layer, the number of neurons in the hidden layer, the number of hidden layers, and the setting of the learning rate δ all affect the training effect of the neural network.

[0063] Figure 7This is a deep learning method for the light source color coordinate estimation system according to the second embodiment of the present invention. Figure 7 Deep learning methods compared to Figure 6 The difference in deep learning methods lies in, Figure 7 The deep learning method further includes step S715, as for Figure 7 Steps S701 to S713 are the same as Figure 6 Steps S601 to S613. Regarding step S715, after each light source completes the deep learning program, it is determined whether the color coordinate error is within a preset error convergence interval. If one of the color coordinate errors is not within the error convergence interval, several other different light sources are provided to execute the deep learning program. For example, originally, twenty different light sources were prepared to train the light source color coordinate estimation system using deep learning, and the preset error convergence interval was -2% to 2%. When the percentage of color coordinate error of one light source is outside the error convergence interval, other light sources different from the original twenty light sources are prepared to train the light source color coordinate estimation system using deep learning.

[0064] When the color coordinate error is within the error convergence interval, the deep learning program is completed. After the deep learning program is completed, a test light source can be provided, and the test light source emits a test beam. The trained light source color coordinate estimation system obtains the test beam and generates the estimated color coordinates of the test beam. The error between the calculated estimated color coordinates of the test beam and the actual recorded color coordinates of the test light source is within -2% to 2%, which meets the requirements of use.

[0065] Figures 8A-8B These are graphs showing the x-axis and y-axis color coordinate error relationships generated by the light source color coordinate estimation system for multiple different light sources. Figures 8A-8B As shown, there are a total of twenty different light sources. These twenty light sources sequentially emit beams onto the light source color coordinate estimation system. Each light source emits a beam with preset x-axis and y-axis color coordinates. The twenty estimated x-axis color coordinates calculated by the light source color coordinate estimation system are compared with the preset twenty x-axis color coordinates to produce a result as shown in the image. Figure 8A The error percentages of the twenty x-axis color coordinates shown range from -5% to 5%, which conforms to the standards for practical use. Similarly, the twenty estimated y-axis color coordinates calculated by the light source color coordinate estimation system will be compared with the preset twenty y-axis color coordinates to produce... Figure 8B The error percentages of the twenty y-axis color coordinates shown range from -5% to 5%, which conforms to the standards for practical use.

[0066] Figure 9 This is a schematic diagram illustrating the first usage state of the light source color coordinate estimation system of the present invention. For example... Figure 9 As shown, the light detectors 1A to 1F are located on the user device T1, which can be, for example, a user's mobile device, laptop, or wearable device. The normalization calculation circuit 2 and the neural network 3 are located on the remote host T2, which can be, for example, a server or a cloud computer. The user device T1 is communicatively connected to the remote host T2. This configuration distributes the more complex calculations in the light source color coordinate estimation system to the hardware of the remote host T2 for processing.

[0067] Figure 10 This is a schematic diagram illustrating a second usage state of the light source color coordinate estimation system of the present invention. For example... Figure 10 As shown, the photodetectors 1A-1F, the normalization calculation circuit 2, and the neural network 3 are disposed in the system chip T3, and the system chip T3 can be disposed in the user's mobile device, laptop, wearable device, etc.

[0068] [Beneficial effects of the embodiments]:

[0069] One of the beneficial effects of this invention is that the light source color coordinate estimation system and its deep learning method provided by this invention, due to the deep learning of the neural network to the sensing differences of different photodetectors, each photodetector has a different detection band, and the sum of the detection bands preferably covers the entire visible light region (380nm~700nm). Therefore, the neural network has high anti-interference capability, resulting in high accuracy of the estimated color coordinates. Furthermore, the normalization calculation circuit normalizes multiple energy integral values ​​every time the light source emits a beam. In this way, the energy integral value of each photodetector is not weakened by the intensity of the light beam emitted by the light source or the reflectivity of the light source surface, also resulting in high accuracy of the estimated color coordinates. Moreover, since the neural network directly estimates the color coordinates, it avoids the errors generated when converting RGB values ​​to color coordinates.

[0070] The above-described content is merely a preferred embodiment of the present invention and is not intended to limit the scope of the claims. Therefore, any equivalent technical changes made using the description and drawings of the present invention are included in the claims of the present invention.

Claims

1. A light source color coordinate estimation system, characterized in that, include: Multiple photodetectors, each having multiple different detection bands, wherein any two adjacent detection bands partially overlap but are not completely identical; When the plurality of photodetectors receive a beam of light emitted by a light source, the plurality of photodetectors output a plurality of spectral responses, the plurality of bands of the plurality of spectral responses being the same as the plurality of detection bands, and each of the spectral responses containing an energy integral value corresponding to that detection band. A normalization calculation circuit is electrically connected to the photodetector. The normalization calculation circuit divides the energy integral value by the largest of the energy integral values ​​to calculate a plurality of normalized energy integral values. as well as A neural network has an input terminal electrically connected to a normalization calculation circuit. The neural network contains multiple neurons connected by multiple neural bonds, each neural bond having a weight value. At least some of the neurons contain an activation function. The input terminal of the neural network receives the normalized energy integral value, and the normalized energy integral value is transformed into an estimated color coordinate through the activation function and the weight value. The output terminal of the neural network outputs the estimated color coordinate.

2. The light source color coordinate estimation system as described in claim 1, characterized in that, The total range of the detection bands covers 380nm to 700nm.

3. The light source color coordinate estimation system as described in claim 1, characterized in that, The neuron comprises multiple input layer neurons, which are located in an input layer of the neural network. The number of input layer neurons is the same as the number of photodetectors. Each input layer neuron obtains the normalized energy integral value.

4. The light source color coordinate estimation system as described in claim 1, characterized in that, The neural network includes an input layer, multiple hidden layers, and an output layer. The input layer and the output layer are respectively connected to two of the hidden layers. Any two adjacent hidden layers are connected to each other. The neurons include multiple hidden layer neurons, and each hidden layer contains a portion of the hidden layer neurons.

5. The light source color coordinate estimation system as described in claim 4, characterized in that, The activation function used by the neurons in this hidden layer is the sigmoid function.

6. The light source color coordinate estimation system as described in claim 1, characterized in that, The detection bands of the optical detectors are all different.

7. The light source color coordinate estimation system as described in claim 1, characterized in that, The optical detector is located in a user device, while the normalization calculation circuit and the neural network are located in a remote host.

8. The light source color coordinate estimation system as described in claim 1, characterized in that, The photodetector, the normalization calculation circuit, and the neural network are housed within a system-on-a-chip.

9. A deep learning method for estimating the color coordinates of a light source, characterized in that, include: Multiple different light sources are provided, and a deep learning program is sequentially executed on the light sources, wherein the deep learning program includes: A light beam emitted by a light source is detected by multiple photodetectors, wherein each photodetector has multiple different detection bands, and any two adjacent detection bands partially overlap but are not completely identical. Multiple photodetectors output multiple spectral responses, and multiple bands of the multiple spectral responses are the same as the multiple detection bands and each has multiple energy integral values; Multiple normalized energy integral values ​​are calculated by dividing the energy integral value by the largest of the energy integral values ​​using a normalization calculation circuit. The normalized energy integral value is received through a neural network, wherein the neural network contains multiple activation functions and multiple weight values; The normalized energy integral value is transformed into an estimated color coordinate based on the activation function of the neural network and the calculation of the weight values. Calculate a color coordinate error between the estimated color coordinates and a preset color coordinate of the light source; and The neural network's at least one weight value is adjusted based on a backpropagation algorithm and the coordinate error.

10. The deep learning method for the light source color coordinate estimation system as described in claim 9, characterized in that, The total range of the detection bands covers 380nm to 700nm.

11. The deep learning method for the light source color coordinate estimation system as described in claim 9, characterized in that, Furthermore, after each light source completes the deep learning program, it determines whether the color coordinate error is within an error convergence range. If one of the color coordinate errors exceeds the error convergence range, it provides multiple other different light sources to execute the deep learning program.

12. The deep learning method for the light source color coordinate estimation system as described in claim 9, characterized in that, When the color coordinate error is within the error convergence interval, the deep learning program is completed; after the deep learning program is completed, a test light source is provided and the test light source emits a test beam, and the light source color coordinate estimation system obtains the test beam and generates an estimated color coordinate of the test beam.

Images