TW201721111A - Chromatic detector of LED light source - Google Patents

Abstract

A set of chromatic detector of LED light source include both a module of detector and a module of signal treatment. In the module of chromatic detector, three economic detected sensors of red, green, and blue are used to get the signal of voltage, respectively. Using the detected voltages and a color standard as the reference are sent to the module of signal treatment which is used the artificial neural network (ANN) to set a neural network replacing the complex function operation of these voltage signals, so the way could detect the color clear and efficiency. The method could detect the color distribution of light source with low cost, small volume, easy calculation, fast detection and high precision.

Description

LED light source color detector

The invention relates to a technical field of an LED light source color detector for measuring a mixed light chromaticity coordinate of an LED lighting fixture, in particular to an LED having the advantages of low cost, small volume, easy correction, fast detection speed and high precision. Light source color detector.

Among the information obtained by the human eye, "color" is an extremely important one, because people's visual information increases its depth and breadth due to color. The color of light observed by people's eyes is made up of different wavelengths of light. Although the current LED lighting is mainly based on single-crystal white LEDs, in order to increase the added value of LED lamps, it can be based on The use of adjustable color light for different use conditions is a trend nowadays. Therefore, in the near future, indoor and outdoor lighting will be dominated by polycrystalline white LEDs to achieve the above-mentioned dimming color. Additional function, and the color of the polycrystalline white LED light source is required to be within a predetermined standard range at the factory to make it conform to the actual use condition, so the color of the light source (such as brightness, chromaticity, etc.) The detection is very important. At present, the LED light source color detector for detecting the chromaticity of the polycrystalline white LED is mainly used as a receiving component by using a color filter and a sensor having high matching with the CIE1931 color matching function, and the color value must be discriminated. Detected by spectrometer or colorimeter, etc., so The measurement operation and structure are very complicated and expensive, especially if it needs to be corrected by the original technology and standard color ticket. Therefore, it is very time-consuming and troublesome to return to the original factory for calibration every time. Moreover, it is often impossible to correct by itself and can only be returned to the original factory calibration based on the time of use. However, there are many factors that cause the detection to lose accuracy. Therefore, the timing of the correction cannot be grasped by the use time alone. It is often the case that there is no immediate correction and there is a long time of inaccurate detection, so it cannot be widely used by LED smart lighting technology.

The conventional LED light source color detector has many problems such as complicated operation, high price, and troublesome time-consuming correction, which cannot be generally used by ordinary people.

The invention provides an LED light source color detector, which comprises a detection module and a processing module. The detection module includes three sensing components and three amplification circuits, and the three sensing components respectively detect the intensity of red, green, and blue light emitted by the LED light source, and convert the same into a sensing voltage signal. The three amplifying circuit can amplify the sensing voltage signal to an appropriate size and then send it out. The processing module includes a processor, an input unit, and a display. The processor can receive the sensing voltage signal detected by the sensing module, and the processor has a built-in neural network system. The neural network system can detect the sensing module first. The standard LED light source obtains the corresponding sensing voltage signal, and cooperates with a pre-input to calculate and learn the color standard value of the standard LED light source to construct a neural network detection model having a corresponding parameter, and then When detecting the LED light source for detection, this place After receiving the signal of the sensing voltage detected by the sensing module, the processor can directly perform calculation according to the pre-constructed neural network detection model to calculate a corresponding color detection value. The input unit is coupled to the processor and the color standard value can be entered. The display is coupled to the processor for displaying the color detection value.

The LED light source color detector provided by the invention can find out the stimulation of the standard LED light source 30 by using the low-cost sensing module and the operation and learning mechanism of the neural network system such as the processing module. The value is related to the input and output relationship between the two sensed voltage signals detected by the sensing module, and then the spectral and color signal errors between the two are corrected to match the CIE1931 function to find an optimized one. After the neural network detection model, the neural network detection model such as the construction can be used to calculate the signal of the sensing voltage detected by the sensing module to detect the LED light source to obtain the color thereof. Detected value. Therefore, it has the effects of low cost, small size, fast detection speed and high precision. In particular, the standard LED light source 30 can be used to construct the neural network detection model 25 for correction at any time, so the correction is also relatively easy, simple and convenient.

10‧‧‧Sensing module

11r, 11g, 11b‧‧‧ sensing components

12‧‧‧Amplification circuit

13‧‧‧Sense voltage signal

20‧‧‧Processing module

21‧‧‧ Processor

22‧‧‧Input unit

23‧‧‧ Display

24‧‧‧ class neural network system

25‧‧‧ Neural Network Detection Model

30‧‧‧Standard LED light source

31‧‧‧Color standard value

32‧‧‧Standard color illuminance meter

33‧‧‧To detect LED light source

34‧‧‧Color detection value

Figure 1 is a schematic diagram of the system architecture of the present invention.

Figure 2 is a schematic diagram of the calibration and detection process of the present invention.

Fig. 3 is a flow chart showing the construction of a neural network detection model such as the present invention.

Figure 4 is a schematic view of the detection distance of the present invention.

Please refer to FIG. 1 and FIG. 2 , which shows that the LED light source color detector of the present invention comprises a sensing module 10 and a processing module 20 . among them:

The sensing module 10 includes three sensing elements 11r, 11g, 11b and three amplifying circuits 12. The three sensing elements 11r, 11g, and 11b are configured to detect the intensity of red, green, and blue light emitted by the LED light source, respectively, and convert it into a sensing voltage signal 13. The three amplifying circuits (OPA) 12 are respectively connected to the three sensing elements 11r, 11g, 11b, and the smaller sensing voltage signals 13 are amplified to an appropriate size and then sent out. In the present invention, the three sensing elements 11r, 11g, 11b use HAMAMATSU Japanese Hamamatsu three-color sensing elements S6428-01, S6429-01, S6430-01 (or sensing elements equal or similar thereto), The price is lower.

The processing module 20 includes a processor 21, an input unit 22, and a display 23. The processor 21 receives the sensing voltage signal 13 detected by the sensing module 10, and the processor 21 A neural network system 24 is built in the system. The neural network system 24 can detect a standard LED light source 30 by the sensing module 10 to obtain a corresponding sensing voltage signal 13 and cooperate with a pre-input. The color standard value 31 (such as the color tristimulus value) of the standard LED light source 30 is calculated and learned to construct a neural network detection model 25 having a corresponding parameter, instead of detecting the sensing module 10 The measured voltage signal 13 and the color standard value 31 are converted into a large and complex function operation required for the nonlinear relationship, and then the processor 21 receives the sensing module 10 when detecting the LED light source 33 to be detected. After detecting the signal 13 of the sensing voltage, it can directly perform calculation according to the pre-constructed neural network detecting model 25 to calculate a corresponding color detection value 34 (such as color three) Stimulus value). The input unit 22 is connected to the processor 21 and can input the color standard value 31. The display 23 is coupled to the processor 21 for displaying the color detection value 34. In the present invention, the input unit 22 can be directly connected to the standard color illuminance meter 32 in addition to the color standard value 31 (such as the Konica INEON color illuminance meter KNOICA MINOLTA CL-200A color illuminance). The standard LED light source 30 is detected by the standard color illuminometer 32 to directly generate the color standard value 31. The standard LED light source 30 and the LED light source 33 to be detected are multi-color mixed light LED array light sources.

Please refer to FIG. 3 again, indicating that the neural network system 24 of the present invention is made by using the mathematical software MATLAB, and the neural network detection model 25 is based on the supervised inverted-transverse nerve. Built by the Back-Propagation Network (BP) algorithm, the supervised inverted-transfer neural network algorithm is based on the Brooklyn Fletcher Goldfard Shanno (BFGS method) and the conjugate gradient method (Fletcher Reeves Conjugate Gradient, The FRCG method, the proportional conjugate gradient method (Scaled Conjugate Gradient, SCG method) and the Levenberg-Marquardt method (LM method) are the algorithms for establishing network learning.

The step of constructing the neural network detection model 25 by using the neural network system 24 is as follows: (1) using the sensing voltage signal 13 and the color standard value 31 of the standard LED light source 30 as the neural network-like vector. And target parameters; (2) using quasi-Newton method, conjugate gradient method, proportional conjugate gradient The method and the Levenberg-Marquardt method use four different algorithms for neural network training; (3) when the neural network training reliability is <0.98, repeat step (1), when the neural network training reliability >0.98, the original function approximation is performed; (4) The approximation function is compared with the color difference for error analysis and correction; (5) The best class of neural network module is saved, that is, the chromaticity coordinate type neural network detection model is completed. Construction.

Referring to FIG. 4, it is pointed out that when the detection module 10 of the present invention detects the standard LED light source 30 or the LED light source 33 to be detected, in order to treat it as a point light source, the measurement distance needs to be in the standard LED. The light source 30 or the LED mixed light array of the LED light source 33 is to be detected more than 10 times diagonal distance, and is at the position of the center point of the LED mixed light array. For example, if the diagonal distance of the LED light mixing array is 8 cm, the measuring distance is preferably 80 to 120 cm from the light source.

The LED light source color detector provided by the present invention can be found by using the low-cost sensing module 10 and the neural network computing and learning mechanism of the neural network system 24 such as the processing module 20 to find out The input and output relationship between the three stimulus values of the standard LED light source 30 and the sensing voltage signal 13 of the red, green and blue light detected by the three sensing elements 11r, 11g, 11b of the sensing module 10, Then correct the spectral and color signal error between the two to match the CIE1931 function to find the optimized neural network detection model 25, and then the neural network detection model 25 can be completed by the construction. The signal 13 detected by the sensing module 10 to detect the sensing voltage of the LED light source 33 is calculated to obtain a color detection value 34 thereof. and so, It has the advantages of low cost, small size, fast detection speed and high precision. In particular, the standard LED light source 30 can be used to construct the neural network detection model 25 for correction at any time, so the correction is also relatively easy, simple and convenient.

10‧‧‧Sensing module

11r, 11g, 11b‧‧‧ sensing components

12‧‧‧Amplification circuit

13‧‧‧Sense voltage signal

20‧‧‧Processing module

21‧‧‧ Processor

22‧‧‧Input unit

23‧‧‧ Display

24‧‧‧ class neural network system

25‧‧‧ Neural Network Detection Model

30‧‧‧Standard LED light source

32‧‧‧Standard color illuminance meter

Claims (10)

An LED light source color detector comprises: a detecting module comprising three sensing components and three amplifying circuits, wherein the three sensing components are respectively configured to detect the intensity of red light, green light and blue light emitted by the LED light source, And converting the signal into a sensing voltage signal, wherein the three amplifying circuits are respectively connected to the three sensing elements and the smaller sensing voltage signals are amplified to an appropriate size and then sent out; and a processing module includes a processing a sensor, an input unit and a display, the processor can receive the sensing voltage signal detected by the sensing module, and the processor has a neural network system built therein, and the neural network system can First, the sensing module detects a standard LED light source to obtain a corresponding sensing voltage signal, and then performs a calculation and learning with a pre-input color standard value corresponding to the standard LED light source to construct one of the corresponding parameters. The neural network detection model, when the LED light source is to be detected, the processor receives the signal of the sensing voltage detected by the sensing module, and then directly according to the pre-constructed class. For detection by the network model calculation to calculate a value corresponding to the color of the detection, the input unit may be connected to the processor and the input color standard value, the display connected to the processor to display the color detection value. The LED light source color detector according to claim 1, wherein the three sensing elements are HAMAMATSU Japanese Hamamatsu three-color sensing elements S6428-01, S6429-01, and S6430-01. The LED light source color detector according to claim 1, wherein the color standard value and the color detection value are color tristimulus values. The LED light source color detector according to claim 3, wherein the color standard value is Can be entered manually. The LED light source color detector according to claim 3, wherein the input unit is directly connectable to a standard color illuminometer, and the standard LED illuminometer is used to detect the standard LED light source to directly generate the color standard value. The LED light source color detector according to claim 5, wherein the standard color illuminance meter is a Konica Minolta color illuminometer KNOICA MINOLTA CL-200A color illuminometer. The LED light source color detector according to claim 1, wherein the standard LED light source and the LED light source to be detected are multi-color mixed light LED array light sources. The LED light source color detector according to claim 1, wherein the neural network detection model is constructed by using a Back-Propagation Network (BP) algorithm. The LED light source color detector according to claim 8, wherein the supervised inverted-transfer neural network algorithm adopts a pseudo-Newton method (Broyden Fletcher Goldfard Shanno, BFGS method) and a conjugate gradient method (Fletcher Reeves Conjugate Gradient, FRCG). The method, the proportional conjugate gradient method (Scaled Conjugate Gradient, SCG method) and the Levenberg-Marquardt method (LM method) are used to establish the algorithm of network learning. The LED light source color detector according to claim 1, wherein the detecting module detects the standard LED light source or the LED light source to be detected, and the measuring distance needs to be mixed in the standard LED light source or the LED to be detected by the LED light source. The diagonal distance of the array is more than 10 times and is the position of the center point of the LED mixing array.