Sensing device, detection system, and urine detection system

Abstract

This disclosure relates to a sensing device, a detection system, and a urine detection system. The device includes: a filter assembly for encoding incident light to obtain imaging information for each sensing channel; and a detection assembly for detecting the imaging information and generating a rectangular grayscale image array. The rectangular grayscale image array is input to a detection model to obtain a detection result using the output of the detection model. The detection model has a mapping relationship between the rectangular grayscale image array and the detection result. The detection result includes the composition of the analyte or the composition of the analyte and the content of each component. The sensing device of this disclosure directly detects the rectangular grayscale image array and inputs it into the detection model to obtain the detection result. It has the advantages of low cost, small size, simplified structure, and high accuracy, thus improving measurement accuracy.

Description

Technical Field

[0001] This disclosure relates to the field of detection technology, and more particularly to a sensing device, a detection system, and a urine detection system. Background Technology

[0002] Spectroscopic sensing is a method for determining the composition and content of a target analyte based on color changes caused by physical and chemical reactions. It has wide applications in medicine, environmental monitoring, agriculture, and other fields. However, current spectral sensing solutions for color detection suffer from problems such as large size, low accuracy, complex structure, and high cost. Summary of the Invention

[0003] According to one aspect of this disclosure, a sensing device is provided, the device comprising:

[0004] A filter assembly is used to encode incident light to obtain imaging information for each sensing channel. The imaging information includes the light intensity value of the incident light. The filter assembly includes multiple different types of filters, each filter corresponding to a sensing channel. Different filters can encode incident light to obtain different imaging information.

[0005] A detection component is used to detect the imaging information and generate a rectangular grayscale image array. The rectangular grayscale image array is used to input into a detection model to obtain a detection result using the output of the detection model. The detection model has a mapping relationship between the rectangular grayscale image array and the detection result. The detection result includes the components of the analyte or the components of the analyte and the content of each component.

[0006] In one possible implementation, the detection model is used to determine the detection result based on the rectangular grayscale image array and the mapping relationship.

[0007] In one possible implementation, the detection model is further used to preprocess the rectangular grayscale image array, wherein the preprocessing method includes at least one of the following:

[0008] Calculate the average value of the corresponding pixels in the multiple rectangular grayscale image arrays;

[0009] The light intensity non-uniformity of each pixel in the rectangular grayscale image array is corrected.

[0010] In one possible implementation, the detection model is based on at least one of least squares, neural networks, support vector machines, Naive Bayes classification, decision trees, k-nearest neighbors, linear discriminant analysis, linear regression, logistic regression, classification and regression trees, learning vector quantization, bagging, and random forest.

[0011] In one possible implementation, if the detection model is based on the least squares method, then the detection model is used for:

[0012] The light intensity of the corresponding rectangular regions of multiple rectangular grayscale image arrays is averaged to obtain multiple average light intensity values;

[0013] The multiple average light intensity values ​​are concatenated to obtain an intensity vector;

[0014] The intensity vector is used as input to perform a least squares operation, and the result is used as the detection result, which includes the content of each component.

[0015] In one possible implementation, if the detection model is based on a neural network, then the detection model is used for:

[0016] Extract the image features of the rectangular grayscale image array;

[0017] The extracted image features are subjected to multiple convolution and fully connected operations to output the detection results.

[0018] In one possible implementation, the filter type includes at least one type of metasurface filter, photonic crystal filter, perovskite quantum dot filter, and colloidal quantum dot filter, with each filter type encompassing multiple different varieties.

[0019] The detection assembly includes at least one of a complementary metal-oxide-semiconductor element, a charge-coupled device, an ultraviolet detection element, and an indium gallium arsenide near-infrared detection element.

[0020] In one possible implementation, the filter combination includes colloidal quantum dot filters, each with a different spectral transmission relationship. The filters encode the incident light based on the spectral transmission relationship and the spectral sensitivity relationship of the corresponding detection component of each filter, thereby obtaining the imaging information of the incident light. The spectral sensitivity relationship represents the relationship between photoresponsivity and light wavelength.

[0021] In one possible implementation, the rectangular grayscale image array includes multiple rectangular regions, with each filter corresponding to one rectangular region, and each rectangular region including multiple pixels.

[0022] In one possible implementation, the filter assembly is determined in the following manner:

[0023] Multiple filter assemblies are formed by selecting a different number of filters from N types of filters. Each of the N types of filters has a different spectral transmission relationship. The N types of filters can encode incident light within the target wavelength range, where N is a positive integer.

[0024] Among the detection results corresponding to filter assemblies with different numbers of filters, the minimum number of filters is selected from the detection results that reach the first preset detection result, and the minimum number is used as the number of filters in the filter assembly. The filter assembly is used to encode incident light into imaging information, and the imaging information includes the light intensity value of the incident light. The filter assembly includes multiple different types of filters, and different filters can encode incident light to obtain imaging information.

[0025] The minimum number of combinations of continuous or skip distributions are determined multiple times from N types of filters;

[0026] Among the detection results corresponding to different combinations of filter elements with the same number of filter elements, the combination of filter elements that achieves the optimal detection result in the second preset detection result is selected as the combination method of filter elements in the filter element assembly; wherein, each filter element combination includes the minimum number of filter elements, and the types and / or arrangements of filter elements in each filter element combination are different.

[0027] According to one aspect of this disclosure, a detection system is provided, the detection system comprising:

[0028] The aforementioned sensing device;

[0029] A light source used to emit detection light;

[0030] A reaction assembly for interacting with the analyte to produce a color change;

[0031] After the probe light emitted by the light source illuminates the reaction component, one or more of the following are obtained: transmitted light, reflected light, or fluorescence, and the light obtained after irradiation is incident on the filter assembly;

[0032] A data processing component is used to obtain detection results based on a rectangular grayscale image array generated by the sensing device using a detection model. The rectangular grayscale image array is input to the detection model to obtain the detection results using the output of the detection model. The detection model has a mapping relationship between the rectangular grayscale image array and the detection results. The detection results include the components of the analyte or the components of the analyte and the content of each component.

[0033] In one possible implementation, the data processing component is further configured to:

[0034] Acquire a first rectangular grayscale image array and a second rectangular grayscale image array output by the sensing device. The first rectangular grayscale image array is the rectangular grayscale image array output by the sensing device when no analyte is added to the reaction component. The second rectangular grayscale image array is the rectangular grayscale image array output by the sensing device when the analyte is added to the reaction component.

[0035] Subtract the intensity of the corresponding pixel from the intensity of the second rectangular grayscale image array and the intensity of the first rectangular grayscale image array to obtain the third rectangular grayscale image array;

[0036] The third rectangular grayscale image array is input into the detection model, and the detection results of the components of the analyte or the components of the analyte and the content of each component are obtained by using the output results of the detection model. The detection model has a mapping relationship between the rectangular grayscale image array and the detection results.

[0037] According to one aspect of this disclosure, a urine detection system is provided, the urine detection system comprising the aforementioned sensor device or the aforementioned detection system.

[0038] In one possible implementation, the urine detection system is used to detect at least one of the following in the urine sample: glucose content, nitrite content, urobilinogen, ketone bodies, bilirubin, protein, red blood cells, white blood cells, and epithelial cells.

[0039] The reaction component of the urine detection system is a reflective component.

[0040] In one possible implementation, the filter assembly in the sensor device is capable of encoding incident light in the range of 450nm-670nm, the number of filters in the filter assembly is 20, and the detection component is made of complementary metal-oxide-semiconductor.

[0041] The sensing device proposed in this disclosure includes a filter assembly for encoding incident light to obtain imaging information for each sensing channel. The imaging information includes the light intensity value of the incident light. The filter assembly includes multiple filters of different types, each filter corresponding to a sensing channel. Different filters can encode incident light to obtain different imaging information. A detection assembly is used to detect the imaging information and generate a rectangular grayscale image array. The rectangular grayscale image array is used as input to a detection model to obtain a detection result using the output of the detection model. The detection model has a mapping relationship between the rectangular grayscale image array and the detection result. The detection result includes the composition of the analyte or the composition of the analyte and the content of each component. The sensing device of this disclosure directly detects the rectangular grayscale image array and inputs it into the detection model to obtain the detection result. It has the advantages of low cost, small size, simple structure, and high accuracy, thus improving measurement accuracy.

[0042] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Other features and aspects of this disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0043] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the specification, serve to illustrate the technical solutions of this disclosure.

[0044] Figure 1 A schematic diagram of a sensing device according to an embodiment of the present disclosure is shown.

[0045] Figure 2 A schematic diagram of the transmission spectrum of various colloidal quantum dot filters according to embodiments of the present disclosure is shown.

[0046] Figure 3 A schematic diagram of a rectangular grayscale image array formed by a detection component according to an embodiment of the present disclosure based on imaging information from a filter component is shown.

[0047] Figure 4a A flowchart illustrating the preprocessing of a rectangular grayscale image array according to an embodiment of the present disclosure is shown.

[0048] Figure 4b A flowchart illustrating a detection model based on the least squares method implemented according to an embodiment of this disclosure is shown.

[0049] Figure 4c A flowchart illustrating a detection process based on a neural network-based detection model according to an embodiment of this disclosure is shown.

[0050] Figure 4dA flowchart illustrating a method for determining the filter of a sensing device according to an embodiment of the present disclosure is shown.

[0051] Figure 5 A schematic diagram of a detection system according to an embodiment of the present disclosure is shown.

[0052] Figure 6 A schematic diagram is shown illustrating the determination of a detection model and the detection of concentration according to an embodiment of the present disclosure.

[0053] Figure 7 A schematic diagram of a filter determination method for a sensing device according to an embodiment of the present disclosure is shown.

[0054] Figure 8 A schematic diagram illustrating the detection of a test object according to an embodiment of the present disclosure is shown. Detailed Implementation

[0055] Various exemplary embodiments, features, and aspects of this disclosure will now be described in detail with reference to the accompanying drawings. The same reference numerals in the drawings denote elements that have the same or similar functions. Although various aspects of the embodiments are shown in the drawings, they are not necessarily drawn to scale unless specifically indicated otherwise.

[0056] In the description of this disclosure, it should be understood that the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this disclosure and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this disclosure.

[0057] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this disclosure, "a plurality of" means two or more, unless otherwise expressly specified.

[0058] In this disclosure, unless otherwise expressly specified and limited, the terms "installation," "connection," "linking," "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this disclosure according to the specific circumstances.

[0059] The term “exemplary” as used herein means “serving as an example, embodiment, or illustration.” Any embodiment illustrated herein as “exemplary” is not necessarily to be construed as superior to or better than other embodiments.

[0060] In this document, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0061] Furthermore, to better illustrate this disclosure, numerous specific details are set forth in the following detailed description. Those skilled in the art will understand that this disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art have not been described in detail in order to highlight the main points of this disclosure.

[0062] Color sensing technologies primarily include visual inspection, RGB image processing, and spectral analysis. Spectral analysis relies on expensive and bulky spectrometers, making it unsuitable for portable, low-cost applications such as clinical testing. Visual inspection and RGB image processing have limited color perception capabilities, affecting identification results. Visual inspection relies on subjective human judgment, especially in clinical medicine, where it depends mainly on the doctor's experience and cannot provide quantitative detection.

[0063] It is evident that the color sensing solutions of related technologies cannot simultaneously achieve aspects such as reducing size, improving accuracy, simplifying structure, and reducing cost.

[0064] This disclosure proposes a sensing device, which includes a filter assembly for encoding incident light to obtain imaging information for each sensing channel. The imaging information includes the light intensity value of the incident light. The filter assembly includes multiple filters of different types, each filter corresponding to a sensing channel. Different filters can encode incident light to obtain different imaging information. A detection assembly is used to detect the imaging information and generate a rectangular grayscale image array. The rectangular grayscale image array is used as input to a detection model to obtain a detection result using the output of the detection model. The detection model has a mapping relationship between the rectangular grayscale image array and the detection result. The detection result includes the composition of the analyte or the composition of the analyte and the content of each component. The sensing device of this disclosure directly detects the rectangular grayscale image array and inputs it into the detection model to obtain the detection result. It has the advantages of low cost, small size, simple structure, and high accuracy, thus improving measurement accuracy.

[0065] Please see Figure 1 , Figure 1 A schematic diagram of a sensing device according to an embodiment of the present disclosure is shown.

[0066] like Figure 1 As shown, the device includes:

[0067] The filter assembly 30 is used to encode the incident light to obtain imaging information of each sensing channel. The imaging information includes the light intensity value of the incident light. The filter assembly includes multiple different types of filters, each filter corresponding to a sensing channel. Different filters can encode the incident light to obtain different imaging information.

[0068] The detection component 40 is used to detect the imaging information and generate a rectangular grayscale image array. The rectangular grayscale image array is used to input into the detection model to obtain the detection result using the output of the detection model. The detection model has a mapping relationship between the rectangular grayscale image array and the detection result. The detection result includes the components of the analyte or the components of the analyte and the content of each component.

[0069] Compared with visual inspection and RGB methods in related technologies, the technical solution for the volume of the sensing device in this disclosure can achieve more accurate color sensing. In the specific color information acquisition process, by acquiring a rectangular grayscale image array instead of reconstructing the spectral curve or color spectral image data, the effect of eliminating the need for calibration and spectral reconstruction can be achieved, simplifying the testing process and requirements.

[0070] The present disclosure does not limit the specific implementation of the filter assembly 30 and the detector assembly 40. Those skilled in the art can adopt appropriate technical means to implement them according to the actual situation and needs. The following is an exemplary description.

[0071] This disclosure does not limit the specific number, type, and arrangement of the filters in the filter assembly 30. Those skilled in the art can adaptively determine the number, type, and arrangement of the filters according to the actual application scenario and needs, thereby further reducing the size, cost, and integration complexity of the filter assembly 30 and improving the detection accuracy.

[0072] In one possible implementation, the filter type may include at least one of various material filter types such as metasurface filter type, photonic crystal filter type, perovskite quantum dot filter type, and colloidal quantum dot filter type. Each filter type includes multiple different types. For example, the filter assembly of this disclosure embodiment may integrate different materials in the same substrate material, that is, a filter assembly may be made of multiple materials, such as integrating perovskite quantum dots, colloidal quantum dots, etc. on the same substrate; of course, it may also be made of one material (such as the filter assembly being made entirely of colloidal quantum dot filters).

[0073] For example, a filter assembly can be composed of multiple (e.g., several, dozens, or hundreds) types of filters. In this embodiment, colloidal quantum dot filters are preferred because the preparation method of colloidal quantum dots is mature, the preparation process is simple, the cost is low, and it is easy to control. Furthermore, multiple colloidal quantum dots can be integrated on the same substrate using ink printing technology to form a filter array (filter assembly). For example, one type of colloidal quantum dot can produce one type of filter. After selecting the type of colloidal quantum dot, multiple colloidal quantum dots can be printed on the same substrate to form a filter assembly (e.g., a filter array including multiple filters). In addition, colloidal quantum dot filters are not affected by the incident light angle, making them particularly suitable for reflective measurement detection systems, and they have significant advantages compared to other filter arrays.

[0074] For example, embodiments of this disclosure prepare colloidal quantum dot filters with different particle sizes by changing the reaction conditions and component ratios during the synthesis of colloidal quantum dots. Different colloidal quantum dot filters have different spectral transmission functions. Colloidal quantum dot filters are simple to prepare, low in cost, easily integrated via liquid-phase printing (such as ink printing), and easily customized for specific applications. Therefore, the detection system of this disclosure can flexibly customize color sensing schemes for different applications, and the device is small in size and low in cost.

[0075] For example, since each filter on the filter assembly has a different spectral transmission function, it can fully sample and encode the incident light, so that the information of each band of the incident light is converted into light intensity information through the filter assembly. In this embodiment of the present disclosure, the light intensity information encoded by the filter assembly can be sampled to form a rectangular grayscale image array, and each type of incident light forms a different rectangular grayscale image array. The rectangular grayscale image array contains the spectral information of the incident light. Compared with traditional large spectrometers, the filter assembly design determined by the filter determination method of this embodiment of the present disclosure avoids the complex structure and large volume of the spectroscopic system, and greatly reduces the volume of the color sensor.

[0076] Please see Figure 2 , Figure 2A schematic diagram of the transmission spectrum of various colloidal quantum dot filters according to embodiments of the present disclosure is shown.

[0077] For example, Figure 2 The 120 colloidal quantum dot filters shown cover a wavelength range of 380nm to 750nm. Each of the 120 colloidal quantum dot filters has a different spectral transmittance function, ensuring that it can encode colors within the 380nm to 750nm range. Each colloidal quantum dot filter encodes the incident light as an intensity value I. i .

[0078] In one possible implementation, the filter array preferably comprises colloidal quantum dot filters, each type of colloidal quantum dot filter having a different spectral transmittance relationship, which represents the correspondence between the spectral transmittance of the filter and the wavelength of light.

[0079] The filter encodes the incident light based on the spectral transmission relationship and the spectral sensitivity relationship of the detection component corresponding to each filter, thereby obtaining the imaging information of the incident light. The spectral sensitivity relationship represents the relationship between spectral responsivity and light wavelength.

[0080] In one possible implementation, the filters encode the incident light based on the spectral transmission relationship and the spectral sensitivity relationship with the detection component corresponding to each filter, to obtain imaging information of the incident light, including:

[0081] The filter encodes the incident light based on the following formula 1 to obtain the imaging information of the incident light:

[0082]

[0083] in, θ represents the spectral transmission relationship of the i-th filter. i (λ) represents the spectral sensitivity relationship between the i-th detector component and each filter, and x(λ) represents the spectrum of the incident light. i Let λ represent the light intensity value corresponding to the i-th filter, λ represent the light wavelength, λ1 represent the minimum wavelength of the light band, and λ2 represent the maximum wavelength of the light band.

[0084] For example, regarding micro-spectrometer technology, It requires a complex, tedious, and costly calibration process to determine, because the spectral reconstruction process requires... The information is missing. However, for the detection scheme proposed in this embodiment, spectral reconstruction is not required, so its value is not known, and calibration is not required to determine it. The value of is such that we only need to know that different colloidal quantum dot filters achieve different encoding results, and then we can obtain the detection result based on the rectangular grayscale image array. For example, we can establish a mapping relationship between the rectangular grayscale image array and the material composition and content, and use this mapping relationship to obtain the detection result based on the rectangular grayscale image array.

[0085] In one possible implementation, the detection component may include at least one of a complementary metal-oxide-semiconductor (CMOS) element, a charge-coupled device (CCD) element, an ultraviolet (UV) detection element, and an indium gallium arsenide (IGaAs) near-infrared (NIIR) detection element. For example, for visible light color sensing applications, the detection component may be a CMOS or a CCD element; for ultraviolet (UV) color sensing applications, the detection component is an UV detection element; and for near-infrared (NIIR) color sensing applications, the detection component is an IGaAs NIIR detection element. Thus, the detection system of this embodiment eliminates the need for an imaging optical lens, further reducing instrument size and cost.

[0086] For example, the detection component acquires image information (light intensity) of the filter component under different colors. The image information can be a grayscale image, which is similar to a barcode distribution. In this embodiment of the disclosure, a rectangular grayscale image array and a pattern recognition algorithm are directly used for qualitative classification and quantitative detection of substances.

[0087] Please see Figure 3 , Figure 3 A schematic diagram of a rectangular grayscale image array formed by a detection component according to an embodiment of the present disclosure based on imaging information from a filter component is shown.

[0088] For example, each colloidal quantum dot filter can cover a rectangular area on the detector component 40, and each rectangular area consists of nearly a hundred pixels. The advantage of using colloidal quantum dots is that their projection function is unaffected by the incident light angle, making them suitable for a wide range of applications, including reflective oblique incidence scenarios. Rectangular grayscale image array technology eliminates the need for imaging lenses, further reducing cost and size compared to other imaging technologies.

[0089] The embodiments disclosed herein acquire a rectangular grayscale image array through the filter assembly 30 and the detector assembly 40 instead of reconstructing spectral curves or color spectral image data. This eliminates the need for calibration and spectral reconstruction processes, reduces processing complexity and cost, simplifies the detection process, and improves measurement accuracy.

[0090] This disclosure does not limit the combination of the filter assembly 30 and the detector assembly 40. Each filter on the filter assembly 30 has a different spectral transmission function, which can fully sample and encode the incident light. This allows the information of each band of the incident light to be converted into light intensity information by the filter assembly, and then sampled by the detector assembly 40 to form a rectangular grayscale image array. For example, each type of incident light forms different imaging information, which includes the spectral information of the incident light and plays a spectral role in representing color changes. Compared with traditional large spectrometers, the design of the filter assembly avoids complex and bulky spectroscopic systems, greatly reducing the size of the color sensor.

[0091] In one possible implementation, the detection model can be used to determine the detection result based on the rectangular grayscale image array and the mapping relationship.

[0092] In one possible implementation, the rectangular grayscale image array can be used as input to a detection model to obtain a detection result using the output of the detection model, wherein the detection model has a mapping relationship between the rectangular grayscale image array and the detection result.

[0093] In one possible implementation, the rectangular grayscale image array may include multiple rectangular regions, each corresponding to a different filter. Each rectangular region corresponds to multiple grayscale values ​​and includes multiple pixels. For example, the grayscale values ​​of the various rectangular regions in the rectangular grayscale image array are different. Of course, the specific shape of the rectangular grayscale image array is not limited in this embodiment. The rectangular grayscale image array can be any shape composed of multiple regions with different grayscale values, each corresponding to a different filter. For example, the rectangular grayscale image array may include multiple rectangular regions arranged in an array of T rows and P columns, where T and P can both be integers greater than 0. The size of each rectangular region can be the same or different. This embodiment does not limit the arrangement of the rectangular grayscale image array, the number of rows, or the number of columns. For example, T can be 8, and P can be 15, meaning a rectangular grayscale image array may include 120 rectangular regions, forming an 8-row, 15-column rectangular grayscale image array.

[0094] This disclosure does not limit the specific type of the test object, the type of detection result, or the specific implementation method of the detection model. Those skilled in the art can determine the test object and select appropriate detection parameters and models based on actual conditions and needs. For example, in one possible implementation, the test object can be a liquid, and the detection result includes components and the content of each component. For instance, the test object can be pesticides, blood, urine, or other liquid test objects, and the detection result can be the composition of the test object and the content of each component, or it can be other classification results. In one possible implementation, the detection result can include the components of the test object or the components of the test object and the content of each component. The detection model is based on at least one of least squares, neural networks, support vector machines, Naive Bayes classification, decision trees, k-nearest neighbors, linear discriminant analysis, linear regression, logistic regression, classification and regression trees, learning vector quantization, bagging, and random forests. This disclosure does not limit the specific methods for establishing and training the detection model; those skilled in the art can use appropriate means to implement it based on actual conditions and needs.

[0095] The sensing device of this disclosure can accurately sense color changes, and the resulting rectangular grayscale image array can be used for qualitative and quantitative identification of substance composition and content. Each part of the sensing device can be customized for specific applications. For example, for pesticide detection, it can acquire liquid-phase absorption color imaging barcodes; for urine component identification, it can acquire imaging barcodes of colors reflected by a paper-based colorimetric array. Specifically, each color corresponds to a unique imaging barcode, and the imaging barcodes combined with pattern recognition algorithms can be used directly to achieve qualitative and quantitative identification of color changes to determine substance composition and content.

[0096] Please see Figure 4a , Figure 4a A flowchart illustrating the preprocessing of a rectangular grayscale image array according to an embodiment of the present disclosure is shown.

[0097] For example, after detecting a rectangular grayscale image array, this embodiment of the present disclosure can preprocess the rectangular grayscale image array using a detection model. The specific method of preprocessing is not limited in this embodiment; those skilled in the art can adopt appropriate preprocessing methods according to actual conditions and needs. For example, in one possible implementation, the detection model is further used to preprocess the rectangular grayscale image array, such as... Figure 4a As shown, the preprocessing method may include at least one of the following:

[0098] Step S11: Calculate the average value of the corresponding pixels of the multiple rectangular grayscale image arrays;

[0099] Step S12: Correct the light intensity non-uniformity of each pixel in the rectangular grayscale image array.

[0100] For example, the detection component 40 can detect multiple rectangular grayscale image arrays obtained by the filter component 30, and the detection model can calculate the average value of the response pixels of the multiple rectangular grayscale image arrays to eliminate random errors caused by accidental factors and improve the stability and accuracy of detection.

[0101] This disclosure does not limit the specific implementation of the light intensity non-uniformity correction method. Those skilled in the art can adopt appropriate correction methods according to actual conditions and needs to correct the light intensity non-uniformity of each pixel in the rectangular grayscale image array.

[0102] As mentioned above, the detection model can be obtained based on at least one of the following: least squares method, neural network, support vector machine, Naive Bayes classification, decision tree, k-nearest neighbor algorithm, linear discriminant analysis, linear regression, logistic regression, classification and regression tree, learning vector quantization, bagging method and random forest. Different types of detection models can be implemented in different ways. This disclosure does not limit the implementation of this model. The following is an exemplary description.

[0103] Please see Figure 4b , Figure 4b A flowchart illustrating a detection model based on the least squares method implemented according to an embodiment of this disclosure is shown.

[0104] In one possible implementation, such as Figure 4b As shown, if the detection model is implemented based on the least squares method, then the detection model is used for:

[0105] Step S211: Calculate the average light intensity of the corresponding rectangular regions of the multiple rectangular grayscale image arrays to obtain multiple average light intensity values;

[0106] Step S212: The multiple average light intensity values ​​are spliced ​​together to obtain an intensity vector;

[0107] Step S213: The intensity vector is used as input to perform least squares operation, and the operation result is used as the detection result, which includes the content of each component.

[0108] This disclosure does not limit the specific method for implementing the detection model based on the least squares method. Those skilled in the art can implement the detection model based on the principle of the least squares method according to the actual situation and needs.

[0109] The detection model based on the least squares method in this embodiment calculates the average light intensity of the corresponding rectangular regions of multiple rectangular grayscale image arrays to obtain multiple average light intensity values. The multiple average light intensity values ​​are then concatenated to obtain an intensity vector. The intensity vector is then used as input for least squares operation, which can quickly obtain the calculation result, thereby obtaining the detection result including the content of each component.

[0110] Please see Figure 4c , Figure 4c A flowchart illustrating a detection process based on a neural network-based detection model according to an embodiment of this disclosure is shown.

[0111] In one possible implementation, such as Figure 4c As shown, if the detection model is implemented based on a neural network, then the detection model can be used for:

[0112] Step S221: Extract the image features of the rectangular grayscale image array;

[0113] Step S222: Perform multiple convolution and fully connected operations on the extracted image features and output the detection results.

[0114] This disclosure does not limit the specific implementation of the detection model based on neural networks. For example, the detection model based on neural networks may include multiple convolutional layers, fully connected layers and other related neural network layers to implement convolutional operations and fully connected operations.

[0115] This disclosure does not limit the specific implementation of step S221 of extracting image features from the rectangular grayscale image array, nor does it limit the specific type of image features. For example, the image features may be the average grayscale intensity of the rectangular grayscale image array or other types of features.

[0116] The detection model based on neural networks in this embodiment extracts image features from the rectangular grayscale image array and performs multiple convolution and fully connected operations on the extracted image features to quickly output detection results. The results include the composition of the analyte and the content of each component.

[0117] This disclosure does not limit the specific training process of the detection model based on the neural network. For example, a training set for model training can be prepared in advance. The training set includes model training data, such as multiple rectangular grayscale image arrays based on the neural network and their corresponding calibration components and component content. In this way, this disclosure can initialize the initial values ​​of each convolution kernel parameter of the neural network, and perform forward computation according to the structure of the neural network using the training set (this disclosure does not limit the specific implementation method of activation computation in forward computation). During the forward computation, it is determined whether the loss function of the neural network has reached a preset value. If so, the structure and parameters of the neural network are saved to obtain the detection model. If the loss function fails to reach the preset value multiple times, the neural network can be back-trained, the weights of each layer of the neural network can be updated, and forward computation can be performed again until the loss function of the neural network reaches the preset value.

[0118] For example, in this embodiment of the present disclosure, a validation set can be prepared in advance. The validation set includes model validation data, such as multiple rectangular grayscale image arrays implemented based on neural networks and their corresponding calibrated components and component contents. In this way, this embodiment of the present disclosure can use the data in the validation set to validate the trained detection model. For example, the rectangular grayscale image array in the validation set is input into the detection model, and the components and component contents output by the detection model are compared with the components and component contents of the corresponding rectangular grayscale image arrays in the validation set. If a large number of comparisons show that the model detection is relatively accurate, the validated detection model can be used for subsequent detection of the test object. If the model detection accuracy is poor, the training set data can be increased for further training to improve the detection accuracy of the detection model.

[0119] The embodiments disclosed herein can optimize the filter of the filter assembly 30 to further simplify the structure and reduce costs.

[0120] Please see Figure 4d , Figure 4d A flowchart illustrating a method for determining the filter of a sensing device according to an embodiment of the present disclosure is shown.

[0121] In one possible implementation, such as Figure 4d As shown, the filter assembly 30 is determined in the following manner:

[0122] Step S31: Select multiple filter chips of different numbers from N types of filter chips to form multiple filter chip assemblies. Each of the N types of filter chips has a different spectral transmission relationship. The N types of filter chips can encode incident light within the target wavelength range. N is a positive integer.

[0123] Step S32: Among the detection results corresponding to filter assemblies with different numbers of filters, select the minimum number of filters from the detection results that reach the first preset detection result, and use the minimum number as the number of filters in the filter assembly. The filter assembly is used to encode incident light into imaging information. The imaging information includes the light intensity value of the incident light. The filter assembly includes multiple different types of filters. Different filters can encode incident light to obtain imaging information.

[0124] Step S33: Determine the minimum number of combinations of continuous or skip distributions from N types of filters multiple times;

[0125] Step S34: Among the detection results corresponding to different combinations of filter elements with the same number of filter elements, select the combination of filter elements that achieves the optimal detection result in the second preset detection results, which is the combination method of the filter elements in the filter element assembly; wherein, each filter element combination includes the minimum number of filter elements, and the types and / or arrangements of the filter elements in each filter element combination are different.

[0126] In one possible implementation, by adaptively determining the number and combination of filters, the present invention can ensure that the number of filters in the filter assembly 30 is the minimum number of detection results that reach a first preset detection result among the detection results corresponding to filter assemblies with different numbers of filters; and that the combination of filters in the filter assembly 30 is a combination that reaches a second preset detection result among the detection results corresponding to different combinations of the same number of filters. The combination method includes the type and arrangement of filters, and each filter combination includes the minimum number of filters. The type and / or arrangement of filters in each filter combination are different.

[0127] This disclosure does not limit the specific form or size of the first preset detection result and the second preset detection result. Those skilled in the art can set them according to actual conditions and needs. For example, the first preset detection result and the second preset detection result can be related to the accuracy of color sensing. In this way, this disclosure can take into account the accuracy, size, cost, etc. of color sensing. For example, the first preset detection result and the second preset detection result can be quantities corresponding to the detection results of the filter assembly. For example, if the detection result is the content of a component, the first preset detection result can be a preset content, or a preset mean square error, root mean square error, etc. If the detection result is a classification result, the first preset detection result can be a preset classification accuracy, etc. For example, the second preset detection result can be a preset boundary line for the better detection result. When the detection result reaches the second preset detection result, it can be determined that the corresponding filter component has a better detection result. For example, there may be multiple values ​​that reach the second preset detection result. In this case, in order to reduce the detection threshold, the embodiments of this disclosure can select the filter component corresponding to the smaller value among the multiple detection results that reach the second preset detection result. Of course, the second preset detection result can also be reasonably set so that the selected filter component has the best detection result.

[0128] This disclosure does not limit the specific method for preparing the filter assembly, nor the specific way of building the sensing device or detection system using the filter assembly. Those skilled in the art can set it according to the actual situation and needs. The following is an exemplary description.

[0129] The embodiments disclosed herein do not limit the specific size of the target wavelength range, which can be determined by those skilled in the art based on actual conditions and needs.

[0130] For example, the spectral transmittance relationship can represent the correspondence between the spectral transmittance of the filter and the wavelength of light.

[0131] For example, after preparing multiple filter assemblies with different numbers of filter elements, embodiments of this disclosure can use each filter assembly to detect the test object and obtain multiple detection results. Among the detection results corresponding to filter assemblies with different numbers of filter elements, the minimum number of filter elements is selected from the detection results that achieve the first preset detection result, and the minimum number is used as the number of filter elements in the filter assembly.

[0132] For example, after determining the number of filters in the filter assembly, this embodiment of the disclosure can determine the minimum number of consecutive or skip distribution combinations from N types of filters multiple times to provide multiple filter assemblies with different filter combination methods. In the detection results corresponding to different filter combinations with the same number of filters, the combination method of the filter combination that reaches the second preset detection result is selected as the combination method of the filters in the filter assembly.

[0133] The following examples illustrate possible implementations of determining the number and combination of filters. It should be understood that the following examples should not be considered as limitations on the embodiments of this disclosure.

[0134] For example, let N be the total number of filters. The filters can be of the colloidal quantum dot type, metasurface structure type, photonic crystal structure type, perovskite quantum dot type, etc. For example, the total number of filters N can range from tens to thousands. Taking colloidal quantum dot filters as an example, this embodiment of the disclosure can synthesize hundreds or thousands of colloidal quantum dots through a chemical synthesis process. Based on the target wavelength of the sensing application, the type and number N of colloidal quantum dots for the target wavelength are determined. Specific methods for synthesizing colloidal quantum dots are not detailed here. For example, if the spectral range of color change is 380nm-750nm, this embodiment of the disclosure can correspondingly synthesize multiple colloidal quantum dots with a transmission rise peak in the 380nm-750nm range.

[0135] For example, in the target wavelength band of the sensing application (e.g., 380nm-750nm), the filters can be arranged from shortest wavelength to longest wavelength according to the transmission rise peak position, with a total of N filters. For instance, according to the arrangement order, 10, 15, 20, 25, 30, ..., N-1 (or other numbers) filters can be selected to form multiple filter assemblies, and imaging information corresponding to the color change of the test object under different numbers of filter assemblies can be obtained, such as a rectangular grayscale image array. A detection model (e.g., a neural network model) is used to classify or fit the imaging information corresponding to different numbers of filter assemblies, such as the rectangular grayscale image array, to obtain the detection results. Relevant evaluation metrics are then used to quantify the classification or fitting effect of the neural network model. For example, the evaluation metrics can be mean square error, root mean square error, classification accuracy, etc.

[0136] For example, assuming the wavelength range of color change of the analyte is 380nm-750nm, and there are 240 types of quantum dot filters within this range, the quantum dot filters can be sorted according to the wavelength of the transmission curve rising from small to large. Following this order, 10, 20, 30, 40, 60, 80, and 120 quantum dot filters are uniformly selected to form various filter assemblies. Different numbers of filter assemblies are used for color measurement of the analyte, acquiring rectangular grayscale image arrays of different components and their contents of the analyte (assuming it is a liquid). A detection model (such as a neural network model) is used for quantitative fitting, and evaluation indicators such as mean square error are used for assessment. The detection model (such as a neural network model) can realize the mapping from the rectangular grayscale image array to the material components and their contents or types. The specific parameters and training process of the detection model (such as a neural network model) can be determined by those skilled in the art according to actual conditions and needs, and will not be elaborated upon in this embodiment.

[0137] For example, embodiments of this disclosure can select a corresponding first preset detection result to determine the minimum number of filters in the filter assembly, such as determining the minimum number M of filters required based on application accuracy requirements. For instance, after obtaining the fitting / classification results of the neural network model under different uniformly distributed filter assemblies, a threshold (first preset detection result) can be set according to the accuracy requirements of the specific color sensing application. The first preset detection result may include classification accuracy, detection limit, etc., and the minimum number of filters required that is greater than the set threshold is determined to be M. For example, the fitting results of each neural network under uniformly distributed filter assemblies of 10, 20, 30, 40, 60, 80, and 120 can be obtained, and the mean square error can be used as the evaluation standard. A mean square error threshold e (first preset detection result) can be set according to the specific application, and the minimum number of filters with a mean square error greater than the threshold e can be selected. Assuming that the minimum number of filters in the filter assembly is M = 30.

[0138] For example, after determining the minimum number of filters in the filter assembly, a number of M filters can be selected from the total number of filters (N types) to determine the combination method of the filters in the filter assembly. For example, it can be a combination method that achieves a second preset detection result among the detection results corresponding to different combinations of filters with the same number of filters.

[0139] In one example, M filters can be selected from a total of N filters, thus determining multiple combinations and forming corresponding filter assemblies. Multiple experiments are then conducted to obtain multiple evaluation metrics for each combination. Since a uniformly distributed set of M filters is not necessarily the optimal result, other combinations such as concentrated continuous distributions and skip distributions are also possible. The step of uniformly selecting M filters is the first step of screening, while the traversal screening is the second step of precise screening. Traversal screening involves significant computational cost. To reduce this, concentrated continuous distributions and uniform distributions can be compared and screened, reducing workload. Alternatively, weights can be assigned based on the main wavelength band of the color change of the analyte, and the filter type can be selected accordingly. For example, assuming there are 240 quantum dot filters (N=240) in the 380nm-750nm wavelength band, the minimum number of filters M=30 requires traversal screening. Due to the large amount of data involved in the computation, this embodiment of the present disclosure can select 30 quantum dot filters in a centralized distribution for screening and comparison. This reduces the number of computations to 210, obtaining the neural network fitting results (i.e., the detection results of the detection model) under 210 consecutive centralized distribution conditions (e.g., 1-30, 2-31, 3-32, ..., 210-239). The neural network fitting results of the 210 cases are compared with the second preset detection results, and the combination of the best evaluation index is selected as the final optimization scheme.

[0140] For example, after selecting M filters and their arrangement, the embodiments of this disclosure can use this combination to guide the fabrication of filter assemblies and color sensors. For instance, after selecting M filters, subsequent mass production of color sensors used for this color sensing application only requires integrating the optimized M filters, instead of integrating N filters, which greatly reduces the number of filters, saves costs, and does not reduce the detection accuracy.

[0141] Of course, for a detailed introduction on the adaptive determination of the number and combination of filters, please refer to the previous description of the detection system, which will not be repeated here.

[0142] In one possible implementation, based on the sensing device proposed in the embodiments of this disclosure, a detection system is proposed to qualitatively and quantitatively identify the composition and content of substances using the sensing device.

[0143] Please see Figure 5 , Figure 5 A schematic diagram of a detection system according to an embodiment of the present disclosure is shown.

[0144] According to one aspect of this disclosure, a detection system is provided, such as Figure 5 As shown, the detection system includes:

[0145] The aforementioned sensing device includes the filter assembly 30 and the detection assembly 40.

[0146] Light source 10 is used to emit detection light;

[0147] Reaction component 20, the reaction component 20 being used to interact with the analyte to produce a color change;

[0148] After the probe light emitted by the light source 10 illuminates the reaction component 20, one or more of the following are obtained: transmitted light, reflected light, or fluorescence, and the light obtained after irradiation is incident on the filter component 30.

[0149] The data processing component 50 is used to obtain a detection result based on the rectangular grayscale image array generated by the sensing device using a detection model. The rectangular grayscale image array is used to input the detection model to obtain the detection result using the output of the detection model. The detection model has a mapping relationship between the rectangular grayscale image array and the detection result. The detection result includes the components of the analyte or the components of the analyte and the content of each component.

[0150] Compared with visual inspection and RGB methods in related technologies, the detection system of this disclosure can achieve more accurate color sensing. In the specific color information acquisition process, by acquiring a rectangular grayscale image array instead of reconstructing spectral curves or color spectral image data, the effect of eliminating the need for calibration and spectral reconstruction can be achieved, simplifying the testing process and requirements.

[0151] In this embodiment, a light source 10 emits probe light to illuminate a reaction component. The light emitted by the light source 10 passes through the reaction component 20 to obtain one or more of the following: transmitted light, reflected light, or fluorescence, and is then incident on a filter assembly 30. The filter assembly 30 encodes the incident light into imaging information, which includes the light intensity value of the incident light. The filter assembly 30 includes multiple filters of different types, and different filters can encode the incident light to obtain different imaging information. The imaging information is detected by a detection component 40 and a rectangular grayscale image array is generated. The detection result is obtained by a data processing component 50 based on the rectangular grayscale image array. This system can achieve accurate color sensing. Furthermore, the system is low-cost, small in size, and the filter array can be optimized according to specific applications to reduce costs and integration complexity, thereby improving measurement accuracy.

[0152] The embodiments disclosed herein do not limit the specific implementation of the light source 10, the reaction component 20, the filter component 30, the detection component 40, and the data processing component 50. Those skilled in the art can adopt appropriate technical means to implement them according to actual conditions and needs. The following is an exemplary description.

[0153] For example, the light source 10 can be customized according to the specific application. For example, for visible light applications, LED light source 10, halogen tungsten lamp light source 10, etc. can be selected, or natural light and other forms of light sources can be used.

[0154] For example, the reaction component 20 includes an analyte. The reaction component 20 can interact with the analyte to produce a detectable color change. When probe light irradiates the reaction component 20, color-related information of the reaction component 20 can be obtained. For example, the reaction component 20 can be filled or loaded with certain reagents that can interact with the analyte to produce a color change. The materials of the reagents include one or more of quantum dot materials, chemical dyes, fluorescent luminescent materials, etc. In one embodiment, the reaction component 20 can be obtained by loading or filling the above-mentioned reagents onto a carrier. The carrier can be set according to actual needs. For example, optional carrier materials include one or more of polytetrafluoroethylene (PTFE), polyvinylidene fluoride (PVDF), polyethylene terephthalate (PET), nylon, non-woven fabric, MCE, PP, etc.

[0155] For example, the reaction component 20 can be a customized placement slot in which the reagent can be placed. Alternatively, in one embodiment, a carrier such as a paper base or film can be placed in the placement slot, and the reagent can be loaded on the carrier. Those skilled in the art can achieve maximum optical efficiency or obtain target optical information by reasonably adjusting the positional relationship between the reaction component 20 and the light source 10 and the detection component 40. Of course, the specific positional relationship between the reaction component 20 and the light source 10 and the detection component 40 is not limited in the embodiments of this disclosure.

[0156] In one possible implementation, the filter type includes at least one of the following: metasurface filter type, photonic crystal filter type, perovskite quantum dot filter type, colloidal quantum dot filter type, etc., and each filter type includes multiple different models.

[0157] In one possible implementation, the detection component 40 may include at least one element such as a complementary metal-oxide-semiconductor element, a charge-coupled device, an ultraviolet detection element, or an indium gallium arsenide near-infrared detection element.

[0158] In one possible implementation, the filter assembly includes colloidal quantum dot filters, each type of colloidal quantum dot filter having a different spectral transmittance relationship, which represents the correspondence between the spectral transmittance of the filter and the wavelength of light.

[0159] The filter encodes the incident light based on the spectral transmission relationship, the spectral sensitivity relationship of the detection component corresponding to each filter, and the spectrum of the incident light to obtain the imaging information of the incident light. The spectral sensitivity relationship represents the relationship between spectral responsivity and light wavelength.

[0160] In one possible implementation, the filters encode the incident light based on the spectral transmission relationship and the spectral sensitivity relationship with the detection component corresponding to each filter, to obtain imaging information of the incident light, including:

[0161] In one embodiment, the filter encodes the incident light based on the following formula to obtain the imaging information of the incident light:

[0162] in, θ represents the spectral transmission relationship of the i-th filter. i (λ) represents the spectral sensitivity relationship between the i-th detector component and each filter, and x(λ) represents the spectrum of the incident light. i Let λ represent the light intensity value corresponding to the i-th filter, λ represent the light wavelength, λ1 represent the minimum wavelength of the light band, and λ2 represent the maximum wavelength of the light band.

[0163] In one possible implementation, the rectangular grayscale image array is a rectangular grayscale image array comprising multiple rectangular regions, each type of filter corresponding to one rectangular region, the multiple rectangular regions corresponding to multiple grayscale values, and each rectangular region comprising multiple pixels. For example, in the above formula... It can be 1, and the corresponding rectangular grayscale image array is a rectangular grayscale image array.

[0164] In one possible implementation, the data processing component 50 includes, but is not limited to, a separate processor, discrete components, or a combination of a processor and discrete components. The processor may include a controller in an electronic device with instruction execution capabilities. The processor may be implemented in any suitable manner, for example, by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components. Within the processor, the executable instructions may be executed by hardware circuitry such as logic gates, switches, ASICs, programmable logic controllers, and embedded microcontrollers.

[0165] In one possible implementation, the data processing component 50 may include a terminal device, a server, or other processing equipment. The terminal device may be a user equipment (UE), mobile device, user terminal, terminal, handheld device, computing device, or in-vehicle device, etc. Examples of terminals include: mobile phones, tablets, laptops, PDAs, mobile internet devices (MIDs), wearable devices, virtual reality (VR) devices, augmented reality (AR) devices, wireless terminals in industrial control, wireless terminals in self-driving, wireless terminals in remote medical surgery, wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities, wireless terminals in smart homes, and wireless terminals in vehicle-to-everything (V2X) networks. For example, the server may be a local server or a cloud server.

[0166] For example, the filter assembly 30 of this disclosure embodiment can be integrated with terminal devices such as smartphones, using natural light or the smartphone's flashlight as the light source 10, or the terminal device's camera as the detection component to achieve a portable color sensing solution. Furthermore, the data processing component 50 can be implemented as software for the terminal device, making operation more convenient and simple. Simultaneously, leveraging the terminal device's Bluetooth and network communication functions, data sharing between multiple devices can be achieved. For instance, any smartphone, when paired with the filter assembly and data processing component 50, can become a color sensor, realizing a portable, low-cost, and compact color sensing solution that meets the needs of various fields such as clinical medicine and environmental monitoring.

[0167] The detection system of this disclosure can accurately sense color changes to qualitatively and quantitatively identify the composition and content of substances. Each part of the detection system can be customized for specific applications. For example, for pesticide detection, it can acquire a rectangular grayscale image array of absorbed colors in liquid phase; for urine component identification, it can acquire a rectangular grayscale image array of reflected colors from a paper-based colorimetric array. Specifically, each color corresponds to a rectangular grayscale image array, and the rectangular grayscale image array combined with a pattern recognition algorithm can be used directly to qualitatively and quantitatively identify color changes to determine the composition and content of substances. Compared to micro-spectrometer technology, whose calibration process relies on expensive instruments and complex experimental procedures, the rectangular grayscale image array technology implemented in this disclosure using filter components and detector components does not require experimental calibration of the filter components and detector components. The rectangular grayscale image array technology is lower in cost, resulting in a correspondingly lower-cost and simpler detection system.

[0168] The following is based on Figure 5 The detection system shown is an example of determining the detection model and performing concentration detection.

[0169] Please see Figure 6 , Figure 6 A schematic diagram is shown illustrating the determination of a detection model and the detection of concentration according to an embodiment of the present disclosure.

[0170] For example, such as Figure 6 As shown, in this embodiment, the incident light is encoded into imaging information by a filter assembly to obtain the light intensity value of each sensing channel. A rectangular grayscale image array is obtained based on the imaging information, and the mapping of the rectangular grayscale image array or imaging information to the material composition and its content is directly realized, thereby achieving qualitative and quantitative detection.

[0171] For example, when detecting substances of different concentrations, the different concentrations of substances cause paper bases and films to exhibit different color changes. The detection system using the embodiments of this disclosure can acquire rectangular grayscale image arrays corresponding to substances of different concentrations, and then use image processing technology and other algorithms for detection.

[0172] For example, such as Figure 6 As shown, to establish a detection model, this embodiment of the disclosure can obtain the average intensity value of each filter region of the rectangular grayscale image array as the light intensity value under the sensing channel of that filter, and then merge them into a high-dimensional vector of the rectangular grayscale image array. Different concentrations of substances correspond to different high-dimensional vectors, and the substance concentration can be fitted using algorithms such as least squares method, neural networks, and various machine learning algorithms to draw a fitting curve, calculate its detection limit (LOD), and obtain the detection model. After the detection model is established, the rectangular grayscale image array obtained by the detection component can be used to determine the unknown concentration.

[0173] Experimental comparisons with RGB image-based methods show that the detection method using the filter assembly obtained in this embodiment significantly improves the fitting effect and reduces the detection limit compared to the RGB image method. This embodiment demonstrates strong color sensing capability. For example,... Figure 6 As shown, the present invention provides a highly accurate method for detecting the composition and content of a solution.

[0174] The following example illustrates the method for determining the filter in a sensing device, using the example of determining the filter assembly in a sensing device for detecting urine.

[0175] Please see Figure 7 , Figure 7 A schematic diagram of a filter determination method for a sensing device according to an embodiment of the present disclosure is shown.

[0176] For example, the glucose content in urine is an important marker for measuring whether or not someone has diabetes and the degree of diabetes, so it is meaningful to develop a low-cost, portable urine glucose detection sensor.

[0177] Figure 7 Figure (1) shows a flowchart of the optimization process for urine glucose detection filters. Exemplarily, this embodiment of the present disclosure uses an array of colloidal quantum dot filters for a urine glucose detection color sensor, for example, using 120 types of colloidal quantum dots (N=120) with transmission rise peaks located in the 380nm-750nm spectral range. However, when fabricating a urine glucose color sensor, simply using 120 types of colloidal quantum dot filters is not feasible because the filter sensing channels are redundant. This redundancy introduces random errors, affecting the detection results of glucose by the detection model formed by pattern recognition methods such as neural networks. Therefore, in this embodiment, multiple filter arrays can be formed by selecting multiple filter arrays of different numbers from N types of filter arrays to provide multiple filter arrays with different numbers of filter arrays. Among the detection results corresponding to filter arrays with different numbers of filter arrays, the minimum number of filter arrays is selected from the detection results that achieve a first preset detection result. This minimum number is used as the number of filter arrays in the filter array. For example, 120 filter arrays can be sorted according to the rising peak from smallest to largest. 20, 30, 40, 60, 90, and 120 uniformly distributed quantum dot filter arrays can be selected to form 7 filter arrays. Further, 7 corresponding color sensors are prepared. The image sensors used are all of the same complementary metal-oxide-semiconductor (CMOS) type, and the same neural network architecture or other pattern recognition method is used to form a detection model for quantitative fitting. In this example, a three-layer fully connected neural network is used, and the mean square error is used as the evaluation index (normalized). The evaluation indices for the 7 color sensors are as follows: Figure 7As shown in (2). The color sensor composed of 120 quantum dot filters is not the optimal result. This is because for glucose sensing, 120 quantum dot filters are redundant. Redundant filters introduce random errors, such as detector noise, which affect the fitting effect. Compared with the sensor composed of 120 filters, the sensor composed of 20 uniformly distributed filters (quantum dot filter numbers are 1, 7, 13, 19, 25, 31, 37, 43, 49, 55, 61, 67, 73, 79, 85, 91, 97, 103, 109, 115; arranged from short wavelength to long wavelength according to the absorption peak position of the filter transmission spectrum) has the best evaluation index. Its evaluation index is nearly an order of magnitude lower than that of the sensor composed of 120 filters. The first step of the filter optimization method proposed in the embodiments of this disclosure reduces the number of quantum dot filters from 120 to 20, improving detection performance (the evaluation index is reduced by nearly an order of magnitude) while reducing costs (cost is reduced to one-sixth). Therefore, for urine glucose sensing, a color sensor composed of 20 different filters is used.

[0178] For example, further, since 20 uniformly distributed filter types are not necessarily the optimal result among all distributions of the 20 filter types, embodiments of this disclosure can determine the optimal distribution of the 20 filter types. For example, the minimum number of consecutive or skip distribution combinations can be determined multiple times from N filter types to provide multiple filter assemblies with different filter combination methods. Among the detection results corresponding to different filter combinations with the same number of filter types, the combination method of the filter combination that reaches the second preset detection result is selected as the filter combination method in the filter assembly. For example, 20 common types are selected from 120 filter types. One selection method involves choosing 20 consecutively distributed filters (1-20, 11-30, ..., 101-120, a total of 11 arrangements) to form a corresponding filter array, which is then used to construct the corresponding color sensor, resulting in 11 color sensors. For example, a detection model constructed using neural network algorithms can be fitted to the detection results of each of the 11 color sensors, using mean squared error (MSE) as the evaluation metric (normalized). Figure 7As shown in (3). The evaluation index of the color sensor with 11 continuously distributed filters is greater than that of the color sensor with uniform distribution. Therefore, in the second step of determining the 20 types of filters, this embodiment of the present disclosure selects 20 types of uniformly distributed filters to reduce the detection limit (LOD). Of course, different optimization results will appear in the second step for different applications. For example, for urine nitrite sensor, in one example, the color sensor composed of continuously distributed 71-90 quantum dot filters has the best performance and the smallest evaluation index. Finally, the color sensor is further prepared by forming an array of the selected 20 uniformly distributed filters (quantum dot filter numbers are 1, 7, 13, 19, 25, 31, 37, 43, 49, 55, 61, 67, 73, 79, 85, 91, 97, 103, 109, 115).

[0179] For example, in transmission color measurement, embodiments of this disclosure may first measure a rectangular grayscale image array without solution; then obtain a rectangular grayscale image array with solution; subtracting the two rectangular grayscale image arrays yields a rectangular grayscale image array of solution absorption.

[0180] Please see Figure 8 , Figure 8 A schematic diagram illustrating the detection of a test object according to an embodiment of the present disclosure is shown.

[0181] In one possible implementation, such as Figure 8 As shown, the data processing component can also be used for:

[0182] Step S41: Obtain the first rectangular grayscale image array and the second rectangular grayscale image array output by the sensing device. The first rectangular grayscale image array is the rectangular grayscale image array output by the sensing device when no analyte is added to the reaction component. The second rectangular grayscale image array is the rectangular grayscale image array output by the sensing device when the analyte is added to the reaction component.

[0183] Step S42: Subtract the intensity of the corresponding pixels in the second rectangular grayscale image array from that in the first rectangular grayscale image array to obtain a third rectangular grayscale image array.

[0184] Step S43: Input the third rectangular grayscale image array into the detection model, and use the output of the detection model to obtain the detection results of the components of the analyte or the components of the analyte and the content of each component, wherein the detection model has a mapping relationship between the rectangular grayscale image array and the detection results.

[0185] This disclosure does not limit the specific type of the analyte, the type of detection result, or the specific implementation method of the detection model. Those skilled in the art can determine the analyte and select appropriate detection parameters and models based on actual conditions and needs. For example, in one possible implementation, the analyte is a liquid, and the detection result includes components and the content of each component. For instance, the analyte can be pesticides, blood, urine, or other liquid analytes, and the detection result can be the composition of the analyte and the content of each component, or it can be other classification results. In one possible implementation, the detection model can be based on at least one of least squares method, neural network, support vector machine, etc. This disclosure does not limit the specific methods for establishing and training the detection model; those skilled in the art can use appropriate means to implement it based on actual conditions and needs.

[0186] In one possible implementation, the reaction component 20 may include a reflective component, a transmissive component, and a fluorescent component, wherein the incident light is any one of the reflected light generated by the reflective component based on the probe light, the transmissive light generated by the probe light penetrating the transmissive component, or the fluorescence generated by the incident light irradiating the fluorescent component.

[0187] In one possible implementation, the filter assembly in the sensing device is capable of encoding incident light in the 380nm-750nm range.

[0188] Compared with visual inspection and RGB methods in related technologies, the detection system of this disclosure can achieve more accurate color sensing. In the specific color information acquisition process, by acquiring a rectangular grayscale image array instead of reconstructing spectral curves or color spectral image data, the effect of eliminating the need for calibration and spectral reconstruction can be achieved, simplifying the testing process and requirements.

[0189] According to one aspect of this disclosure, a urine detection system is provided, the urine detection system comprising the aforementioned sensor device or the aforementioned detection system.

[0190] In one possible implementation, the urine detection system is used to detect at least one of the following in the urine sample: glucose content, nitrite content, urobilinogen, ketone bodies, bilirubin, protein, red blood cells, white blood cells, and epithelial cells.

[0191] The reaction component of the urine detection system is a reflective component.

[0192] In one possible implementation, the filter assembly in the sensor device is capable of encoding incident light in the range of 450nm-670nm, the number of filters in the filter assembly is 20, and the detection component is made of complementary metal-oxide-semiconductor.

[0193] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A sensing device, characterized in that, The device includes: A filter assembly is used to encode incident light to obtain imaging information for each sensing channel. The imaging information includes the light intensity value of the incident light. The filter assembly includes multiple different types of filters, each filter corresponding to a sensing channel. Different filters can encode incident light to obtain different imaging information. A detection component is used to detect the imaging information and generate a rectangular grayscale image array. This rectangular grayscale image array is input to a detection model to obtain a detection result using the output of the detection model. The detection model has a mapping relationship between the rectangular grayscale image array and the detection result. The detection result includes the composition of the analyte or the composition of the analyte and the content of each component. The rectangular grayscale image array comprises multiple rectangular regions arranged in an array of T rows and P columns, where T and P are both integers greater than 0.

2. The apparatus according to claim 1, characterized in that, The detection model is used to determine the detection result based on the rectangular grayscale image array and the mapping relationship.

3. The apparatus according to claim 1, characterized in that, The detection model is also used to preprocess the rectangular grayscale image array, wherein the preprocessing method includes at least one of the following: Calculate the average value of the corresponding pixels in the multiple rectangular grayscale image arrays; The light intensity non-uniformity of each pixel in the rectangular grayscale image array is corrected.

4. The apparatus according to claim 1, characterized in that, The detection model is based on at least one of the following: least squares method, neural network, support vector machine, naive Bayes classification, decision tree, k-nearest neighbor algorithm, linear discriminant analysis, linear regression, logistic regression, classification and regression tree, learning vector quantization, bagging method and random forest.

5. The apparatus according to claim 1, characterized in that, If the detection model is implemented based on the least squares method, then the detection model is used for: The light intensity of the corresponding rectangular regions of multiple rectangular grayscale image arrays is averaged to obtain multiple average light intensity values; The multiple average light intensity values ​​are concatenated to obtain an intensity vector; The intensity vector is used as input to perform a least squares operation, and the result is used as the detection result, which includes the content of each component.

6. The apparatus according to claim 1, characterized in that, If the detection model is based on a neural network, then the detection model is used for: Extract the image features of the rectangular grayscale image array; The extracted image features are subjected to multiple convolution and fully connected operations to output the detection results.

7. The apparatus according to any one of claims 1 to 6, characterized in that, The types of filters include at least one type of metasurface filter, photonic crystal filter, perovskite quantum dot filter, and colloidal quantum dot filter, with each type encompassing multiple different varieties. The detection assembly includes at least one of a complementary metal-oxide-semiconductor element, a charge-coupled device, an ultraviolet detection element, and an indium gallium arsenide near-infrared detection element.

8. The apparatus according to any one of claims 1 to 6, characterized in that, In the filter assembly, the filter is a colloidal quantum dot filter. Each colloidal quantum dot filter has a different spectral transmission relationship. The filter encodes the incident light based on the spectral transmission relationship and the spectral sensitivity relationship of the detection component corresponding to each filter, thereby obtaining the imaging information of the incident light. The spectral sensitivity relationship represents the relationship between photoresponsivity and light wavelength.

9. The apparatus according to any one of claims 1 to 6, characterized in that, The rectangular grayscale image array includes multiple rectangular regions, with each type of filter corresponding to one rectangular region, and each rectangular region includes multiple pixels.

10. The apparatus according to any one of claims 1 to 6, characterized in that, The filter assembly is determined in the following manner: Multiple filter assemblies are formed by selecting a different number of filters from N types of filters. Each of the N types of filters has a different spectral transmission relationship. The N types of filters can encode incident light within the target wavelength range, where N is a positive integer. Among the detection results corresponding to filter assemblies with different numbers of filters, the minimum number of filters is selected from the detection results that reach the first preset detection result, and the minimum number is used as the number of filters in the filter assembly. The filter assembly is used to encode incident light into imaging information, and the imaging information includes the light intensity value of the incident light. The filter assembly includes multiple different types of filters, and different filters can encode incident light to obtain imaging information. The minimum number of combinations of continuous or skip distributions are determined multiple times from N types of filters; Among the detection results corresponding to different combinations of filter elements with the same number of filter elements, the combination of filter elements that achieves the optimal detection result in the second preset detection result is selected as the combination method of filter elements in the filter element assembly; wherein, each filter element combination includes the minimum number of filter elements, and the types and / or arrangements of filter elements in each filter element combination are different.

11. A detection system, characterized in that, The detection system includes: The sensing device as described in any one of claims 1 to 10; A light source used to emit detection light; A reaction assembly for interacting with the analyte to produce a color change; After the probe light emitted by the light source illuminates the reaction component, one or more of the following are obtained: transmitted light, reflected light, or fluorescence, and the light obtained after irradiation is incident on the filter assembly; A data processing component is used to obtain detection results based on a rectangular grayscale image array generated by the sensing device using a detection model. The rectangular grayscale image array is input to the detection model to obtain the detection results using the output of the detection model. The detection model has a mapping relationship between the rectangular grayscale image array and the detection results. The detection results include the components of the analyte or the components of the analyte and the content of each component.

12. The system according to claim 11, characterized in that, The data processing component is also used for: Acquire a first rectangular grayscale image array and a second rectangular grayscale image array output by the sensing device. The first rectangular grayscale image array is the rectangular grayscale image array output by the sensing device when no analyte is added to the reaction component. The second rectangular grayscale image array is the rectangular grayscale image array output by the sensing device when the analyte is added to the reaction component. Subtract the intensity of the corresponding pixel from the intensity of the second rectangular grayscale image array and the intensity of the first rectangular grayscale image array to obtain the third rectangular grayscale image array; The third rectangular grayscale image array is input into the detection model, and the detection results of the components of the analyte or the components of the analyte and the content of each component are obtained by using the output results of the detection model. The detection model has a mapping relationship between the rectangular grayscale image array and the detection results.

13. A urine detection system, characterized in that, The urine detection system comprises the sensing device according to any one of claims 1-10 or the detection system according to any one of claims 11-12.

14. The urine detection system according to claim 13, characterized in that, The urine testing system is used to detect at least one of the following in the urine sample: glucose content, nitrite content, urobilinogen, ketone bodies, bilirubin, protein, red blood cells, white blood cells, and epithelial cells. The reaction component of the urine detection system is a reflective component.

15. The urine detection system according to claim 13, characterized in that, The filter assembly in the sensing device is capable of encoding incident light in the range of 450nm-670nm. The number of filters in the filter assembly is 20. The detection component is made of complementary metal-oxide-semiconductor.

Images