阅读说明：本技术 比色测量分析移动平台设备 (Colorimetric measurement analysis mobile platform equipment ) 是由 刘钢 于 2019-10-24 设计创作，主要内容包括：本发明提供了一种比色测量分析移动平台设备,包括：便携装置、照明控制箱和传感带支架；便携装置包括：相机组件、图像处理组件和处理组件；相机组件捕获测试中的感测条的图像；图像处理组件对相机组件获取的图像进行处理,通过比色分析直接从图像中提取像素信息数据；处理组件对图像处理组件提取的像素信息数据进行校准以便将其转换为当前的照明条件,将校准后的数据拟合到预先计算的线性回归模型中以估计浓度,线性回归模型是利用已知的选定样品浓度基于该选定样品浓度的感测结果预先计算出来的；便携装置和测试条带支架在使用时在封闭的不透光的照明控制箱内相对布置,照明控制箱的外表面具有控制便携装置的相机组件的控制组件。(The invention provides a mobile platform device for colorimetric measurement and analysis, which comprises: the portable device, the lighting control box and the sensing belt bracket; the portable device includes: a camera component, an image processing component and a processing component; the camera assembly captures an image of the sensor strip under test; the image processing component processes the image acquired by the camera component and directly extracts pixel information data from the image through colorimetric analysis; the processing component calibrates the pixel information data extracted by the image processing component to convert it to current lighting conditions, fits the calibrated data to a pre-computed linear regression model to estimate concentration, the linear regression model being pre-computed based on sensing of a selected sample concentration using the known selected sample concentration; the portable device and the test strip holder are arranged in use in opposition within a closed light-tight lighting control box, the outer surface of which has a control assembly that controls the camera assembly of the portable device.)

1. A colorimetric measurement and analysis mobile platform device, comprising: the portable device, the lighting control box and the sensing belt bracket; wherein, portable device includes: a camera component, an image processing component and a processing component; a camera assembly for capturing an image of a sensor strip under test; the image processing component processes the image acquired by the camera component and directly extracts pixel information data from the image through colorimetric analysis; the processing assembly first calibrates the pixel information data extracted by the image processing assembly to convert it to current illumination conditions, fits the calibrated data to a pre-calculated linear regression model to estimate concentration, wherein the linear regression model is pre-calculated based on the sensing of the selected sample concentration using the known selected sample concentration; the portable device and the test strip holder are arranged in use in opposition within a closed light-tight lighting control box, the outer surface of which has a control assembly that controls the camera assembly of the portable device.

2. A colorimetric measurement analysis mobile platform device as claimed in claim 1, wherein the processing component visualizes the estimated concentration for presentation to a user.

3. A colorimetric analytical mobile platform device as claimed in claim 1 or 2, wherein the camera component creates its own camera space that provides information to the user about the current focal length, ISO value and exposure compensation.

4. A colorimetric measurement analysis mobile platform device according to claim 1 or 2 in which the portable device and test strip support are arranged, in use, at opposite ends or sides of the closed lighting control box.

5. A colorimetric analytical mobile platform device according to claim 1 or 2 in which the illumination control box is provided with a reflective material between the position at which the portable device is arranged and the position at which the test strip support is arranged.

6. A colorimetric measurement analysis mobile platform device as claimed in claim 1 or 2, wherein the reflective material is a reflective material in fact formed by a sticker with bubble-like protrusions.

7. A colorimetric measurement analysis mobile platform device according to claim 1 or 2, wherein the housing of the illumination control box has a lying plane inclined at an angle of 30 ° to the side, and the groove of the test strip holder communicates to the opening at the housing of the illumination control box.

8. A colorimetric measurement analysis mobile platform device as claimed in claim 1 or claim 2 in which for ovulation analysis the values of the red, green and luminance channels are selected for use in thresholding the image.

9. A colorimetric analytical mobile platform device as claimed in claim 1 or 2, in which, for glucose estimation, FAST signature detection is used to locate the profile of the strip support.

10. A colorimetric measurement analysis mobile platform device as claimed in claim 1 or 2, wherein the camera component is a smartphone.

Technical Field

The invention relates to the field of colorimetric measurement and analysis, in particular to a mobile platform device for colorimetric measurement and analysis.

Background

Solution concentration sensing, which is directly related to medical diagnosis, is now an important research topic. In this field, many products have been developed for patients at home or medical personnel in clinics to monitor health indicators, such as glucose concentration. The most widely used tools are for example chemical sensor strips, which provide only qualitative results, compared to those sophisticated analytical devices in the laboratory.

For example, traditional colorimetric measurements are widely used for chemical concentration estimation. Traditional colorimetric measurements are only suitable implementations for self-diagnosis at home. Colorimetric measurements provide only limited information about the test chemical solution. However, the general health awareness of the general population is constantly increasing and the need for more sophisticated data for clinical analysis requires a new solution.

In particular, there is a case where, on a sensor strip of a test strip, a plurality of types of sensor arrays each embedding an enzyme for a chemical reaction are disposed. When the array is in direct contact with the target solution, the chemical reaction process causes a change in the color of the top layer of the array. These color changes can be easily seen by the naked eye under normal/indoor lighting conditions. When the reaction is complete, an approximate range of solution concentrations can be read immediately using the reference chart provided. However, it is difficult to accurately quantify the amount of analyte, and thus great improvement is required.

There has been much research on colorimetric sensing. One of the most valuable jobs is the tool that can perform high precision measurements under a variety of environmental conditions to quantify the color of colorimetric diagnostic assays. Rather than using RGB intensities directly, the tool uses colorimetric values extracted from images taken with a smartphone to construct a calibration curve of analyte concentration to overcome the deficiencies of simple RGB analysis, such as low sensitivity to dark colors. This setup requires first a standard color reference chart to calibrate the lighting conditions, and then the RGB values of the point of interest are measured and mapped to the international commission on illumination (CIE)1931 color space for chromaticity value quantification. Finally, it performs regression on the 3D space to calculate the measurements. A good way of performing colorimetric measurements is provided and compensation of lighting conditions is discussed in detail with reference to the graphs; however, this tool still has some drawbacks that prevent its use at the clinical level. First, non-linear regression in 3D space to calculate the measurement is an expensive method, which means that it takes a long time to execute the algorithm. Second, the compensation of the tool for ambient light, using the color reference map to compensate for the position of the light source, light temperature or outdoor lighting environment, because the measured intensity has a linear relationship between different ambient light conditions. However, this solution is not universal. According to the observation of the present inventors, the calibration reference is not necessarily linear unless the illumination conditions include exposure time, exposure rate, ISO sensitivity, and even focal length, etc. Finally, the position and angle of the handheld smartphone camera relative to the test strip has a significant impact on the measurements that are ignored by the tool.

The present invention aims to overcome the above-mentioned drawbacks.

Disclosure of Invention

The technical problem to be solved by the present invention is to provide a mobile platform device for colorimetric measurement and analysis, which is capable of analyzing quantitative data while still retaining the advantages of qualitative methods such as low cost, in view of the above-mentioned drawbacks of the prior art.

According to the invention, the mobile platform equipment for colorimetric measurement and analysis comprises a portable device, an illumination control box and a sensor strip support; wherein, portable device includes: a camera component, an image processing component and a processing component; a camera assembly for capturing an image of a sensor strip under test; the image processing component processes the image acquired by the camera component and directly extracts pixel information data from the image through colorimetric analysis; the processing assembly first calibrates the pixel information data extracted by the image processing assembly to convert it to current illumination conditions, fits the calibrated data to a pre-calculated linear regression model to estimate concentration, wherein the linear regression model is pre-calculated based on the sensing of the selected sample concentration using the known selected sample concentration; the portable device and the test strip holder are arranged in use in opposition within a closed light-tight lighting control box, the outer surface of which has a control assembly that controls the camera assembly of the portable device.

Preferably, the processing component visualizes the estimated concentration for presentation to the user.

Preferably, the camera component creates its own camera space that provides the user with information about the current focal length, ISO value and exposure compensation.

Preferably, the portable device and test strip holder are arranged, in use, at opposite ends or sides of the closed lighting control box.

Preferably, the lighting control box is arranged with a reflective material between the position where the portable device is arranged and the position where the test strip holder is arranged.

Preferably, the reflective material is a reflective material actually formed of a sticker with bubble-like protrusions.

Preferably, the housing of the lighting control box has a placement plane with an inclination angle of 30 ° to the side, and the recess of the test strip holder communicates to the opening at the housing of the lighting control box.

Preferably, for ovulation analysis, the values of the red, green and intensity channels are selected for thresholding the image.

Preferably, for glucose measurements, FAST signature detection is used to locate the profile of the strip scaffold.

Drawings

A more complete understanding of the present invention, and the attendant advantages and features thereof, will be more readily understood by reference to the following detailed description when considered in conjunction with the accompanying drawings wherein:

fig. 1 schematically illustrates a system block diagram of a colorimetric measurement analysis mobile platform device in accordance with a preferred embodiment of the present invention.

Fig. 2 schematically illustrates a block diagram of a portable device of a colorimetric analytical mobile platform device in accordance with a preferred embodiment of the present invention.

Fig. 3 schematically illustrates a top view configuration of a test strip holder of the colorimetric analytical mobile platform device in accordance with a preferred embodiment of the present invention.

Fig. 4 schematically shows a comparison of an original image and an edge detected version thereof.

Fig. 5 shows an example of the maximum profile detected with the four corners marked as red circles.

Fig. 6 shows an example of a detected rectangle.

FIG. 7 shows an example of linear regression of glucose.

Figure 8 shows an example of ovulation linear regression.

It is to be noted, however, that the appended drawings illustrate rather than limit the invention. It is noted that the drawings representing structures may not be drawn to scale. Also, in the drawings, the same or similar elements are denoted by the same or similar reference numerals.

Detailed Description

In order that the present disclosure may be more clearly and readily understood, reference will now be made in detail to the present disclosure as illustrated in the accompanying drawings.

The colorimetric measurement and analysis mobile platform device according to the preferred embodiment of the present invention has two main functions: sensing health indicators (e.g., glucose concentration, etc.) and estimating pregnancy index (e.g., ovulation concentration, etc.).

Fig. 1 schematically illustrates a system block diagram of a colorimetric measurement analysis mobile platform device in accordance with a preferred embodiment of the present invention.

As shown in fig. 1, the colorimetric measurement analysis mobile platform device according to the preferred embodiment of the present invention includes: a portable device 10, an illumination control box 20, and a test strip holder 30.

Fig. 2 schematically illustrates a block diagram of the portable apparatus 10 of the colorimetric analytical mobile platform device in accordance with a preferred embodiment of the present invention.

As shown in fig. 2, the portable device 10 includes a camera assembly 100, an image processing assembly 200, and a processing assembly 300.

Therein, a camera assembly 100 is used to capture an image of a sensor strip under test.

Preferably, the camera assembly 100 is a smartphone. Also, preferably, the camera assembly 100 creates its own camera space (called a surface mount) that can provide the user with more information about the current focal length, ISO value, and exposure compensation, as these parameters have a direct impact on the concentration estimation. It may also prompt the user to enter the name of the photograph in order to more easily track the image data.

The image processing component 200 processes the image acquired by the camera component 100 and extracts pixel information data directly from the image by colorimetric analysis.

For example, the image processing component 200 may use the OpenCV library to provide various feature detection operations to locate regions of interest in the input image and extract pixel information. In the current setup, the position of the test area is fairly consistent in the input image, which means that the image offset can be simply calculated to locate and collect pixel information. However, this is only one possible hard-coded solution as an example, and other solutions may actually be adopted. Therefore, it is very important to automatically identify the attention point by detecting the feature. Furthermore, the method can compensate for small variations in the placement of the stripes due to imperfections in the peripheral design. The data extraction for glucose and ovulation is partially different. The data extracted from the input image will then be sent to the processing component 300 for analysis and visualization.

The processing assembly 300 first calibrates the pixel information data extracted by the image processing assembly 200 to convert it to current illumination conditions, fits the calibrated data to a pre-calculated linear regression model (i.e., mapped to empirical data) that is pre-calculated based on sensing of a selected sample concentration using a known selected sample concentration to estimate the concentration.

Further, the processing component 300 visualizes the estimated concentration for presentation to the user. The visualization facilitates collecting trends for each user based on the test date to draw a trend graph. In a preferred embodiment, the image data generated during the calculation will also be visualized for verification purposes.

The portable device 10 and the test strip holder 30 are disposed in use in opposition within the enclosed, light-tight, lighting control box 20 (e.g., the portable device 10 and the test strip holder 30 are disposed in use at opposite ends or sides of the enclosed lighting control box 20), with the exterior surface of the lighting control box 20 having control components that control the camera assembly 100 of the portable device 10.

It can be seen that the illumination control box 20 isolates light from the camera assembly 100 other than the flash for the test strip.

Preferably, the lighting control box 20 is made of aluminum.

Preferably, the lighting control box 20 is arranged with a reflective material between the position where the portable device 10 is arranged and the position where the test strip holder 30 is arranged. Preferably, the reflective material is a reflective material actually formed of a sticker with bubble-like protrusions. The purpose of these bumps is to increase the reflectivity of the tunnel wall so that it can scatter the illumination of the flash. Because of the attempt to simulate a controlled daylight environment when taking test images, the diffuse flash is critical to reducing the impact of light focusing as a single light model.

In addition to controlling light conditions while taking a picture, another use of the housing is to provide a skeleton for holding the test strip on the test strip holder.

Fig. 3 schematically illustrates a top-view configuration of a test strip holder 30 of the colorimetric analytical mobile platform device in accordance with a preferred embodiment of the present invention. As shown in fig. 3, the side of the test strip holder 30 facing the camera assembly 100 of the portable device 10 is arranged with a groove 31 for mounting a test strip.

Preferably, the housing of the lighting control box 20 has a placing plane with an inclination angle of 30 ° to the side, and the groove 31 of the test strip holder 30 communicates to an opening (the opening can be closed) at the housing of the lighting control box 20, so that when the lighting control box 20 is placed on the ground or a table, a user can smoothly slide the strip holder into the device while keeping the sensing strip free from contamination when placing the strip to be tested on the strip holder because the housing is designed to have an inclination angle of 30 °. All the user needs to do is thus insert the test strip into the peripheral package and then operate the control unit.

Preferably, on the back plate, there is another component for lighting condition calibration. It is a paper holder that supports a color reference chart printed using filter paper. The reason for using the filter paper is that it is very rough on the surface, and thus can reduce the reflection of the flash lamp.

The color reference chart contains a series of color blocks, covering a range of colors from standard RGB to grayscale. When calculating the colorimetric values, it has been used to calibrate the lighting conditions to adjust the linear regression parameters. To calibrate the lighting conditions, it is first known what the true RGB values of black (0,0,0) and white (255 ) are captured by the camera. After extracting this information, the corresponding pixel value of the point of interest may be calculated.

Among the parameters set in the camera interface, there are two worth discussing. First is the picture size parameter. This parameter controls the size or resolution of the image captured by the camera. Theoretically, any image resolution suitable for the mobile phone screen can be defined. In a preferred example, it is set to 3264x 2448. However, another parameter when retrieving an image from external memory is white balance. White balancing is a process of removing unrealistic color shifts, and therefore objects that appear white in a photograph appear white. In a preferred example, white balance may be a key issue in extracting pixel information, leading to identifying deviations in RGB channel values, since it will compensate on some channels. In the procedure, the white balance is selected as follows: automatic, cloudy, daylight, fluorescent, incandescent, shadow, dusk, warm fluorescent. In the present case, automatic white balancing can simply be used to let the camera decide which mode should be applied. The present invention is concerned with the consistency of white balance. Since in the inventive arrangement the only light source is the flash, the automatic white balancing will always invoke the same mode.

By definition, a color space is an organization of colors. Furthermore, it can be seen as a system that reproducibly represents color in an analog and digital manner. Together with the color space, often with reference to an abstract mathematical model, it is described how colors are represented as tuples of numbers as a color model. In this section, the two most widely used primary color spaces of interest will be discussed based on an analysis of the linearity of the different concentration variations.

The first color space is RGB. It is any additive color space based on the RGB color model. In the color model, each color can be represented by specifying a numerical value in each of the three chromaticities of the red, green, and blue additive primaries. Throughout this disclosure we refer to them as red, green and blue channels. In each channel, the tuples range from 0 to 255 therefore, we can represent a total of 256x256x256 — 16777216 different colors. For example, (0,0,0) indicates pure black, and (255,255,255) indicates pure white. The RGB color space is the most commonly used color space in colorimetric analysis. It was noted that changes in glucose concentration had a significant effect on the green channel.

The second color space of interest is HSB (also called HSV). It is the most common cylindrical coordinate representation of points in the RGB color model. In HSB, H represents a hue from 0 to 360, which represents an angle with respect to the color wheel. S represents saturation, describing the distance from full color to gray. B is brightness, telling you how dark the color is.

In Android, the RGB pixel information is typically first read out from the input image and then converted to HSB color space for further analysis. To convert the color space, the following algorithm is generally employed, for example.

procedure RGB2HSB(A)

maxRGB＝MAX(R，G，B)

minRGB＝MIN(R，G，B)

B＝maxRGB

S＝(maxRGB-minRGB)/maxRGB

Δ＝maxRGB-minRGB

if maxRGB＝R then

H＝(G-B)/Δ

end if

if maxRGB＝G then

H＝2+(B-R)/Δ

end if

if maxRGB＝B then

H＝4+(R-G)/Δ

end if

H＝(60*H)mod360

Return H，S，B

end procedure

For ovulation analysis, the color change of the input image is not reasonably linear on any single color channel. Thus, the values of the red, green and luminance channels are selected for thresholding the image.

The main task in detecting glucose in an application is to correctly locate the coordinates of the glucose reagent area. The method performed is to first find the white area of the strip support in the image. Since the strip is always constrained to the strip support and the relative position of one particular reagent zone is fixed, the offset can be used to calculate an approximate zone of glucose reagent. In the future, even if the position of the strip support changes in the setting (as in the image), it can still be detected correctly.

Another reason for choosing the first-time-detection-strip support is that it has a strong color contrast with its neighbourhood. This is useful when we apply Canny edge detection to clean up all sharp edges in an image. The Canny edge detector is an edge detection operator that uses a multi-level algorithm to detect various edges in an image. It is usually performed starting from noise reduction using a 5x5 gaussian filter. It then tries to find the intensity gradient of the image. Finally, it performs non-maximum suppression and hysteresis thresholding to find the best result as an edge. A detailed description of these terms is to be excluded from the present invention because it is a well-established algorithm. In fig. 4, the original image and its edge-detected version are compared. As you see, although we filter and threshold the image, there is still a lot of noise that can confuse our detection. Therefore, the best way to exclude this post-detection noise is to find the largest contour over the entire image.

By careful inspection of the canny edge detection image, it can be seen that most of the edge is open-loop, meaning that it cannot connect back to itself to form a closed loop. The find contours algorithm will eliminate these edges and only preserve the contours of the closed loop. The outline with the largest area is then selected and displayed. The extraction is the profile of the strap holder. Rarely, the algorithm may falsely detect the maximum profile. This is because some parts of the strip holder edge have strong reflections of the smartphone flash when capturing images, which results in the canny edge detector not being able to detect the strip holder profile as closed loop. Retaking the picture can solve this problem.

In actual code, the image needs to be converted into mat type first. The Mat class is the original class in OpenCV, which represents an n-dimensional dense numerical single or multi-channel array. This helps us to process the data in the selected color channel and image more efficiently. To use the canny edge function in OpenCV, the image needs to be mean segmented first. This function essentially outputs a filtered "color-separated" image in which the color gradient and refinement gain texture are flattened. For each pixel (X, Y) of the input image, the function performs the following operations:

(χ，y)：χ-sp≤χ≤χ+sp，Y-sp≤y≤Y+sp，||(R，G，B)-(r，g，b)||≤sr

where sp is the spatial window radius and sr is the color window radius. Empirically, in the example, sp of 10 and sr of 20 yields the best output results. The source is then converted to a grayscale image and the thresholding function is performed on all pixels in the image using the following operations:

it is referred to as binary thresholding because it is set to the declared maximum or to zero depending on whether the data is greater than a threshold. In practice, the maximum value is set to 255 and the threshold value is set to 50. Since the invention is only interested in white swath areas, anything below the threshold will be set to 0 and the step eliminated from the next canny edge detection.

For the canny edge detector, the lower threshold is defined as 60 and the upper threshold is defined as 100, with this setting, the best output image is retrieved. After the contours are accurately detected and found, all contours are stored in the list as Mat points. The list is traversed and the contour with the largest display area is found.

To date, the outline of the strip support has been found and drawn in white on a black background. Feature detection needs to be performed to detect the corners of the profile. There are many algorithms available to detect corners in computer vision, such as SIFT and SURF; however, since the image obtained by the present invention has almost no noise (only white and black), a feature detector called a FAST segment test Feature (FAST) algorithm is used. It is a high-speed corner detection algorithm, but is not robust to high-level noise, which completely conforms to our situation. The basic idea of FAST is to examine a circle of 16 pixels around the pixel under test. It then detects a set of consecutive pixels in the circle, which are all brighter or darker than the threshold. More details about FAST will not be given in the present invention since it is a well-documented algorithm.

In fact, the rectangular profile of the strip support was detected without four sharp corners. Instead, each corner is actually a non-smooth curve. It is clear that FAST corner detection will detect corners on each curve, and the task of the present invention is to find the coordinates that best represent the shape of the contour. Thus, for example, the following procedure may be employed to select these points:

Select Corner Points

procedure SELECT(Points)

Sort(Points.X)

Sort(Points.Y)

TopLeft＝min(Points.X)，min(Points.Y)

TopRight＝max(Points.X)，min(Points.Y)

BotLeft＝min(Points.X)，max(Points.Y)

BotRight＝max(Points.X)，max(Points.Y)

Return TopLeft，TopRight，BotLeft，BotRight

end procedure

basically, all points are sorted by X and Y coordinates and an extremum is selected for each dimension. Once the four corner points are selected, the offset of the glucose reagent area can be measured to extract pixel information. An example of the maximum profile detected with the four corners marked as red circles is shown in fig. 5.

FAST signature detection was used to locate the profile of the strip scaffold. This is sufficient for glucose estimation. However, in order to estimate the ovulation concentration, another rectangular detection step needs to be applied. There is no distinct reagent area on the ovulation strip compared to the glucose-sensing strip. Our goal is to find the left and right boundaries of the stripe to perform a vertical pixel scan along the stripe from top to bottom. This requires us to first locate the coordinates of the boundary correctly. Fortunately, there is a black rectangle on the ovulation strip, which can be easily located using a rectangle detection algorithm, based on which the leftmost and rightmost coordinates can be collected to perform a vertical pixel scan.

The first few steps to locate the band profile are exactly the same as we have for the glucose band. The original image is then cropped and only the strip-holding area is retained. Also, the accurate edge detection is performed to 0 for an image with a low threshold, and the high threshold is set to 255 because the black rectangle sought to be detected has a strong contrast with the adjacent color. The next step involves using a gaussian blur filter to reduce image noise and find contours on the smoothed image. The last step is to examine all the contours to determine if the contours are square. Of all square contours, only those with a width of more than 30 pixels and a width to height ratio of more than 1 are of interest, substantially rectangular.

As soon as the correct rectangle is detected, the vertical scan can be started using the coordinates of the upper left and right corners as boundaries. An example of a detected rectangle is shown in fig. 6.

Image processing using the OpenCV library requires a long time to process. It is desirable that these executions actually occur in the background. In Android OS, the application generates the main thread as a single UI thread, which corresponds to any UI interaction on the phone screen. The camera is started or the image is loaded only as a separate activity, so no other thread is needed. When applying computer vision algorithms (e.g. exact edge detection and FAST feature detection), it usually takes 5 to 10 seconds to process, depending on the image size. If we start these heavy calculations on the main thread, the UI becomes unresponsive. Therefore, the best way to solve this problem is to start another worker thread that processes the computing work in the background and, after completion, merges the results back to the main thread. In this case, the user interface will remain responsive at all times. This technique is called multithreading, and is typically implemented using the AsyncTask or Thread class in the Android interface. For example, all calculations related to image processing may be performed using AsyncTask.

For glucose measurement, the main task is to extract pixel information from the reagent area using an offset. The height/width ratio is first used to find points located within the reagent region and the neighborhood size is defined as the region of interest. Empirically, this pad size was set to 30x30, which contained 900 pixels. As previously described, the green channel of these pixels is linear with glucose concentration due to the enzymatic reaction. Thus, the RGB information for each pixel is accessed and each green channel value is stored.

It is usually simple to take the average of these green values and assign a representative reading to this test for input to the linear regression model. However, this assumes that the solution is evenly distributed over the reagent area. It also assumes that the enzyme is evenly distributed in the reagent zone during the manufacturing process and that there is no over-or under-reaction during the testing process. In practice, this is impractical. Therefore, it is desirable to filter out non-dominant values. For example, for a certain photograph, the 30x30 image that was cropped from the glucose reagent area is of interest. Obviously, there may be a non-negligible number of pixels that are inconsistent with their neighbors. Some of them are light colored, near white, either because of reflection of the flash or because no reaction has taken place at this time. Some of them are dark, nearly black, due to excessive reaction or accumulation of enzymes. They have a deviation in calculating the green channel value representing the average value of the area. Most counts are present in the middle range and fall on high and low green values, such as a gaussian distribution. It was therefore decided to set the experiment-based presence threshold at 15 times, which means that any green values present less than 15 times are considered as a minority and eliminated from the average calculation.

For ovulation measurement, the task is to correctly locate the red line and calculate the number of pixels that meet our criteria. All pixels within the boundary are scanned from top to bottom vertically based on the rectangle located in the last section. For each pixel, the red channel value, the green channel value and the converted luminance value are compared to a threshold value. In this section, there is no skill to select the correct corresponding pixel, only experimentation. First, pixels with high luminance values are excluded from the data set because the main part of the ovulation bar is white and tends to be more reflective than the red line region. The threshold for the luminance channel is set to 0.9. Then, note that if thresholding is performed using only the red channel, some shadows in the image may be a calculation error; thus, the threshold for the red channel is set to 220 and the threshold for the green channel is set to 180. Again, the number of pixels passing the threshold along the vertical scan path is stored and sorted, where the ovulation sample solution (8500mIU/ml) is used for ingestion. Then, the total number of qualified pixels is calculated and subtracted by the "common background". By "common background" is meant pixels that are known to have no response in a region but still not qualify. These pass pixels exist even without control lines or test lines; therefore, they are excluded from the count.

Colorimetric linear regression means that a linear mathematical model is fitted using color information to predict new data. The key part is to find the most reliable linear relationship between the two data sets. A regression baseline is first generated using a solution of known concentration. The Apache general math library was then used to fit a linear regression of the data. For glucose estimation, the green channel value is exposed to solid linearity with concentration; thus, the graphs in table 1 and fig. 7 were obtained by linear regression using the green channel value as the Y data and the concentration as the X data.

TABLE 1 glucose Experimental data

Green channel values were obtained for each concentration by repeated tests using the same setup and different glucose sensor strips. To ensure stable values, the mean and standard deviation of each test listed in table 1 were calculated, and a standard deviation around.5 was a relatively stable reading. For those concentrations with high standard deviations, more tests are necessary to obtain generalized experimental data. However, it still produces a rather good result.

For ovulation estimation, the ratio between the number of qualified pixels in the test line and the number of qualified pixels in the control line is used, instead of using any color channel. The ratio was fitted to the data as Y data and the logarithmic concentration (base 10) as X data to obtain the graphs in linear regression table 2 and fig. 8.

TABLE 2 ovulation test data

The concentration of the ovulation solution tested was 0.85,8.5,85,850,8500 mIU/ml. These solutions were stored at a temperature of 40 ° F and tested within one week of preparation. The results of the linear regression are reasonable because the reaction is so weak for concentrations 0.85 and 8.5 that the red test line is hardly visible to the naked eye.

Subject to the experimental conditions and without the more diverse glucose and ovulation calibration concentrations, it was also observed that the linearity is more convincing in subdivision; therefore, the nearest neighbor method is used for estimation. In glucose testing, the input linear fit range is first calculated. Four data are then selected, which are closest to the input on the fit line. Based on these points, a new linear regression model is constructed from which the final estimate is calculated.

For data visualization, Fragment interfaces are used in Android OS and grapeview open source libraries. The Fragment interface provides a number of convenient functions for formatting a scrollable report page as a tab. The GraphView library supports a variety of graphical layouts and is very easy to customize. For each run, the user needs to provide the name of the current test, which should typically be the owner of the solution under examination. This name serves as the only key in the storage system implemented using the SharePreference interface in Android and is associated with a data list containing concentration value and test date pairs. In visualizing the data, the program will iterate through the corresponding hashmaps to retrieve all values sorted by date tested. They will then be visualized as fragments through the grapeview pipeline. For example, two graphical visualizations may be provided for a user. One is a trend graph showing the concentration tested by date, and the user can click on these points to get more detail. The other is a debugging map, which can display the pixel value in glucose or the pixel response in ovulation test for debugging.

In addition, it should be noted that the terms "first", "second", "third", and the like in the specification are used for distinguishing various components, elements, steps, and the like in the specification, and are not used for representing a logical relationship or a sequential relationship between the various components, elements, steps, and the like, unless otherwise specified.

It is to be understood that while the present invention has been described in conjunction with the preferred embodiments thereof, it is not intended to limit the invention to those embodiments. It will be apparent to those skilled in the art from this disclosure that many changes and modifications can be made, or equivalents modified, in the embodiments of the invention without departing from the scope of the invention. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention are still within the scope of the protection of the technical solution of the present invention, unless the contents of the technical solution of the present invention are departed.