Method and system for correcting inter-instrument variation

Abstract

Embodiments of the present invention relate to systems, methods, and computer program products for correcting for inter-instrument variation between color measurement devices. In one specific embodiment, a method for correcting for inter-instrument variation between color measurement devices includes obtaining a set of color measurements of an item being analyzed. The described method also includes accessing a transformation model, wherein a transformation module is generated using one or more artificial neural networks (ANNs) that backpropagate through a set of data points, wherein the data points correspond to measurements using a color standard from devices from a control device batch and a test device batch. Using the transformation module, a processor is configured to transform the set of color measurements into a calibrated set of color measurements. The calibrated set of color measurements is then output to at least one display, a memory, or a remote computing device.

Description

Technical Field

[0001] The present invention relates to a spectrum measurement device and method for correcting for inter-instrument variations between similar spectrum measurement devices. Background Art

[0002] In the field of color quality control, color is measured using industry-standard colorimeters for the purposes of quality control and color parameter communication. These colorimeter devices are well understood in the art and produce accurate color measurements. When these devices are mass-produced, maintaining product-to-product color consistency between manufacturing units becomes crucial. Specifically, when color or spectral measurement instrumentation is used as part of a product quality control program, each color or spectral measurement device used must produce consistent results, not only from measurement to measurement, but across all measurement devices. For example, it is crucial that two different users can use two different spectrophotometers of the same make and model from the same manufacturer and receive nearly identical color measurements.

[0003] There are several well-known methods and techniques that can be used to determine whether the color measured by a color measurement device is accurate. However, these methods do not provide a mechanism for ensuring that each measurement instrument performs similarly to all other measurement instruments with the same production characteristics.

[0004] Furthermore, various methods exist in the field to calibrate measurement results between different instruments. However, in many cases, traditional calibration methods fail to improve inter-instrument agreement (IIA). Failures in IIA often have no single cause, as the underlying causes are complex and cannot be addressed using traditional methods.

[0005] Furthermore, some color measuring instruments are unable to provide spectral information. Where such instruments are only capable of outputting color tristimulus values ​​(e.g., colorimeters), traditional and conventional techniques for obtaining inter-instrument agreement are inappropriate.

[0006] Therefore, what is needed in the art are systems, methods, and processes for evaluating and compensating for differences in the measurement outputs of different instruments of the same or similar make and model. Furthermore, what is needed in the art are systems, methods, and means for allowing correction of the color tristimulus values ​​output by a measurement device to ensure uniform measurement across different batches of devices produced at different times and / or using different components. Summary of the Invention

[0007] Embodiments of the present invention relate to systems, methods, and computer program products for correcting inter-instrument variation between color measurement devices. In one embodiment, a method for correcting inter-instrument consistency between color measurement devices includes obtaining a set of color measurements of an item being analyzed. The method also includes accessing a transformation model generated using one or more artificial neural networks (ANNs), the ANNs trained on a set of input data obtained from one or more test color measurement devices that share common production parameters with the color measurement devices. Using the transformation model, a processor is configured to transform the obtained set of color measurements into a calibrated set of color measurements by applying the ANN to the ANN. The calibrated set of color measurements is then output to at least one of a display, a memory, or a remote computing device. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The present invention is illustrated in the accompanying drawings, which are provided by way of example and not limitation, in which like reference numerals designate like or corresponding parts, and in which:

[0009] Figure 1 An inter-instrument color measurement system according to one embodiment of the present invention is shown.

[0010] Figure 2A A flowchart of an inter-instrument variation correction system using a neural network according to an embodiment of the present invention is shown.

[0011] Figure 2B A flow chart illustrating certain elements of an inter-instrument variation correction system for training a neural network according to one embodiment of the present invention.

[0012] Figure 2C A flow chart illustrating other specific elements of an inter-instrument variation correction system for training a neural network according to one embodiment of the present invention.

[0013] Figure 3 A block diagram detailing modules used to implement one embodiment of the present invention is shown.

[0014] Figure 4 A diagram detailing a representative node layer arrangement of an artificial neural network is shown.

[0015] Figure 5 A diagram detailing a representative node layer arrangement of an artificial neural network is shown. DETAILED DESCRIPTION

[0016] As an overview, various embodiments of the systems and methods described herein relate to using a trained neural network to correct for differences between measurement devices. In further embodiments, the various embodiments described relate to correcting for inter-instrument variations in instruments of the same make and model but produced under different production conditions.

[0017] By way of further explanation, in order to produce consistent results in operations using multiple color measurement devices, it is crucial that each device of a particular make and model of measurement instrument has inter-instrument measurement consistency. For example, two different users can measure the same color sample using two different spectrophotometers, colorimeters, or other color or spectrum measurement devices of the same model from the same manufacturer, and ideally, the results obtained from the two instruments should be nearly identical.

[0018] However, it is common to see small differences in the measured values ​​of two or more different devices of the same make or model, even if the same components and techniques are used to manufacture the devices, if they are from different production batches. This output difference is true even if the instruments are produced by the same manufacturer in the same factory.

[0019] To correct for these differences, traditional profile-based corrections are often used. For example, the following equation can be used to compensate for small systematic differences between instruments:

[0020] Rci=A+B Rmi+C R'mi+DR”mi.+E Rmi(100-Rmi)(EQN.1)

[0021] where Rmi is the i-th measured reflectivity, Rci is the i-th corrected reflectivity, all Rci, Rmi, and variables A, B, C, D, and E implicitly depend on wavelength, and ' and " represent the first and second derivatives of Rmi with respect to wavelength, respectively. The correction in EQN.1 is represented by the following parameters: offset (A), gain variation (B), wavelength scale variation (C), bandwidth variation (D), and some nonlinear characteristics (E).

[0022] In many cases, this traditional correlation method can improve inter-instrument agreement (IIA). However, in cases where the underlying causes of instrument differences are caused by the complex interaction of components of the color measurement device, traditional methods of measurement correction such as EQN.1 cannot improve the differences between similar instruments.

[0023] Furthermore, this conventional method of correction for differences (such as, but not limited to, that provided in EQN. 1) is not suitable for instruments that provide color tristimulus values ​​as an output of a color measurement process. Devices such as colorimeters require an alternative method to correct for inter-instrument measurement variations.

[0024] Generally speaking, when producing spectral measurement devices (such as color measurement devices), it is important to ensure that the measurements from a given device match some master or control measurement. For example, when manufacturing devices in batches, it is crucial that the devices produced accurately measure color (evaluated relative to a set of known reference color measurements) and achieve a high level of inter-instrument agreement with a control group or batch of devices. In other words, it is important that devices within multiple different production batches produce accurate measurements and that these measurements are within predetermined tolerance levels relative to the output of a control, master, or standard batch of devices.

[0025] As a non-limiting example, a reference batch of measurement devices can be determined by obtaining a group of devices manufactured at the same time that use substantially identical components. For example, a first manufacturing batch of color measurement devices produced using substantially identical components can constitute a first or reference batch. In this context, substantially identical means that all components are of the same make and model, provided by the same supplier, and manufactured by the same manufacturer.

[0026] Each color measurement device produced in the control batch undergoes a quality control analysis. For example, each color measurement device is used to measure a batch of known color references. The resulting measurement values ​​are compared to the reference values, and when the measurement values ​​are within a predetermined range, the devices from that control batch are considered control devices. This process can be repeated for each device produced within the control batch, so that at the end of the process, each measurement device within the control batch has been determined to have the desired level of accuracy.

[0027] As is often the case during the production of complex, multi-component products, components can be changed during the manufacturing process without altering the overall functionality of the device. For example, in a later production batch of color measurement devices, some components (due to a number of potential reasons) were replaced with different but substantially equivalent components. For example, the substantially equivalent components could be of a different make and / or model than those used in a control batch of products, but perform the same function as the control batch components. When measuring known color samples, the control batch measurement devices exhibited low levels of inter-measurement variation and accuracy. Similarly, the measurement devices from this second production batch were also examined for accuracy and inter-measurement variation between members of the second production batch. However, when the members of the second production batch were compared to the average of the control batch, inter-device variation exceeding acceptable levels became apparent.

[0028] Devices of the same make and model, produced with identical functionality but different components, hardware, or software, or produced at different facilities, without any explanation or reason, can introduce complex and subtle differences in the operation of the measurement device, thereby reducing IIA between batches.

[0029] Without correction, when different batches of color measurement devices are used, the measured color output values ​​of these devices will vary significantly. Therefore, in certain embodiments, one or more methods are used to correct for differences found in subsequent production batches relative to a control batch.

[0030] In one particular method, described in greater detail herein, an artificial neural network is used to calibrate measurements taken from a subsequent production batch such that the output measurements produced by the devices of the subsequent production batch have a high degree of IIA relative to the devices of a control batch. For example, both the devices of the control batch and the devices of the subsequent production batch are used to measure a batch of color references. Using an artificial neural network or other machine learning algorithm, a processor can generate one or more models, algorithms, coefficients, or values ​​that, when applied to the measurement outputs of the color measurement devices of the subsequent production batch, result in the measurements being within or within a predetermined threshold range of the measurements taken using the measurement devices of the control batch.

[0031] An artificial neural network (ANN) is used to model the differences in color measurements between different units of the same make and model of color measurement devices using color measurements obtained from a production batch of measurement devices and a control batch of measurement devices. For example, such an ANN is configured to generate a function or set of coefficients that, when applied to any given member of a subsequent batch of color measurement devices, corrects the measurements taken so that the resulting output of the subsequent batch of measurement devices is substantially similar to the output produced by one or more measurement devices from the control batch. In a specific embodiment, using this model, the function or coefficients can be applied to direct color measurements made by the color measurement devices of the subsequent batch to convert the measurements into color measurements made by an average or representative device of the control batch.

[0032] This approach improves the accuracy of each unit in subsequent production batches and represents a technological advancement in the field of color measurement. Specifically, each individual measurement device utilizing the model, or functions or coefficients derived therefrom, is able to provide measurements that are consistent with those made by the instrument of the reference batch.

[0033] In the area of ​​measurement calibration and inter-instrument agreement, current methods are unable to compensate for varying inter-instrument agreement. This is, in part, because each member of a subsequent batch may provide a color measurement within the quality control threshold of a given reference measurement, but such measurements may deviate from those of the control batch. However, simply adjusting the measurement of each individual unit does not actually correct for inter-instrument agreement issues.

[0034] Thus, the presently described systems and methods employ an unconventional, non-traditional approach to the problem of ensuring that devices manufactured under varying production conditions produce consistent measurements across batches. Furthermore, the described systems and methods involve using sample devices from a production batch to train an ANN to generate coefficients of inter-instrument agreement that can be used to ensure that measurements of other members of the production batch are consistent with expected values ​​from a control batch of measurement devices.

[0035] Turning to illustrative examples of the described systems and methods, Figure 1 Detailed description: Color measurement device 102 is configured to obtain measurements of sample 104. The measurements obtained by color measurement device 102 are directly or indirectly transmitted to computer or processor 104 for evaluation and / or further processing. Processor 104, configured by one or more modules stored in memory 105, is configured to transmit the measurements to artificial neural network device 106. In an optional configuration, processor 104 can access one or more coefficients from database 108 for application to the measurements obtained by color measurement device 102, so as to provide updated or corrected color measurements to database 108 or display 110. In one embodiment, the coefficients used to convert measured color values ​​into output color values ​​are stored as a data set in database 108.

[0036] Continue to refer Figure 1 In one embodiment, sample 103 is a color card, a fan plate, a color sample, a product, an article, or an object. For example, sample 103 is a collection of color samples with known color values. In another embodiment, sample 103 is any object whose CIE color value is unknown or needs to be clarified.

[0037] In one embodiment, color measurement device 102 is a colorimeter. In another embodiment, color measurement device 102 is a collection or array of photometers, light-sensing elements, or other similar devices. In another embodiment, color measurement device 102 is one or more cameras or image acquisition devices, such as CMOS (complementary metal oxide semiconductor), CCD (charge coupled device), or other color measurement devices. Such sensors may include data acquisition devices and associated hardware, firmware, and software for generating color values ​​for a given sample. In a specific embodiment, color measurement device 102 is used to generate CIE color values ​​or coordinates for color sample 103. According to one embodiment, color measurement device 102 is a standalone device configured as one or more components, interfaces, or connection structures to one or more processors, networks, or storage devices. In such an arrangement, color measurement device 102 is configured to communicate with associated processors, networks, and storage devices using one or more USB, FireWire, Wi-Fi, GSM, Ethernet, Bluetooth, and other wired or wireless communication technologies suitable for transmitting color, image, spectral, or other related data and / or metadata. In an alternative arrangement, the color measurement device 102 is a component of a smartphone, tablet computer, mobile phone, workstation, test stand, or other computing device.

[0038] In one or more embodiments, the color measurement device 102 is configured to output one or more values ​​when measuring the sample 103. For example, when the color measurement device 102 is a colorimeter, the output generated complies with the CIE color standard. For example, the color measurement device 102 is configured to output L*, a*, and b* values ​​according to the CIE standard. As used herein, CIE color values ​​refer to color coordinates defined by the International Commission on Illumination (CIE), where the L*a*b* color space is modeled according to the color-opponent theory, and where L* represents lightness, a* is the red / green coordinate, and b* is the yellow / blue coordinate.

[0039] In one or more embodiments, the processor 104 is configured, via one or more modules, to receive the CIE values ​​generated by the color measurement device 102. However, in an alternative configuration, the color measurement device 102 is configured to output a raw or unformatted digital or analog signal. In this configuration, the processor 104 is configured to receive the raw or unformatted signal and construct the CIE values ​​therefrom.

[0040] Further references Figure 1The processor 104 is a computing device, such as a commercial microprocessor, processing cluster, integrated circuit, chip computer, or other data processing device. In one or more configurations, the processor is one or more components of a mobile phone, smartphone, laptop, or desktop computer that is configured to receive color measurement data captured by the color measurement device 102 directly or via a communication link. The processor 104 is configured with code executed therein to access various peripheral devices and network interfaces. For example, the processor 104 is configured to communicate with one or more remote servers, computers, peripheral devices, or other hardware over the Internet using standard or custom communication protocols and settings (e.g., TCP / IP, etc.).

[0041] In one configuration, the processor 104 is a portable computing device, such as an Apple or Device, or other commercial mobile electronic device executing a commercial or customized operating system, such as an operating system implementation based on MICROSOFT WINDOWS, APPLE OSX, UNIX or LINUX. In other embodiments, processor 104 is or includes a customized or non-standard hardware, firmware or software configuration. For example, processor 104 includes one or more of a microcomputing element, a chip computer, a home entertainment console, a media player, a set-top box, a prototyping device or a collection of "hobby" computing elements. Depending on the specific embodiment, processor 104 may include a single processor, multiple discrete processors, a multi-core processor or other types of processors known to those skilled in the art.

[0042] In one or more embodiments, the processor 104 is directly or indirectly connected to one or more memory storage devices (memory) to form a microcontroller structure. In addition to one or more software modules 107, the memory is a permanent or non-permanent storage device (such as memory 105) for storing an operating system. According to one or more embodiments, the memory includes one or more volatile and non-volatile memories, such as read-only memory ("ROM"), random access memory ("RAM"), electrically erasable programmable read-only memory ("EEPROM"), phase change memory ("PCM"), single in-line memory ("SIMM"), dual in-line memory ("DIMM"), or other memory types. As known to those skilled in the art, such memory can be fixed or removable, for example, through the use of removable memory cards or modules. In one or more embodiments, the memory of the processor 104 provides storage for application programs and data files. The one or more memories provide program code that the processor 104 reads and executes upon receiving a startup or start signal. The computer memory may also include auxiliary computer memory, such as a magnetic disk drive or optical disk drive or flash memory, which provides long-term storage of data in a manner similar to the permanent memory device 105. In one or more embodiments, the memory 105 of the processor 104 provides storage for applications or modules 107 and data files as needed.

[0043] As shown, memory 105 and permanent storage 108 are examples of computer-readable tangible storage devices. A storage device is any hardware capable of storing information (e.g., data, program code in functional form, and / or other suitable information) on a temporary basis and / or a permanent basis. In one or more embodiments, memory 105 includes random access memory (RAM). According to the present invention, RAM can be used to store data such as venue data. Generally, memory can include any suitable volatile or non-volatile computer-readable storage device. Software and data are stored in permanent storage 108 for access and / or execution by processor 104 via one or more memories of memory 105.

[0044] In one particular embodiment, persistent storage 108 includes a magnetic hard drive. Alternatively, or in addition to the magnetic hard drive, persistent storage 108 may include a solid-state hard drive, a semiconductor memory device, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), flash memory, or any other computer-readable storage device capable of storing program instructions or digital information.

[0045] The database 108 may be embodied as solid-state memory (e.g., ROM), a hard drive system, a RAID, a disk array, a storage area network ("SAN"), a network attached storage ("NAS"), and / or any other suitable system for storing computer data. In addition, the database 108 may include a cache, including a database cache and / or a web cache. Programmatically, the database 108 may include a flat file data store, a relational database, an object-oriented database, a hybrid relational object database, a key-value data store such as HADOOP or MONGODB, in addition to other systems for data structure and data retrieval known to those skilled in the art.

[0046] The media used by persistent storage 108 may also be removable. For example, a removable hard drive may be used for persistent storage 108. Other examples include optical and magnetic disks that are inserted into a drive for transfer to another computer-readable storage medium that is also part of persistent storage 108, thumb drives, and smart cards.

[0047] In these examples, communication or network interface unit 112 provides communication with other subsystems or devices. In one embodiment, communication unit 112 can provide a suitable interface to the Internet or other suitable data communication network to connect to one or more servers, resources, API hosts, or computers. In these examples, communication unit 112 can include one or more network interface cards. Communication unit 112 can provide communication using one or both physical and wireless communication links.

[0048] In one or more embodiments, the display device 106 is a screen, monitor, display, LED, LCD or OLED panel, augmented or virtual reality interface, or electronic ink display device.

[0049] Those having ordinary skill in the necessary technical field will recognize that additional features (such as power supplies, power supplies, power management circuits, control interfaces, relays, interfaces and / or other components for powering and interconnecting electronic components and controlling activation) should be appreciated and understood to be included.

[0050] As described in more detail above, the present systems and methods involve correcting for instrument variation using at least one trained artificial neural network ("ANN"). In one embodiment, the ANN is trained on a set of measurements taken using devices selected from production batches and control batches, such that the trained ANN is capable of providing values, models, or coefficients that convert production batch measurements to control batch measurements. Figure 2A Such a process is described in detail.

[0051] As mentioned above, artificial neural networks ("ANN") are well known in the art. The function of ANN is to perform non-parametric, non-linear, multivariate mapping from one set of variables to another set of variables. For example, Figure 4 The ANN 500 of FIG. 5 illustrates such a mapping by operating on an input vector 502 to produce an output vector 506. To perform this mapping, a training algorithm is applied to derive the input / output relationship from example data. Such ANN training algorithms are also well known in the art.

[0052] Further references Figure 4 , the nodes of the ANN are organized into multiple layers for transforming and weighting input data from a source (e.g., color measurement data generated by one or more color measurement devices) according to one or more algorithms. In certain configurations, the neural network has three (3) or more layers. For example, the neural network implemented or accessed by processor 104 has three (3) layers: an input layer, a hidden layer, and an output layer. In configurations where there are more than three (3) layers, there are typically two (2) or more hidden layers.

[0053] The source data provided to the input layer nodes is distributed to the nodes located in the hidden layer. In one or more configurations, a single input node provides an output value to one or more hidden layer nodes. Therefore, in one embodiment, each hidden layer node is configured to accept multiple input signals. The nodes of the hidden layer combine the input signals pointing to a specific hidden layer node and compare the combined (and weighted) value with the activation function to determine the output signal. The output signal generated by each node in the one or more hidden layers is distributed to one or more nodes in the next layer of the network. Upon reaching the output layer, the output node is configured to accept multiple inputs and generate an output signal or value by using an activation function.

[0054] If a signal propagates from one node to another, then that path can be defined and assigned initial weight values, transforming the input into the output. However, for an ANN to provide useful output values ​​based on the input values, the entire ANN must be trained so that, for a given set of inputs, it provides the expected set of outputs. Training an ANN involves adjusting the weights assigned to the paths connecting the various nodes.

[0055] For example, before training, an ANN is initialized by randomly assigning values ​​to free parameters called weights. The training algorithm takes an unorganized ANN and a set of training input and output vectors and adjusts the weight values ​​through an iterative process. Ideally, at the end of the training process, the presentation of the input vectors from the training data to the ANN results in activations (outputs) at the output layer that exactly match the appropriate training data outputs.

[0056] Figure 5The basic units that make up an artificial neural network are variously referred to as artificial neurons, neurons, or simply nodes. Each ANN node has multiple variable inputs, one (or more) constant weights (also called bias or bias inputs), and an output. In a specific embodiment, the variable inputs correspond to the outputs of the previous nodes in the ANN. Each input of the node (including the bias) is multiplied by the weight associated with that particular input of that particular node.

[0057] Figure 5 An exemplary ANN is shown in FIG, and includes an input layer communicatively connected to one or more hidden layers. The hidden layers can be connected to each other, to the input layer, and to the output layer. All layers are composed of one or more nodes. Thus, data or other information flows from left to right, from each layer to the next adjacent layer.

[0058] When the network is generated or initialized, the weights are randomly set to values ​​close to zero. At the beginning of the ANN training process, as expected, the untrained ANN does not perform the desired mapping well. A training algorithm combined with some optimization technique must be applied to change the weights to provide an accurate mapping. Training is performed in an iterative manner as dictated by the training algorithm. Optimization techniques fall into one of two categories: stochastic or deterministic.

[0059] The selection of training data is often a nontrivial task. An ANN only uses the data used to train it as a representation of the functional map. Any features or characteristics of the map that are not included (or implied) in the training data will not be represented in the ANN. Choosing a good representative sample requires analyzing historical data.

[0060] The training algorithm or neural network structure itself is stored in the neural network device 106 for access by the processor 104. The processor 104 is configured by the program code 107 to evaluate the color value data obtained from the color measurement device 102. In one or more optional configurations, the software module 107 contains appropriate instructions for implementing the training algorithm or neural network structure. For example, the neural network is a software application stored in a memory location locally accessible by the processor 104. Optionally, Figure 1 A neural network is provided (indicated by the dashed line) stored on a remote device 106 accessible by processor 104. In one or more configurations, the neural network is stored or hosted on a remotely accessible storage device (e.g., cloud storage and storage website implementations) that allows for dynamic allocation of additional processors, hardware, or other resources on an "on-demand" or elastic as-needed basis.

[0061] In one or more embodiments, the method includes training an ANN to generate coefficients and / or functions to apply to the CIE colors obtained by measuring devices in a production batch to match the expected output of one or more devices in a reference batch.

[0062] As noted, in one embodiment, the control group of measurement devices refers to the first production batch of devices generated using the same components. It should be understood that this first production batch is also evaluated against one or more calibration tests to ensure that the devices produce the expected measurement results. As long as an individual unit within this first production batch provides an output value within a predetermined threshold of known measurement values, that individual unit is included in the control batch. In another embodiment, the color values ​​of a collection of reference tiles are measured and recorded by each unit in the first production batch. In one embodiment, each unit is used to measure 12 BCRA tiles. For each batch (e.g., the first production batch), an average color value for each of the 12 BCRA tiles is calculated by averaging the values ​​for each tile from all units in the given batch. The 12 average color values ​​are then designated as the virtual color center for the given batch. Each measurement made by each unit in the given batch is compared to this virtual color center to determine the color difference of each unit in the given batch relative to the virtual center of the batch. In further embodiments, unit selection is performed based on the amount of difference between the measurement of any given unit and the measurement of the virtual center. For example, selection of the members of the batch of units includes selecting the 10 units with the color difference closest to the virtual center. These selected units are the units used for ANN training.

[0063] As further shown in step 216, one or more measuring devices in the control group are used to obtain measurements of the known color samples. In one non-limiting embodiment, one or more processors of one or more color measuring devices (such as color measuring device 102) are configured by the ANN control data capture module 316. In one embodiment, as in step 216, an NCS color sector plate having 1950 colors is measured by at least two (2) different measuring devices (e.g., 102) forming the control group. In another embodiment, each unit (i.e., 10 units) in the control batch is used to measure a color reference standard. It should be understood that each measuring device has the same construction and model so that all devices in the control group include substantially the same components. In another embodiment, five (5) different color measurement units (102) selected from the control batch are used to measure the NCS color sector plate.

[0064] continue Figure 2A and Figure 3The ANN control data capture module 316 configures the processor 104 to access the CIE color values ​​(e.g., L*a*b* results) generated by one or more measurement devices in the control batch when measuring each color in the NCS color sector, as shown in step 216. Furthermore, the color values ​​measured by each of the color measurement devices are averaged by one or more of the submodules of the ANN control data capture module 316 to obtain each of the 1950 NCS colors in the color chart. The combined average L*a*b* values ​​generated by each measurement device in the control batch are stored in one or more databases, such as database 108, as a control database or data set by the processor 104.

[0065] Proceeding to step 218, a second group of color measurement devices, referred to herein as a production batch (e.g., subsequently produced color measurement devices 102), are used to measure the same color charts as the control batch. In one embodiment, the second group of color measurement devices has a substantially similar make and model as the color measurement devices of the control batch. However, the members of the production batch were manufactured at different times, from different assemblies, or at different locations. In a particular embodiment, the individual color measurement devices selected to be in the production batch are based on a quality control process. For example, one or more color measurement devices of the production batch are used to measure a reference or calibration standard to ensure that the device functions within an acceptable tolerance. Devices that provide an acceptable level of tolerance are included in the production batch. In a particular embodiment, there are at least two (2) measurement devices in the production group. In another embodiment, there are at least five (5) measurement devices in the production group.

[0066] In a specific implementation, as shown in step 218, one unit of a production batch is used to measure each of the 1950 NCS colors in the fan plate and generate corresponding CIE color values ​​(e.g., L*a*b* results). In this configuration, the production batch only includes one unit. However, in a specific embodiment, the selected member of the production batch is used to measure the color in the fan plate, and the measurement results of the production batch measurement are averaged. In one or more embodiments, one or more submodules of the ANN production data capture module 318 configure one or more processors to obtain the measured value (L*a*b* value) of each color of the fan plate or color reference. In one or more other embodiments, each measured value of the fan plate for evaluation obtained by each measuring device (e.g., one or more color measuring devices) in the production batch is stored in one or more databases. For example, one or more submodules of the ANN production data capture module 318 configure a processor (e.g., processor 104) to store the production batch measurement data in a test database 108 or a data set.

[0067] Going to step 220, the ANN is trained using data obtained from one or more units of the control batch and the production batch. For example, a processor (such as, but not limited to, processor 104) is configured by the ANN training module 320 to instantiate or generate an ANN having input nodes, hidden layers, and output nodes. For example, referring to Figure 5 , one or more submodules of the ANN training module 320 configure the processor to generate an ANN 600 having 3 input nodes, 20 hidden layer nodes, and 3 output nodes. In one or more further embodiments, one or more submodules of the ANN training module 320 configure the processor to generate default W and b parameter values ​​to be adjusted during training.

[0068] continue Figure 2A 、 Figure 3 and Figure 5 In the example provided in FIG, the processor is configured by the ANN training module 320 or a submodule thereof to train the ANN by providing the L*a*b* values ​​obtained from the device of the production batch as input to the ANN. In a specific embodiment, the input value is the average value of the ANN measurement for each color of the NCS color sector plate. For example, the processor is configured to access one or more data values, objects, or files representing data to be input to the ANN 600 from the database 108.

[0069] like Figure 5 As shown, the ANN is configured to generate output values ​​based on input values. However, the output of the ANN 600 can be predetermined so that the output values ​​are assigned to the L*a*b* results stored in the comparison database 108.

[0070] By way of illustration, the ANN training module 320 is configured to set the input values ​​to the ANN to be the average L*a*b* value of each NCS color (making the total input size 3x1950), or the L*a*b* value of each unit stacked together obtained from one or more production batches of color measurement units (so the total input size would be 3x9750, where five (5) measurement devices are used). Likewise, stored color values ​​obtained from one or more control batches of units (making the value three CIE values ​​x 1950 or the same set of control database (size 3x9750) stacked five times) are used.

[0071] According to the ANN training step 220, the processor is configured by the ANN training module 320 to train the ANN. In a specific embodiment, the processor is configured by the ANN training module 320 to train the ANN using one or more forward or backpropagation algorithms. For example, the ANN is trained by the processor configured by the ANN training module 320 to determine appropriate parameter values ​​and weights to convert input values ​​(obtained from the production batch units) into output values ​​(obtained from the control batch units).

[0072] As described herein, the back propagation algorithm is based on a gradient descent training process. In one or more common methods in the art, ANN utilizes forward propagation and back propagation components. In forward propagation, the input information from the input layer is transmitted to be processed via the hidden layer for processing and is eventually transmitted to the output layer. The state of the neurons in each node layer only affects the next layer of neurons. Therefore, when the output value obtained from the ANN training using forward propagation is inconsistent with the expected value (so that the output cannot match the control value when implemented), back propagation is used to calculate the difference between the expected output (that is, the measured value obtained from the control batch unit) and the actual output based on the input (that is, the output obtained by recursively running the measured value obtained from the production batch unit through the node layer by layer).

[0073] Once the ANN has been trained according to step 220, the ANN 600 can be used or accessed to perform live or real-time corrections to measurements taken by any color measurement device for a given production batch. By way of non-limiting embodiment, the ANN or coefficients or models derived from the ANN are made available to a processor in operative communication with the color measurement device 102.

[0074] like Figure 2B and Figure 3 As shown, in a particular embodiment, any device or unit that may be included in a production batch may utilize an ANN or coefficients or models derived therefrom to correct for measurement differences relative to a control group. For example, a processor 104 coupled or communicatively linked to a color measurement device that may be included in the production batch may be configured by code stored in memory 105. As shown in step 204, the processor is configured to capture a color measurement of sample 103. Such code includes one or more software modules that configure processor 104 to instruct color measurement device 102 to capture a color measurement of sample 103. In one or more further embodiments, processor 104 is configured to instruct color measurement device 102 to obtain a CIE color value for sample 103. For example, in step 204, processor 104 is configured by color capture module 302, which initiates a capture routine or process to be executed by color measurement device 102.

[0075] In one non-limiting embodiment, the color capture module 302 causes one or more color sensing elements of the color measurement device 102 to activate or record data corresponding to the analyzed sample 103. In another embodiment, the color capture module 302 sends a capture instruction or flag to one or more processors of the color measurement device, thereby initiating a color measurement process.

[0076] Continuing with step 208, the ANN access module 308 configures the processor 104 to access a trained ANN or models or coefficients derived therefrom. In a particular configuration, the ANN access module 308 configures the processor 104 to access one or more pre-generated or stored analytical models or coefficients generated using an ANN training process (such as, but not limited to, the training process outlined in steps 216 to 220) for a given production batch.

[0077] like Figure 2B As shown, where the ANN training process generates one or more coefficients or functions to be applied to each of the CIE color values, the ANN access module 308 is configured to access these coefficients or functions. In one embodiment, the coefficients or functions accessed by the processor 104 are accessed from the database 108 or other remote storage location.

[0078] Using the accessed ANN-derived coefficients or functions, the measurement conversion module 310 configures the processor 104 to convert the measurements obtained by a given color measurement device 102 for a production batch using the ANN-derived coefficients or functions (as in step 204). Step 210 details the conversion of the initial measurements into corrected measurements. Specifically, given a sufficiently large and representative training data set for the ANN, the ANN-derived coefficients and functions enable the measurements of instruments with characteristics characteristic of the production batch to be compensated or corrected such that the measurements output by the corrected measurement device have a lower degree of inter-instrument variability relative to the control batch. For example, the measurements output by the corrected measurement device are equivalent to the output of the measurement instrument that was part of the control batch.

[0079] After the converted measurement values ​​are generated, the converted measurement values ​​are output to at least one display 110 or data storage device 108, or one or more remote computing devices for further processing. For example, in one embodiment, the converted measurement values ​​and the original measurement values ​​are stored in a local or remote data storage location, as shown in step 212. In one or more further embodiments, the CIE output values ​​are provided on the display 110 or transmitted to a remote device using the communication interface 112.

[0080] As shown in FIG2c , if a new production batch has overlapping similarities with a previous production batch that has been used to train the ANN, the previously used ANN can be used to correct for the new production batch without the need to train a new batch-specific ANN.

[0081] In one specific embodiment, each unit produced in a new, current production batch undergoes a quality control process to check the device's performance. For example, each color measurement device produced in the current production batch is evaluated against a set of reference colors. In one embodiment, the reference colors are ceramic tiles. When measuring the reference colors, the output values ​​are compared to the known values ​​of the reference colors. The difference between the measured and known values ​​determines whether the unit was manufactured correctly. Because the output values ​​of each unit are evaluated against this reference standard, an average measurement value for the entire batch can be calculated. The relative relationship between the average measurement value of the current production batch and the average measurement value of the control group can be considered the measurement signature of that particular production batch. These measurement signatures for these production batches are then stored in one or more databases for further use.

[0082] When a new production batch of color measurement devices is generated, each unit within the batch is evaluated for quality control purposes. The result of this quality control evaluation is the average batch signature for that particular production batch. As shown with respect to step 222, the average batch signature of the new production batch is evaluated relative to the stored production batch signatures used to train the ANN. In another embodiment, one or more production parameters are also searched relative to a batch. Here, the production parameters may include the type and manufacturer of the component, software, hardware modules, production location, and production date. It should be understood that during the ANN training steps 216-220, at least one production parameter of the production batch is stored in the database 108 along with the ANN and all of its coefficients. Therefore, as in step 224, the current production batch signature is used to query the database 108 of the ANN and associated coefficients. In another embodiment, both the production batch signature and the one or more production parameters are used to query the stored database of ANN models and associated coefficients.

[0083] As shown in step 226, where the similarity between the current production batch signature and the stored previous production batch signature is within a predetermined threshold, as in steps 210-212, the identified previous production ANN coefficients or model can be applied to each member of the current production batch to correct future measurements.

[0084] As described in detail herein, the generation of an ANN or its coefficients and its application to a color measurement device results in an increase in inter-batch agreement with a control batch. Thus, for all future batches, the measurement results of any unit in a future production batch can be such that the result has a high degree of inter-instrument agreement with the expected average response of the device in the control batch. Therefore, the described method provides a novel and unconventional application of an ANN for the technical problem of inter-instrument variation. The described method configures a color measurement unit belonging to a production batch through one or more ANN modules, and when measuring the same color, generates an output value substantially the same as that of the device of the control batch, or an expected average value between the devices of the control group. This method demonstrates an improvement in the reliability and accuracy of the measuring device. For example, the described method eliminates the need for the root cause and mechanism of the inter-instrument differences between instruments made of components of the same type but different constructions and models, thereby reducing the difficulty of correcting inter-instrument differences. Since an artificial neural network is a nonlinear system, the described method can compensate for unknown and complex system differences that are not obvious between instruments of the same manufacturer and model. As a result, this particular approach in the field of color measurement achieves inter-instrument agreement without preventing or preempting other approaches to achieving inter-instrument agreement.

[0085] Although this specification contains many specific embodiment details, these details should not be interpreted as limitations on the scope of any embodiment or the scope of what may be claimed, but rather as descriptions of features that may be specific to a particular embodiment. Certain features described in this specification in the context of separate embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented in multiple embodiments individually or in any suitable subcombination. Furthermore, although features may be described above as functioning in certain combinations and even initially claimed as such, in some cases one or more features from a claimed combination may be excised from the combination, and a claimed combination may be directed to subcombinations or variations of subcombinations. In another specific embodiment or example of the methods, systems, and computer-implemented methods described herein, a method for correcting for inter-instrument variation between production batches of color measurement devices comprises: obtaining at least one measurement of a sample target using a sample color measurement device selected from a production batch of color measurement devices; accessing, using a processor configured by code, a transformation module, wherein the transformation module is generated using one or more artificial neural networks (ANNs) trained on a set of production input values ​​obtained from members of the production batch and control output values ​​obtained from members of a control batch of color measurement devices; processing the set of color measurements using the transformation module; obtaining a set of calibration measurements from the set of color measurements; and generating one or more visual indicators of the calibrated color measurements of the sample on a display device. The method of any preceding embodiment or example, wherein the color values ​​are expressed in CIE color coordinate values. The method of any preceding embodiment or example, wherein the color measurement device is a colorimeter. A method according to any preceding embodiment or example, wherein the ANN model is trained using at least one backpropagation algorithm. A method according to any preceding embodiment or example, wherein the ANN model is generated using a control dataset as an output node and a production dataset as an input node. A method according to any preceding embodiment or example, wherein the control data set is generated by obtaining measurements of a plurality of known color standards from a control batch of color measurement devices, wherein each color measurement device is constructed using identical components such that each color measurement device in the control batch generates an output measurement value within a predetermined threshold of an average of the measurements of one or more calibration standards. A method according to any preceding embodiment or example, wherein the number of the plurality of color measurement devices is greater than or equal to 10.A method according to any preceding embodiment or example, wherein the sample color measurement device and the production data set are generated by obtaining measurements of a plurality of known color standards from one or more members of a production batch of color measurement devices, wherein each color measurement device of the production batch is constructed using substantially identical components such that each color measurement device in the production batch generates an output measurement within a predetermined threshold of an average of the measurements of the one or more calibration standards. A method according to any preceding embodiment or example, wherein the number of color measurement devices in the production batch is greater than or equal to 10. A method according to any preceding embodiment or example, wherein the production batch of color measurement devices and the sample color measurement device are constructed using substantially identical components such that each color measurement device in the production batch generates an output measurement within a predetermined threshold of each other.

[0086] In an alternative configuration, a method for correcting for inter-instrument variation is provided, comprising: obtaining a set of color measurements of at least one quality control calibration standard using a plurality of production color measurement devices; generating an average of the measurements of the set of color measurement devices; determining a subset of the set of color measurement devices having minimal color difference from the generated average; obtaining a set of measurements of a training color set using each color measurement device of the measured subset; applying the set of measurements as input to an artificial neural network (ANN), wherein output values ​​are obtained from a control set of measurement devices; generating a conversion coefficient using the ANN; and applying the generated conversion coefficient to at least one measurement made by any of the remaining members of the production set. A method according to any preceding embodiment or example, wherein the plurality of production color measurement devices are all made from the same components. A method according to claim 11, wherein the set of test inputs to the ANN is derived from measurements of a plurality of known color standards made by units of one or more test batches of color measurement devices, wherein each color measurement device of the test batch is constructed using the same components and such test batches have at least one production parameter that differs from the production equipment. A method according to any preceding embodiment or example, wherein the production parameter is one of a production location, a production time, and at least one component used in production.

[0087] In another embodiment, a system for correcting for inter-instrument variation is provided, comprising: a color measurement device selected from a production batch of color measurement devices and configured to measure a color value of a sample; and a processor having a memory and configured to: receive one or more data values ​​corresponding to the measured color value of the sample from the color measurement device; access at least one evaluation model from the memory, wherein the evaluation model receives the color value as input and is generated using one or more artificial neural networks (ANNs) trained on a set of production input values ​​obtained from at least one member of the production batch and control output values ​​obtained from multiple members of a control batch of color measurement devices; evaluate the received data values ​​using the model; obtain an evaluated measurement value; and transmit the evaluated measurement value to at least one of a remote computer, a display device, or a human-perceivable indicator. The system of any preceding embodiment or example, wherein the conversion module is generated using one or more artificial neural networks. The system of any preceding embodiment or example, wherein the artificial networks are trained using at least one back-propagation algorithm.

[0088] Similarly, although operations are depicted in a particular order in the accompanying drawings, this should not be understood as requiring that such operations be performed in the particular order shown or in a sequential order, or that all illustrated operations be performed, in order to achieve the desired results. In certain circumstances, multitasking and parallel processing are advantageous. Furthermore, the separation of various system components in the above-described embodiments should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0089] The terms used herein are for the purpose of describing particular embodiments only and are not intended to limit the present invention. As used herein, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should be further understood that when used in this specification, the terms "comprises" and / or "comprising" specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0090] It should be noted that the use of ordinal numbers such as "first," "second," "third," etc. in the claims to modify claim elements does not itself imply any priority, precedence, or order of one claim element relative to another claim element or the temporal order in which method acts are performed, but merely serves as a label to distinguish one claim element having a certain name from another element having the same name (but used in ordinal terms) to distinguish the claim elements. Furthermore, the words and terms used herein are for descriptive purposes and should not be construed as limiting. The use of "including," "comprising," or "having," "containing," "involving," and variations thereof herein is intended to encompass the items listed thereafter and their equivalents as well as additional items.

[0091] Specific embodiments of the subject matter described in this specification have been described. Other embodiments are within the scope of the following claims. For example, the actions recited in the claims can be performed in a different order and still achieve the desired results. As an example, the processes depicted in the accompanying figures do not necessarily require the specific order shown or sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing may be advantageous.

[0092] Publications and references representing various systems of known registered trademarks are cited in this application, the disclosures of which are incorporated herein by reference. Citation of any of the above publications or documents does not constitute an admission that any of the above is related to the prior art, nor does it constitute any admission of the contents or date of these publications or documents. All references cited herein are incorporated by reference to the same extent as if each individual publication or reference was specifically and individually indicated to be incorporated by reference.

[0093] Although the present invention has been particularly shown and described with reference to the preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made without departing from the spirit and scope of the invention. Therefore, the present invention is not limited by the above discussion, but by the following key points, the corresponding features listed in these key points, and the equivalents of these features.

Claims

1. A method for correcting inter-instrument variation between multiple production batches of a color measurement device, comprising: a. obtaining at least one measurement of a sample target using a sample color measuring device selected from a production batch of color measuring devices; b. accessing, with a processor configured by code, a conversion model, wherein the conversion model is generated using one or more artificial neural networks (ANNs) trained on a set of production input values ​​obtained from one member of the production batch and control output values ​​obtained from a plurality of members of a control batch of color measurement devices; c. processing a set of multiple color measurements using the conversion model; d. obtaining a calibration measurement set from the plurality of color measurement sets; e. generating one or more visual indicators of the calibrated color measurements of the sample on a display device; Wherein the production lot of the color measurement devices and the sample color measurement devices are constructed using substantially identical components such that each color measurement device in the production lot generates output measurements within a predetermined threshold of each other. The method according to claim 1 , wherein the color value is expressed as a CIE color coordinate value. The method of claim 1 , wherein the color measuring device is a colorimeter.

4. The method of claim 1, wherein the artificial neural network model is trained using at least one back-propagation algorithm.

5. The method of claim 4, wherein the artificial neural network model is generated using a control dataset as an output node and a production dataset as an input node.

6. The method of claim 5 , wherein the control data set is generated by obtaining measurements of a plurality of known color standards from a control batch of the color measurement devices, wherein each of the color measurement devices is constructed using identical components such that each color measurement device in the control batch generates an output measurement value within a predetermined threshold of an average of the measurement values ​​of one or more calibration standards. 7 . The method according to claim 5 , wherein the number of the plurality of color measurement devices is greater than or equal to 10.

8. The method of claim 5 , wherein the sample color measurement device and production data set are generated by obtaining measurements of a plurality of known color standards by one or more members of a production batch of color measurement devices, wherein each color measurement device of the production batch is constructed using substantially identical components such that each color measurement device in the production batch generates an output measurement within a predetermined threshold of an average of the measurements of one or more calibration standards.

9. The method of claim 8, wherein the number of color measuring devices in a production batch is greater than or equal to 10.

10. A method for correcting for inter-instrument variation, comprising: a. obtaining a set of multiple color measurements of at least one quality control calibration standard using a plurality of production color measurement devices in a first group; b. generating an average of multiple measurements of a collection of color measurement devices; c. determining a subset of the set of color measurement devices in the first group that has a minimum color difference from the generated average value; d. obtaining a set of measurements of the training color set using each color measurement device of the determined subset; e. applying the set of measurements as input values ​​to an artificial neural network (ANN), wherein the output values ​​are obtained from a control set of measurement devices; generating conversion coefficients using the artificial neural network; f. Applying the generated conversion coefficients to measurements made by the production color measurement devices in the first group but not included in the determined subset.

11. The method of claim 10, wherein each of the plurality of production color measurement devices is constructed from identical components.

12. The method of claim 10 , wherein the set of test inputs to the artificial neural network is derived from measurements of a plurality of known color standards by units of one or more test lots of color measurement devices, wherein each color measurement device of the test lots is constructed using identical components, and wherein such test lots have at least one production parameter that differs from that of the production device.

13. The method according to claim 12, wherein the production parameters are: production location, production time, and at least one component used for production.

14. A system for correcting for inter-instrument variation, comprising: a color measurement device, which is selected from a production batch of color measurement devices, and is configured to measure the color value of the sample; as well as b. A processor having a memory and configured to: i. receiving one or more data values ​​corresponding to the measured color values ​​of the sample from the color measurement device; ii. accessing at least one measurement conversion model from a memory, wherein the measurement conversion model receives a color value as an input and is generated using one or more artificial neural networks (ANNs) trained on a set of production input values ​​obtained from at least one member of the production batch and control output values ​​obtained from a plurality of members of a control batch of color measurement devices; iii. evaluating the received data value using the model; iv. obtaining measurements for evaluation; and v. transmitting said evaluated measurement to at least one of a remote computer, a display device, or a human perceptible indicator.

15. The system of claim 14, wherein the conversion model is generated using one or more artificial neural networks.

16. The system of claim 14, wherein the artificial network is trained using at least one back-propagation algorithm.

Images