Che­mo­me­trics soft­ware is a powerful tool for ana­ly­zing com­plex data sets from various fields of che­mis­try, bio­lo­gy and engi­nee­ring. It can help rese­ar­chers and prac­ti­tio­ners to obtain meaningful infor­ma­ti­on, opti­mi­ze pro­ces­ses and sol­ve pro­blems. In this blog post, we pre­sent some of the fea­tures and bene­fits of the che­mo­me­trics soft­ware and show how it can be used for various applications.

Introduction: What is chemometric software?

Che­mo­me­tric soft­ware packa­ges are indis­pensable tools for ana­ly­ti­cal che­mists who need to pro­cess lar­ge amounts of mul­ti­va­ria­te data pre­cis­e­ly. The­se soft­ware packa­ges uti­li­ze various tech­ni­ques such as line­ar dis­cri­mi­nant ana­ly­sis, mul­ti­va­ria­te data ana­ly­sis and quan­ti­ta­ti­ve ana­ly­sis to pro­vi­de com­pre­hen­si­ve solu­ti­ons for the ana­ly­sis of unknown samples, clas­si­fi­ca­ti­on models and the detec­tion of cri­ti­cal mate­ri­al attri­bu­tes in raw mate­ri­als. In this artic­le, we will explo­re the various appli­ca­ti­ons and bene­fits of che­mo­me­tric soft­ware in dif­fe­rent indus­tries, inclu­ding agri­cul­tu­re, che­mi­cals and pro­duct manu­fac­tu­ring, and final­ly pre­sent our own solution.

Examp­les of con­cre­te appli­ca­ti­ons of che­mo­me­tric data models are…

  1. Pre­dic­tion of qua­li­ty cha­rac­te­ristics of soy pro­ducts: A che­mo­me­tric model can be deve­lo­ped to pre­dict the qua­li­ty cha­rac­te­ristics of soy pro­ducts, such as pro­te­in con­tent, oil con­tent and mois­tu­re con­tent. This model could be used by food manu­fac­tu­r­ers to accu­ra­te­ly mea­su­re the­se cri­ti­cal com­pon­ents in their products.
  2. Detec­tion of impu­ri­ties in medi­cinal pro­ducts: A che­mo­me­tric model can be deve­lo­ped to detect traces of impu­ri­ties in phar­maceu­ti­cals. This model could be used by the phar­maceu­ti­cal indus­try to ensu­re that their pro­ducts com­ply with safe­ty stan­dards and regulations.
  3. Iden­ti­fi­ca­ti­on of off-fla­­vors in bever­a­ges: A che­mo­me­tric model can be deve­lo­ped to iden­ti­fy off-fla­­vors in bever­a­ges. This model could be used by bevera­ge manu­fac­tu­r­ers to redu­ce or eli­mi­na­te off-fla­­vors in their pro­ducts and thus impro­ve cus­to­mer satisfaction.
  4. Iden­ti­fi­ca­ti­on of pla­s­tic poly­mers: A che­mo­me­tric model has been deve­lo­ped to accu­ra­te­ly iden­ti­fy dif­fe­rent types of pla­s­tic poly­mers used in the auto­mo­ti­ve and con­su­mer goods indus­tries. This model can be used to ensu­re that the cor­rect poly­mer is used for a par­ti­cu­lar application.
  5. Coa­ting thic­k­ness mea­su­re­ment: A che­mo­me­tric model can be used to mea­su­re the thic­k­ness of thin films and coa­tings used in the elec­tro­nics and pho­to­vol­taic indus­tries. This model can be used to ensu­re that the lay­ers are within the requi­red tolerances.
  6. Esti­ma­ti­on of the vege­ta­ti­on index: A che­mo­me­tric model was deve­lo­ped to esti­ma­te vege­ta­ti­on indi­ces from satel­li­te images. This model can be used for remo­te sens­ing appli­ca­ti­ons as well as for moni­to­ring agri­cul­tu­ral crops and land use patterns.
  7. Clas­si­fi­ca­ti­on of feed ingre­di­ents: A che­mo­me­tric model can be used to clas­si­fy feed ingre­di­ents in ani­mal feed. This model can be used to pre­cis­e­ly iden­ti­fy the dif­fe­rent types of ingre­di­ents in a par­ti­cu­lar feed mixture.
  8. Metro­lo­gy: A che­mo­me­tric model could be deve­lo­ped for metro­lo­gi­cal appli­ca­ti­ons such as sur­face pro­fil­ing, dimen­sio­nal mea­su­re­ments and 3D scan­ning. This model can be used to ensu­re accu­ra­te mea­su­re­ment and con­trol of cri­ti­cal com­pon­ents in manufacturing.

A gra­phi­cal user inter­face is an essen­ti­al part of che­mo­me­tric soft­ware packa­ges as it pro­vi­des users with an intui­ti­ve and user-fri­en­d­­ly inter­face. The­se inter­faces allow users to easi­ly navi­ga­te the soft­ware, ana­ly­ze and visua­li­ze com­plex data sets and opti­mi­ze their work­flows. In addi­ti­on, the­se inter­faces allow ana­lysts to pre-pro­cess their data, sel­ect the spec­tral ran­ge and per­form mul­ti­va­ria­te ana­ly­ses with mini­mal effort.

Dif­fe­rent ana­ly­ti­cal methods have dif­fe­rent spec­tral ran­ges, noi­se rati­os and stan­dard errors, and che­mo­me­tric soft­ware can effec­tively ana­ly­ze data regard­less of the source method. The­se soft­ware tools can ana­ly­ze spec­tra such as deri­va­ti­ve spec­tra, libra­ry spec­tra, mass spec­tra, NIR spec­tra, visi­ble spec­tra and absorp­ti­on spectra.

In sum­ma­ry, che­mo­me­tric soft­ware packa­ges are powerful tools that can assist in the deve­lo­p­ment of com­plex models for data pro­ces­sing, mul­ti­va­ria­te data ana­ly­sis and the iden­ti­fi­ca­ti­on of cri­ti­cal mate­ri­al pro­per­ties. With an intui­ti­ve gra­phi­cal user inter­face and the fle­xi­bi­li­ty to work with dif­fe­rent ana­ly­ti­cal methods, the­se tools make che­mo­me­tric ana­ly­sis acces­si­ble to a wider ran­ge of users.

What are typical applications for NIR spectroscopy?

NIR spec­tro­sco­py is a powerful instru­ment that is used in various fields and indus­tries for the non-inva­­si­­ve and non-des­­truc­­ti­­ve ana­ly­sis of samples. The sec­tors in which NIR spec­tro­sco­py is typi­cal­ly used for ana­ly­sis include agri­cul­tu­re, phar­maceu­ti­cals and the food and bevera­ge industry.

Howe­ver, the appli­ca­ti­on of NIR spec­tro­sco­py is not limi­t­ed to the­se indus­tries. It is also used in the pla­s­tics indus­try, par­ti­cu­lar­ly for recy­cling and pro­cess con­trol. The tech­no­lo­gy offers con­sidera­ble advan­ta­ges for the­se appli­ca­ti­ons as it is fast and can pro­vi­de pre­cise infor­ma­ti­on on the che­mi­cal com­po­si­ti­on of raw materials.

One of the main advan­ta­ges of NIR spec­tro­sco­py is its abili­ty to quick­ly pro­vi­de infor­ma­ti­on on cri­ti­cal mate­ri­al pro­per­ties such as the che­mi­cal com­po­si­ti­on of raw mate­ri­als. This makes it an indis­pensable tool for indus­tries whe­re under­stan­ding the com­po­si­ti­on of mate­ri­als is crucial.

NIR spec­tro­sco­py is often used during inspec­tion to deter­mi­ne pro­duct strength in the phar­maceu­ti­cal indus­try. By ana­ly­zing the NIR spec­tra of the pro­duct, it is pos­si­ble to deter­mi­ne its strength and con­sis­ten­cy, for exam­p­le. It is also used to ana­ly­ze the com­pon­ents of agri­cul­tu­ral pro­ducts such as corn chaff samples. It is used in the food and bevera­ge indus­try to moni­tor fer­men­ta­ti­on processes.

NIR spec­tro­me­ters are sin­g­le-pixel came­ras in the NIR wave­length ran­ge. With the right tech­no­lo­gy, they can be built very com­pact­ly. Many NIR spec­tro­me­ters are built wit­hout optics, i.e. the mea­su­re­ments must be car­ri­ed out by contact.

In sum­ma­ry, NIR spec­tro­sco­py offers a wide ran­ge of appli­ca­ti­ons in indus­tries such as agri­cul­tu­re, phar­maceu­ti­cals, food and bever­a­ges and pla­s­tics. Due to its non-inva­­si­­ve and non-des­­truc­­ti­­ve ana­ly­sis capa­bi­li­ties, it is par­ti­cu­lar­ly useful for deter­mi­ning the che­mi­cal com­po­si­ti­on of raw mate­ri­als and for con­trol pur­po­ses, e.g. in the phar­maceu­ti­cal indus­try. It can also be used to ana­ly­ze the com­pon­ents of agri­cul­tu­ral pro­ducts and to moni­tor fer­men­ta­ti­on pro­ces­ses in the food and bevera­ge industry.

What is hyperspectral imaging?

Hyper­spec­tral ima­ging (HSI) is a powerful ana­ly­ti­cal tech­ni­que that enables the iden­ti­fi­ca­ti­on and cha­rac­te­riza­ti­on of a sample’s com­po­si­ti­on and struc­tu­re by ana­ly­zing its spec­trum across many nar­row wave­length bands. In hyper­spec­tral ima­ging, light is pas­sed through or reflec­ted from a sam­ple and the resul­ting spec­trum is recor­ded with a sui­ta­ble detec­tor. This spec­trum con­ta­ins infor­ma­ti­on about the che­mi­cal com­po­si­ti­on of the sam­ple, which can be used to crea­te a detail­ed pic­tu­re of its struc­tu­re and properties.

Hyper­spec­tral ima­ging has a wide ran­ge of appli­ca­ti­ons in che­mo­me­trics. One of the most important appli­ca­ti­ons is the ana­ly­sis of agri­cul­tu­ral pro­ducts, whe­re it is used to deter­mi­ne the qua­li­ty, com­po­si­ti­on and nut­ri­tio­nal value of plants such as fruit and vege­ta­bles. Hyper­spec­tral ima­ging can also be used to check the qua­li­ty of finis­hed pro­ducts such as phar­maceu­ti­cals, cos­me­tics and food to iden­ti­fy cri­ti­cal mate­ri­al pro­per­ties that could affect their per­for­mance or safety.

One of the main advan­ta­ges of hyper­spec­tral ima­ging is that it enables non-inva­­si­­ve and non-des­­truc­­ti­­ve ana­ly­sis of a wide ran­ge of samples, inclu­ding solids, liquids and gases. This makes them ide­al for use in indus­tries whe­re the inte­gri­ty of samples is cru­cial, such as the phar­maceu­ti­cal, food and cos­me­tics indus­tries. Hyper­spec­tral ima­ging can also be used to ana­ly­ze samples in situ wit­hout the need for sam­ple pre­pa­ra­ti­on or mani­pu­la­ti­on, saving time and redu­cing ana­ly­sis costs.

HSI came­ras are available in the visi­ble, near-infrared and mid-infrared wave­length ran­ges. They are equip­ped with eit­her 320 or 640 pixels and ope­ra­te at a fre­quen­cy of approx. 300 Hz to cap­tu­re spec­tra over the enti­re ran­ge. The­se came­ras can be used con­ti­nuous­ly and are sui­ta­ble for inline appli­ca­ti­ons. Optics and light­ing should be sel­ec­ted careful­ly, as HSI came­ras mea­su­re at a distance of 20 cm to 50 cm.

In sum­ma­ry, hyper­spec­tral ima­ging is a valuable tool for che­mo­me­tri­ci­ans as it allows the com­po­si­ti­on and struc­tu­re of a sam­ple to be ana­ly­zed across many wave­length bands. Their abili­ty to ana­ly­ze a wide ran­ge of sam­ple types non-inva­­si­­ve­­ly and non-des­­truc­­tively makes them useful for a varie­ty of appli­ca­ti­ons, such as agri­cul­tu­ral pro­duct ana­ly­sis, qua­li­ty con­trol and iden­ti­fi­ca­ti­on of cri­ti­cal mate­ri­al properties.

What is the chemotmetric software used for?

The data from NIR spec­tro­me­ters con­tain a lot of infor­ma­ti­on, and this is even more true for the data from HSI came­ras. Wit­hout che­mo­me­tric soft­ware, you can­not per­form data ana­ly­sis for the clas­si­fi­ca­ti­on and quan­ti­ta­ti­ve ana­ly­sis of unknown samples. This soft­ware enables ana­ly­ti­cal che­mists to deve­lop mul­ti­va­ria­te models, such as mul­ti­va­ria­te cali­bra­ti­on models and clas­si­fi­ca­ti­on models. The soft­ware has a user-fri­en­d­­ly gra­phi­cal inter­face that faci­li­ta­tes the pre-pro­ces­­sing of spec­tral data in a wide ran­ge of spec­tral ran­ges and noi­se ratios.

The most important criteria for finding the best chemometrics software for your application

If you work in the field of ana­ly­ti­cal che­mis­try, you alre­a­dy know how important it is to have the right soft­ware tools to help you ana­ly­ze and inter­pret your data accu­ra­te­ly. Che­mo­me­tric soft­ware is a powerful tool that can help you pro­cess and ana­ly­ze lar­ge amounts of data to iden­ti­fy pat­terns, clas­si­fy samples and crea­te models. Howe­ver, with so many che­mo­me­tric soft­ware packa­ges available on the mar­ket, choo­sing the right soft­ware can be a dif­fi­cult task. In this artic­le, we will give you some insights into the most important cri­te­ria for fin­ding the best che­mo­me­trics soft­ware for your application.

Your experience

Your level of expe­ri­ence with che­mo­me­trics soft­ware can influence your choice. If you are fami­li­ar with pro­gramming lan­guages such as Python, you should use them to ana­ly­ze che­mo­me­tric data. Howe­ver, if you are not fami­li­ar with pro­gramming, you should use soft­ware with a gra­phi­cal user inter­face (GUI), which is more user-fri­en­d­­ly and easier to operate.

Required functions: PCA, LDA, PLS and more?

Func­tions: The func­tions offe­red by the soft­ware can also influence your choice. Some soft­ware is bet­ter sui­ted to spe­ci­fic types of ana­ly­sis, such as mul­ti­va­ria­te data ana­ly­sis, while others offer a wider ran­ge of fea­tures that can be used for a varie­ty of applications.

Offline analysis or inline processing?

Befo­re pro­cee­ding, you should deter­mi­ne whe­ther the data models are requi­red exclu­si­ve­ly for off­line ana­ly­sis or whe­ther an inline HSI came­ra is to be used with the models later. In the lat­ter case, indus­­tri­al-gra­­de soft­ware is requi­red that can com­mu­ni­ca­te with the inten­ded came­ra and exe­cu­te the models in real time.

The fol­lo­wing sec­tion con­ta­ins infor­ma­ti­on on the various steps in the deve­lo­p­ment of data models, inclu­ding pre-pro­ces­­sing and fre­quent­ly used filters.

Data pre-processing and model development

Data pre-pro­ces­­sing and model deve­lo­p­ment are cru­cial steps in che­mo­me­tric soft­ware ana­ly­sis. Data pre-pro­ces­­sing invol­ves clea­ning, noi­se reduc­tion and con­ver­si­on of raw data into a for­mat sui­ta­ble for mode­ling, while model deve­lo­p­ment invol­ves trai­ning models with the pre-pro­ces­­sed data to make pre­dic­tions or clas­si­fy new data. The­se two steps are essen­ti­al to ensu­re accu­ra­te ana­ly­sis and inter­pre­ta­ti­on of data sets in various appli­ca­ti­ons such as qua­li­ty con­trol, inspec­tion of agri­cul­tu­ral pro­ducts and non-inva­­si­­ve ana­ly­sis of samples. In the fol­lo­wing, data prepro­ces­sing and model deve­lo­p­ment are dis­cus­sed in detail to under­stand their importance and the key fac­tors to con­sider when imple­men­ting them in che­mo­me­tric soft­ware packages.

Data (pre)processing

Data pro­ces­sing is a cru­cial step in che­mo­me­tric soft­ware that invol­ves various tech­ni­ques to prepa­re the data for accu­ra­te ana­ly­sis and inter­pre­ta­ti­on. Accu­ra­te data ana­ly­sis is essen­ti­al for the deve­lo­p­ment of relia­ble models that assist in the iden­ti­fi­ca­ti­on of unknown samples and raw mate­ri­als, clas­si­fi­ca­ti­on models and mul­ti­va­ria­te cali­bra­ti­on models. The pro­cess of data pro­ces­sing usual­ly beg­ins with the pre-pro­ces­­sing of the data, inclu­ding the import of spec­tral data, for­mat­ting and cleansing.

Nor­ma­liza­ti­on is a com­mon tech­ni­que used in che­mo­me­tric soft­ware to cor­rect data for varia­ti­ons in their ori­gi­nal inten­si­ty levels due to varia­ti­ons in expe­ri­men­tal con­di­ti­ons, such as absor­ban­ce. Sca­ling is ano­ther method of adap­ting the data to a stan­dard form for com­pa­ri­son. When sca­ling, the data is usual­ly divi­ded by the stan­dard devia­ti­on or the mean value in order to obtain a ratio that faci­li­ta­tes data comparison.

Base­line cor­rec­tion is ano­ther method for eli­mi­na­ting sys­te­ma­tic errors that can affect the accu­ra­cy of the data. When back­ground noi­se and sys­te­ma­tic errors are remo­ved using this tech­ni­que, accu­ra­te data ana­ly­sis and eva­lua­ti­on is gua­ran­teed. Noi­se sup­pres­si­on is ano­ther tech­ni­que for impro­ving the signal-to-noi­­se ratio of the data.

The detec­tion of out­liers is ano­ther important aspect of data pro­ces­sing in che­mo­me­tric soft­ware. Out­liers can signi­fi­cant­ly affect the accu­ra­cy of the results obtai­ned, and their detec­tion and eli­mi­na­ti­on are neces­sa­ry to obtain relia­ble results. Out­liers are usual­ly iden­ti­fied by means of mul­ti­va­ria­te sta­tis­ti­cal ana­ly­sis. Once out­liers are iden­ti­fied, they can be remo­ved from the data set, enab­ling the crea­ti­on of more accu­ra­te and relia­ble models.

To sum­ma­ri­ze, data pro­ces­sing is a cru­cial step in che­mo­me­tric soft­ware that con­tri­bu­tes signi­fi­cant­ly to accu­ra­te data ana­ly­sis and inter­pre­ta­ti­on. Nor­ma­liza­ti­on, sca­ling, base­line cor­rec­tion and noi­se reduc­tion are some com­mon data pro­ces­sing tech­ni­ques, while the detec­tion and rem­oval of out­liers helps to gene­ra­te relia­ble results. Accu­ra­te models and ana­ly­ti­cal methods deve­lo­ped through the use of relia­ble data pro­ces­sing tech­ni­ques faci­li­ta­te, among other things, the ana­ly­sis of agri­cul­tu­ral pro­ducts and non-inva­­si­­ve and non-des­­truc­­ti­­ve test­ing of pro­duct strength.

Development of models

This sec­tion pres­ents seve­ral important fil­ters for the deve­lo­p­ment of models.

Multivariate data analysis

Mul­ti­va­ria­te data ana­ly­sis (MVA) is a powerful tool used in che­mo­me­tric soft­ware packa­ges for the simul­ta­neous eva­lua­ti­on of dif­fe­rent mea­su­re­ments. MVA allows the ana­ly­ti­cal che­mist to eva­lua­te the inter­ac­tion of com­pon­ents within a sam­ple, which can lead to more accu­ra­te pre­dic­ti­ve models and clas­si­fi­ca­ti­ons. This is achie­ved using various tech­ni­ques and models, inclu­ding prin­ci­pal com­po­nent ana­ly­sis (PCA) and regres­si­on analysis.

PCA is a tech­ni­que fre­quent­ly used in MVA to deter­mi­ne cor­re­la­ti­ons bet­ween dif­fe­rent varia­bles. To do this, the ori­gi­nal data is con­ver­ted into a new coor­di­na­te sys­tem in which the first few prin­ci­pal com­pon­ents explain the majo­ri­ty of the data varia­bi­li­ty. The­se prin­ci­pal com­pon­ents can then be used to iden­ti­fy pat­terns, detect out­liers and crea­te more robust models.

Regres­si­on ana­ly­sis is ano­ther MVA tech­ni­que that enables the crea­ti­on of mul­ti­va­ria­te pre­dic­tion models. By ana­ly­zing the rela­ti­onship bet­ween one or more inde­pen­dent varia­bles and a depen­dent varia­ble, regres­si­on ana­ly­sis can esti­ma­te the effects of the­se varia­bles on the depen­dent variable.

Various types of mea­su­re­ments can be used in MVA, inclu­ding near-infrared (NIR) spec­tra, mass spec­tra and absorp­ti­on spec­tra. NIR spec­tra are a non-des­­truc­­ti­­ve and non-inva­­si­­ve ana­ly­sis tech­ni­que that pro­vi­des infor­ma­ti­on about the com­po­si­ti­on of a sam­ple. Mass spec­tra mea­su­re the mas­ses and rela­ti­ve abun­dan­ces of ions in a sam­ple, while absorp­ti­on spec­tra mea­su­re the amount of light absor­bed by a sam­ple at dif­fe­rent wavelengths.

In sum­ma­ry, MVA is a powerful ana­ly­ti­cal tech­ni­que that can be used to eva­lua­te the inter­ac­tions bet­ween dif­fe­rent com­pon­ents in a sam­ple. By using models and tech­ni­ques such as prin­ci­pal com­po­nent ana­ly­sis and regres­si­on ana­ly­sis, as well as mea­su­re­ment tech­ni­ques such as NIR spec­tra and mass spec­tra, MVA enables accu­ra­te pre­dic­tions and clas­si­fi­ca­ti­ons in various are­as, inclu­ding agri­cul­tu­ral pro­ducts, whe­re it can be used to check pro­duct strength.

What is PCA?

PCA (Prin­ci­pal Com­po­nent Ana­ly­sis) is a sta­tis­ti­cal method that is often used in che­mo­me­tric soft­ware packa­ges for the ana­ly­sis and visua­liza­ti­on of mul­ti­va­ria­te data. It aims to trans­form a set of poten­ti­al­ly cor­re­la­ted input varia­bles into a smal­ler, uncor­re­la­ted set of varia­bles cal­led prin­ci­pal com­pon­ents. The­se prin­ci­pal com­pon­ents repre­sent the main sources of varia­ti­on in the ori­gi­nal data and can reve­al hid­den struc­tures, pat­terns and rela­ti­onships bet­ween variables.

PCA is based on line­ar alge­bra and eigenva­lue ana­ly­sis and can pro­cess lar­ge and com­plex data sets. It is often used in che­mo­me­trics to ana­ly­ze and inter­pret spec­tro­sco­pic, chro­ma­to­gra­phic and other ana­ly­ti­cal data, espe­ci­al­ly for the quan­ti­ta­ti­ve ana­ly­sis and clas­si­fi­ca­ti­on of unknown samples and raw mate­ri­als. PCA can be used to crea­te clas­si­fi­ca­ti­on models and mul­ti­va­ria­te cali­bra­ti­on models that can be used to pre­dict the pro­per­ties of new samples with high accuracy.

PCA has seve­ral advan­ta­ges, inclu­ding the abili­ty to redu­ce the dimen­sio­na­li­ty of the data, high­light the most important sources of varia­ti­on, and remo­ve noi­se and red­un­dan­cy. PCA can sim­pli­fy data visua­liza­ti­on and help iden­ti­fy out­liers, trends and clus­ters. Howe­ver, PCA has some limi­ta­ti­ons, inclu­ding sen­si­ti­vi­ty to out­liers, sus­cep­ti­bi­li­ty to over­fit­ting and the assump­ti­on of linea­ri­ty. In addi­ti­on, the inter­pre­ta­ti­on of PCA results can be a chall­enge, espe­ci­al­ly for non-experts.

Alter­na­ti­ve methods to PCA include par­ti­al least squa­res (PLS) regres­si­on, dis­cri­mi­nant ana­ly­sis and fac­tor ana­ly­sis. PLS regres­si­on is sui­ta­ble for cases in which the­re is a strong cor­re­la­ti­on bet­ween the input and out­put varia­bles. Dis­cri­mi­nant ana­ly­sis is useful for clas­si­fy­ing samples into pre­de­fi­ned groups or clas­ses. Fac­tor ana­ly­sis is used when it is hypo­the­si­zed that the varia­bles are asso­cia­ted with a smal­ler num­ber of latent fac­tors that can­not be obser­ved directly.

To sum­ma­ri­ze, PCA is a powerful tool for data ana­ly­sis in che­mo­me­tric soft­ware. Like any ana­ly­sis tech­ni­que, it has its advan­ta­ges and limi­ta­ti­ons and should be used appro­pria­te­ly. Rese­ar­chers and prac­ti­tio­ners need to under­stand the con­cepts and assump­ti­ons under­ly­ing PCA in order to get the most out of it and choo­se the most appro­pria­te method for each spe­ci­fic scenario.

What is LDA?

Line­ar dis­cri­mi­nant ana­ly­sis (LDA) is a machi­ne lear­ning method that is used to clas­si­fy and ana­ly­ze data. It is often used in che­mo­me­tric soft­ware packa­ges to distin­gu­ish bet­ween dif­fe­rent clas­ses of samples or com­pon­ents within a sam­ple. With LDA, the data is pro­jec­ted onto a vec­tor space in which the samples can be dif­fe­ren­tia­ted accor­ding to their desi­gna­ti­ons or classifications.

In con­trast to PCA, which is a method for dimen­sio­na­li­ty reduc­tion, LDA was deve­lo­ped for clas­si­fi­ca­ti­on. This invol­ves fin­ding the line­ar com­bi­na­ti­on of fea­tures that best sepa­ra­tes two or more clas­ses of data points. In che­mo­me­trics, it is used to distin­gu­ish bet­ween dif­fe­rent com­pon­ents in a sam­ple or to reco­gni­ze pat­terns in spec­tral and chro­ma­to­gra­phic data.

Our chemometric software solution

Our che­mo­me­tric soft­ware solu­ti­on, pro­vi­ded by our tech­no­lo­gy part­ner, is a powerful tool for per­forming mul­ti­va­ria­te data ana­ly­sis, deve­lo­ping clas­si­fi­ca­ti­on and cali­bra­ti­on models, and ana­ly­zing a varie­ty of spec­tral data. With an easy-to-use gra­phi­cal user inter­face, we can offer a com­ple­te packa­ge for pro­ces­sing and ana­ly­zing spec­tral data.

One of the big­gest chal­lenges for ana­ly­ti­cal che­mists is the deve­lo­p­ment of accu­ra­te clas­si­fi­ca­ti­on models to ana­ly­ze and eva­lua­te the quan­ti­ta­ti­ve ana­ly­sis of unknown samples and raw mate­ri­als. With our solu­ti­on, this chall­enge is easi­ly over­co­me thanks to the advan­ced mul­ti­va­ria­te methods that enable the crea­ti­on of robust clas­si­fi­ca­ti­on models. The soft­ware is also capa­ble of deve­lo­ping mul­ti­va­ria­te cali­bra­ti­on models, making it a sui­ta­ble soft­ware tool for test­ing pro­duct strength and cri­ti­cal mate­ri­al properties.

The soft­ware offers seve­ral key fea­tures for opti­mal per­for­mance, such as the gra­phi­cal, user-fri­en­d­­ly inter­face for quick and easy ana­ly­sis of spec­tral data. It also includes pre-pro­ces­­sing func­tions that can redu­ce the noi­se ratio and enable the use of deri­va­ti­ve spec­tra, libra­ry spec­tra, mass spec­tra, NIR spec­tra, visi­ble spec­tra and absorp­ti­on spec­tra for che­mo­me­tric analysis.

Embedded system, real-time processing, C++, .NET, Python API: A truly industrial solution

Our che­mo­te­tric soft­ware is also desi­gned to pro­cess a wide ran­ge of spec­tral data, making it com­pa­ti­ble with various ana­ly­ti­cal methods used in the agri­cul­tu­ral indus­try. In addi­ti­on, it has an API for C++, .NET and Python so that sci­en­ti­fic rese­ar­chers and method deve­lo­pers can inte­gra­te their exis­ting soft­ware with it wit­hout worry­ing about compatibility.

One of its uni­que capa­bi­li­ties is the abili­ty to cal­cu­la­te hyper­spec­tral ima­ging (HSI) came­ra data in real time. This func­tion is par­ti­cu­lar­ly useful for the non-inva­­si­­ve and non-des­­truc­­ti­­ve ana­ly­sis of com­pon­ents in samples, such as corn chaff samples. It includes addi­tio­nal vali­da­ti­on samples and sam­ple sets, an important fea­ture to redu­ce errors in ana­ly­sis and impro­ve accuracy.

Final­ly, our sys­tem offers various fil­ter opti­ons, such as ter­rain cor­rec­tion and atmo­sphe­ric cor­rec­tion. The­se fil­ter opti­ons are par­ti­cu­lar­ly neces­sa­ry when working with HSI data and spec­tral mul­­ti-ang­­le images.

The complete list of currently available filters

  • Clas­si­fi­ca­ti­on
    • Distance Clas­si­fier Filter
    • Sup­port Vec­tor Machi­nes Filter
    • Class Map­ping Filter
    • Min/Max Clas­si­fi­ca­ti­on Filter
    • Per-Pixel Decis­i­on Graph Fil­ter Reference
  • Color
    • Color Deter­mi­na­ti­on Filter
    • Color Con­ver­si­on Filter
    • Simi­la­ri­ty Based Color Map Filter
    • ΔE Cal­cu­la­ti­on Fil­ter Reference
  • Decom­po­si­ti­on
    • End­mem­ber Extra­c­tion Filter
    • Abun­dance Deter­mi­na­ti­on Filter
  • Object Detec­tion
    • Mask-Based Object Detec­tor Filter
    • Simi­la­ri­­ty-Based Object Detec­tor Filter
  • Object Pro­ces­sing
    • Per-Object Aver­aging Filter
    • Per-Object Coun­ter Filter
    • Per-Object Scat­ter Cor­rec­tion Filter
    • Per-Object Distance Varia­ti­on Filter
    • Object Reclas­si­fi­ca­ti­on Filter
    • Object Regi­on Aver­aging Filter
    • Per-Object Sta­tis­tics Fil­ter Reference
    • Per-Object Decis­i­on Graph Fil­ter Reference
  • Dimen­sio­na­li­ty Reduction 
    • Prin­ci­pal Com­po­nent Ana­ly­sis Filter
    • Line­ar Dis­cri­mi­nant Ana­ly­sis Filter
    • Subspace Pro­jec­tion Filter
  • Image Pro­ces­sing
  • Ker­nel Ope­ra­ti­on Filter

In sum­ma­ry, our soft­ware solu­ti­on is a relia­ble and advan­ced che­mo­me­tric soft­ware solu­ti­on that pro­vi­des the neces­sa­ry tools and func­tions for per­forming mul­ti­va­ria­te data ana­ly­sis, deve­lo­ping clas­si­fi­ca­ti­on and cali­bra­ti­on models and ana­ly­zing a wide ran­ge of spec­tral data. With its user-fri­en­d­­ly inter­face, it is aimed at various sci­en­ti­fic rese­ar­chers, ana­ly­ti­cal che­mists and method deve­lo­pers in dif­fe­rent indus­tries. The addi­tio­nal soft­ware engi­ne runs models in real-time on embedded sys­tems and can be remo­te­ly con­trol­led using Python, .NET and C++.

About us - Solid Scanner

Let’s take respon­si­bi­li­ty and recy­cle more pla­s­tics - ask us for sui­ta­ble solu­ti­ons. Our port­fo­lio includes solu­ti­ons ran­ging from small, por­ta­ble solu­ti­ons to indi­vi­du­al solu­ti­ons based on hyper­spec­tral came­ra sys­tems for simp­le, auto­ma­ted iden­ti­fi­ca­ti­on of pla­s­tics in the sort­ing pro­cess and for inline pro­cess con­trol, e.g. for homogeneity.