The uptake by the industry of the existing knowledge about the online NIR analysis is being much slower, compared to the acceptance of the at-line analysis. The Research Group of the authors since 2001 has been in close collaboration with the largest Spanish rendering plant to evalate the ability of different for the quality control of animal protein processed by-products. Since 2017, and after several years of research, the company decided to invest in a on-line project. The work done until, for moving from at line to on line analysis in the rendering plant will be summarised in the Conference.

Iberian pork meat has exceptional sensory and nutritional attributes, which are related to the breed and the feeding regime of the animals. Regarding the breed purity, two categories can be considered: 100% Iberian products and Iberian products coming from crossed animals (Iberian x Duroc). The aim of this work was to evaluate the viability of using portable Near-infrared sensors for the in situ authentication of Iberian pork fresh meat according to its breed. Models were developed using partial least squares discriminant analysis. The results confirm the viability of using NIRS to guarantee the authenticity of the Iberian pork meat.

Software tools for chemometric analysis of NIRS data have existed since the first NIRS instruments appeared on the market in the late 1970s. Generally, these software appear attached to a certain instrumentation. Recently, some works have started to use open-source software, such as R and Python, but the development status is still in its infancy, particularly in the case of the latter. This work tries to generate information on the potential of the open-source Python software for the implementation of multivariate algorithms and signal pre-treatment methods for the quantitative and qualitative NIRS analysis of olive oils.

Mosquitoes are the most life-threatening insect to human on Earth. Main disease vector mosquitoes inhabiting in Korea cause Zika fever, Yellow fever, Malaria, West Nile fever, Japanese encephalitis, etc. Only Malaria has cure among them. Usually, the disease vector mosquito species are counted and identified manually with optical microscopy, which needs huge labor and causes human error. Although the recent mosquito trap devices are developed, they can only count the number of mosquitoes without the species identification. This study proposes a deep learning image analysis technique to identify species along with the population of mosquitoes using SWIN-transformer model. The non-maximum suppression (NMS) technique for both RGB and fluorescence images has been applied for the improvement of prediction accuracy. Results revealed that the proposed model has achieved good performance for mosquito identification.

This study was conducted to confirm the localization possibility of the automated pork carcass grading machines in Korea. This experiment has used a total of 174 carcasses. Image analysis was conducted in three main steps: 1) image preprocessing, 2) feature extraction, 3) regression model build-up. For features extraction and model building, we used the U-net and Gaussian processing regression respectively. The Analysis was done for prediction of LMP and seven different prime cuts. The prediction results were satisfactory to the European minimum standards thus making the localization of the pork carcass grading machine possible.

Future space crop production systems will require that plant health and food safety is determined with minimal crew intervention. A prototype hyperspectral and chlorophyll fluorescence imaging system was designed for early symptom detection of abiotic plant stress in crop production systems. A watering system was developed for imposing water stress treatments (mild or severe drought, flooding) on candidate leafy green crops to be grown on the International Space Station. Daily images recorded changes in crop reflectance and chlorophyll fluorescence during 28-day growouts. Harvest data recorded leaf area, fresh weight, dry weight, plant height and leaf number. The imaging and harvest data were used to evaluate the ability of the prototype imaging system to differentiate between the water stress treatments.

This experiment was conducted to predict fatty acid contents in beef using short-wave infrared (SWIR) hyperspectral imaging (HSI) technique. The HSI datasets were acquired from the longissimus dorsi and further evaluated the quality parameter of these samples. A partial least square regression (PLSR) model with spectral preprocessing was applied to predict the fatty acid in beef samples. The obtained results showed high coefficients of determination (R2 > 0.8). The overall outcomes suggest that the SWIR-HSI technique might be utilized as non-destructive measurement tool for the determination of fatty acid content in beef.

Corn is commonly used as a good source of food and feed, as well as for producing cooking oil and starch. However, corn is among the many agricultural staples that can be easily contaminated with aflatoxin, a poisonous mycotoxin produced by molds that can have serious effects on human and animal health, and rapid and effective methods for detecting aflatoxin in the corn are lacking for on-site use in food processing operations. This study investigated the use of short-wavelength infrared (900 - 2500 nm) hyperspectral image data for detecting aflatoxin in ground maize, using measurements of aflatoxin content via chemical analysis for sample reference. Preliminary results are reported for the development of a detection model using deep learning to detect aflatoxin-contaminated corn powder.

Hyperspectral images are typically acquired at high spatial and spectral resolutions, being essential the reduction of data for the implementation of this technology at industrial level. The aim of this work was the optimization and development of algorithms for the selection of the region of interest in oranges hyperspectral data. PLS and its multilinear version, NPLS, were used to model the internal quality of oranges. The results obtained in external validation enabled to carry out a screening of the product according to its flavour, confirming that the use of multilinear models could reduce the noise and data redundancy.

It is crucial to improve the efficiency of plant breeding and crop yield in order to fulfill rising food demands. In plant phenotyping study, the capability to correlate morphological traits of plants plays an important role. However, measuring the plant phenotypes manually is prone to human errors, labor intensive and time-consuming. Hence, it is important to develop techniques for measurement of plant phenotypic data accurately and rapidly. The objective of this study was to find out the feasibility of point cloud data of 3D LiDAR including RGB image for plant phenotyping. The obtained results were then verified through the manually acquired data for sorghum and soybean plant samples. The overall results showed remarkable correlation between point cloud data and manually acquired data for plant phenotyping. This correlation indicates that the 3D Lidar imaging system have potential to measure phenotypes of crops in rapid and accurate way.

The contamination and sanitation inspection and disinfection (CSI-D) system has been developed to enable rapid detection, immediate intervention, and documentation of organic residue (debris), saliva, respiratory droplets on surfaces that may cause contamination disease spread. Novel aspects of the CSI-D solution include the combination of contamination identification and immediate remediation of the potential threat (bacteria, virus) using UVC light disinfection, and documenting this process to provide traceable evidence of disinfection. CSI-D reveals invisible contamination and truly defines cleanliness with measurements and documentation to provide reassurance to staff and promote cleanliness to customers.

Sanitation inspection is an ongoing concern for food distributors, restaurant owners, caterers, and others who handle and serve food. They must prevent food contamination but now must also deal with potential infection spread among workers and customers. Beyond zero tolerance legal requirements and damage to institutional or restaurant reputation, loss of trust with workers and customers can be very costly. We provide fluorescence imaging results that were measured, analyzed, and recorded on different high touch surfaces in restaurants and institutional kitchens. We have developed an algorithm to classify cleanliness levels based on the extent of organic residues detected.

Poultry meat is the most consumed meat in the US. To ensure a wholesome and safe product, carcasses and the viscera are inspected for disease and other conditions indicating they should be condemned as unfit for consumption. Septicemia-Toxemia (SepTox) is the most common carcass condemnation observed and reported. In this paper, we present a fast, convenient, and easy-to-use handheld system to detect SepTox for condemnation in post-eviscerated poultry carcasses. We provide fluorescence imaging measurements and analysis on poultry carcasses for developing machine learning models to classify carcasses with SepTox for future high speed process line automated imaging.

We evaluate a handheld multispectral fluorescence imaging device to detect a bacterial colony on leafy greens. The most common diseases causing illness transmitted by leafy vegetables are norovirus, Shiga toxin-producing E. coli (STEC), and Salmonella, according to a CDC. Listeria and Cyclospora can also cause these illnesses. We will test the efficacy of a Contamination, Sanitization Inspection, and Disinfection (CSI-D) system using light at two fluorescence excitation wavelengths, ultraviolet C (UVC) at 275 nm and violet at 405 nm. Tests will evaluate the detection efficacy of device on inoculated control specimens of leafy greens during a time lapsed study.

Meat and poultry can be contaminated by pathogens like E. coli and salmonella. Animal fecal matter and ingesta host these pathogens, so developing a method to detect contamination on meat surfaces is crucial. We visited four meat processing facilities and used a handheld fluorescence imaging device to detect fecal matter or ingesta on carcasses. We investigated the efficiency and reliability of a state-of-the-art semantic segmentation algorithm to segment fecal or ingesta contaminated regions in meat surfaces images. Industry could use CSI-D to train employees in trimming carcasses as part of their Hazard Analysis Critical Control Point zero tolerance plan.

This paper introduces an autonomous robot system with an intelligent contaminant detection and disinfection device. The system can maneuver and using a robotic arm can detect, disinfect, and document invisible organic contamination on surfaces that may host pathogens. We will present repeated autonomous detection of hard-to-see potato starch biofilms on a conveyor belt surface. The system will be designed to report the time and location of the detected contamination on a digital floor plan. The system also records the amount of germicidal energy dosage to the surface by calculating the optical power, exposure time and distance to the surface.