Method and system for made-tea quality assessment in manufacturing process

A system for assessing made-tea quality, comprising:

• memory storing instructions;
• one or more communication interfaces;
• and
• one or more hardware processors coupled to the memory via the one or more communication interfaces, wherein the one or more hardware processors are configured by the instructions to:
• randomly collect a physical grade of a made-tea sample from multiple sections of a production batch in a tea factory, wherein the made-tea liquor sample is formulated with a liquor containing water;
• acquire by a spectral camera under controlled illumination, a spectral image of the made-tea sample indicative of chemical composition, wherein the spectral camera captures the spectral image over a predetermined wavelength range to form a multi-dimensional spectral hypercube;
• preprocess each slice of the acquired spectral image by,
• i. converting each spectral image into grayscale to reduce colour complexity and to improve intensity variations,
• ii. applying Gaussian blur on each grayscale of corresponding spectral image to smooth random noise and soften sharp edge, and
• iii. performing histogram equalization on each blurred image to obtain improved contrast of distinguishable spectral features;
• apply a spectral segmentation on the pre-processed spectral image to mark one or more regions of interests corresponding to tea particle;
• create a bounding box on each candidate region among the one or more region of interests;
• extract a plurality of features from each candidate region, wherein the plurality of features includes (i) a spectral reflectance data indicative of chemical profile from the bounding box region superimposed on raw spectral hypercube, (ii) a total dissolved solutions (TDS) measurement value obtained from the liquor used to prepare made-tea sample indicative of approximate dissolved compounds, and (iii) a physical grade anomaly detection parameter identifying presence of foreign particles;
• detect from the plurality of features a plurality of anomalies comprising a spectral anomaly, a visual anomaly and a physical grade anomaly, wherein the spectral anomaly detects deviations with the spectral reflectance data attributable to heavy metal contamination(s), the visual anomaly detects at least one physical contaminants discrepancies, and the physical grade anomaly validates a plurality of morphological features in comparison with factory provided grade standards;
• assign a negative score to each detected anomaly based on severity and type of impact, and recommending corrections for each detected anomaly based on the negative score, wherein the negative score includes a high negative indicating critical discrepancies, a moderate negative indicating moderate discrepancies, and a low negative indicating minor discrepancies;
• estimate a chemical composition of the made-tea by correlating the spectral reflectance data with the TDS measurement value;
• compute a quality level score by combining the chemical composition, the physical grade anomaly and the negative score, and categorizing the quality level score into at least one of a high, medium and low based on a predefined threshold;
• assess a final grade value for the made-tea sample based on the quality level score (QLS) and a valuation grade score (VLS), and the final grade value is compared with a predefined grade threshold to quantify factory standards, wherein the valuation grade score is a quantitative measure of body and zing attributes using the spectral reflectance data and the TDS value;
• and
• recommend the tea factory, a quality assessment report indicating personalized improvements based on the final grade value indicative of acceptable quality, the quality level score, and a proximity value, wherein the proximity value is determined based on factory-specific reference spectral profiles.

One or more non-transitory machine-readable information storage mediums comprising one or more instructions which when executed by one or more hardware processors cause:

• randomly collecting a physical grade of a made-tea sample from multiple sections of a production batch in a tea factory, wherein the made-tea sample liquor is formulated with a liquor containing water;
• acquiring by a spectral camera under controlled illumination, a spectral image of the made-tea sample indicative of chemical composition, wherein the spectral camera captures the spectral image over a predetermined wavelength range to form a multi-dimensional spectral hypercube;
• preprocessing each slice of the acquired spectral image by,
• i. converting each spectral image into grayscale to reduce colour complexity and to improve intensity variations,
• ii. applying Gaussian blur on each grayscale of corresponding spectral image to smooth random noise and soften sharp edge, and
• iii. performing histogram equalization on each blurred image to obtain improved contrast of distinguishable spectral features;
• applying a spectral segmentation on the pre-processed spectral image to mark one or more regions of interests corresponding to tea particle;
• creating a bounding box on each candidate region among the one or more region of interests;
• extracting a plurality of features from each candidate region, wherein the plurality of features includes (i) a spectral reflectance data indicative of chemical profile from the bounding box region superimposed on raw spectral hypercube, (ii) a total dissolved solutions (TDS) measurement value obtained from the liquor used to prepare made-tea sample indicative of approximate dissolved compounds, and (iii) a physical grade anomaly detection parameter identifying presence of foreign particles;
• detecting from the plurality of features, a plurality of anomalies comprising a spectral anomaly, a visual anomaly and a physical grade anomaly, wherein the spectral anomaly detects deviations with the spectral reflectance data attributable to heavy metal contamination(s), the visual anomaly detects at least one physical contaminants discrepancies, and the physical grade anomaly validates a plurality of morphological features in comparison with factory provided grade standards;
• assigning a negative score to each detected anomaly based on severity and type of impact, and recommending corrections for each detected anomaly based on the negative score, wherein the negative score includes a high negative indicating critical discrepancies, a moderate negative indicating moderate discrepancies, and a low negative indicating minor discrepancies;
• estimating chemical composition of the made-tea by correlating the spectral reflectance data with the TDS measurement value;
• computing a quality level score by combining the chemical composition, the physical grade anomaly and the negative score, and categorizing the quality level score into at least one of a high, medium and low based on a predefined threshold;
• assessing a final grade value for the made-tea sample based on the quality level score (QLS) and a valuation grade score (VLS), and the final grade value is compared with a predefined grade threshold to quantify factory standards, wherein the valuation grade score is a quantitative measure of body and zing attributes using the spectral reflectance data and the TDS value;
• and
• recommending the tea factory, a quality assessment report indicating personalized improvements based on the final grade value indicative of acceptable quality, the quality level score, and a proximity value, wherein the proximity value is determined based on factory-specific reference spectral profiles.