Method and system for tracking quality of farmed aquatic products based on lipidomic fingerprinting

A method for tracking quality of farmed aquatic products based on lipidomic fingerprinting, comprising the following steps:

• S100, identifying a farmed aquatic product circulation process and selecting sampling points in the aquatic product circulation process;
• S200, performing HS-SPME-based treatment on farmed aquatic products at each sampling point to obtain aquatic product volatile samples;
• S300, collecting GC-MS total ion chromatogram information about the aquatic product volatile samples and constructing peak area vectors;
• S400, calculating a quality attenuation density according to the peak area vector of each sampling point;
• and
• S500, feeding back a circulation step where a risk of quality attenuation occurs to a manager or management system through the quality attenuation density, wherein
• a time interval formed by all sampling points is set as a sampling period, and sampling points in reverse and forward time directions of each sampling point are defined as backward-direction and forward-direction points thereof respectively;
• cosine distances between peak area vectors of each sampling point and all forward-direction points thereof are calculated, and the forward-direction point with the maximum cosine distance is denoted as a first marker point;
• the average of the cosine distances of all the forward-direction points is denoted as a forward-direction cosine distance;
• a first distance condition is set to be that the cosine distance is greater than the forward-direction cosine distance, and a second distance condition is set to be that at least one intermediate forward-direction point having a cosine distance less than the forward-direction cosine distance exists between the sampling point and the forward-direction point;
• a first marker point that satisfies the first and second distance conditions is searched in reverse time order from the sampling points and is denoted as a second marker point;
• the number of backward-direction points that have the same second marker point as a current sampling point in the reverse time direction of the sampling points is counted and denoted as a reverse-charging number, and a reverse-charging coefficient Rc.idx is preset, Rc.idx∈[0.3, 1];
• the product of the reverse-charging number and the reverse-charging coefficient is rounded up to a reverse-charging value Rc.vl, a third marker point of the sampling point is defined as the Rc.vl-th forward-direction point of the second marker point in the forward time direction, and a time interval between the sampling point and the third marker point is denoted as a reference domain of the sampling point;
• and the ratio of cosine distances between the second and third marker points is denoted as a reverse-charging regression ratio Rc.rvt of the sampling points;
• for any sampling point, the peak area vector of the sampling point is subtracted from that of the first sampling point in the reverse time direction thereof to obtain a step difference vector, test objects corresponding to step difference elements with values greater than zero in the step difference vector are classified as class-adding objects, otherwise, are classified as class-lossing objects, and a write sequence of each class-adding object is denoted as a class-adding sequence, the step difference element referring to any element in the step difference vector;
• a percentile value of the step difference element of the class-adding object in a set of step difference elements corresponding to various identical class-adding objects in the reference domain is denoted as a class-adding order value, an accumulated value of class-adding order values corresponding to all class-adding objects at the sampling point is denoted as a class-adding accumulated value Icv, and forward-direction points in the reference domain of which the class-adding accumulated value is less than the class-adding accumulated value corresponding to the current sampling point are denoted as adaptive forward-direction points;
• and a quality attenuation density Pdt is calculated through the class-adding accumulated value and the reverse-charging regression ratio:
• Pdt(i0)=Rc.rvti0Σi1=1len exp(Icvi0−Icvi1)−diff{Pmksi0,Pmksi1};
• wherein i0 is the serial number of the sampling point, Pdt(i0) represents the quality attenuation density of the i0th sampling point, i1 is the accumulated variable, len is the number of adaptive forward-direction points corresponding to the sampling points, expo is the exponential function with the natural constant e as the base, and Icvi0 and Icvi1 are the class-adding accumulated values of the i0th sampling point and the i1th adaptive forward-direction point;
• and diff{ } is the Euclidean distance function, which returns the Euclidean distance between two calling sequences, Pmksi0 and Pmksi1 represent the class-adding reading sequences of the i0th sampling point and the i1th adaptive forward-direction point respectively, and the class-adding reading sequence is composed of the class-adding sequences when the i0th sampling point and the i1th adaptive forward-direction point correspond to various step difference elements.