The invention relates to the field of agricultural product
quality monitoring, in particular to an agricultural product tracing link-oriented abnormal
event monitoring method, which comprises the following steps of: S1, defining a batch coding rule and an
event type dictionary, and establishing a tracing relation model; s2, based on the
traceability relation model, deploying
sensing data acquisition equipment at each key
control point of an agricultural
product traceability link, deploying an
edge computing gateway at each place, and completing credible data chaining; s3, unifying data of different sources to a standard timeline, establishing association keys according to batches, places and devices, and performing preprocessing; s4, constructing an
anomaly detection model which comprises a
time sequence prediction model, a visual anomaly model, a rule anomaly model and a graph anomaly model, and carrying out
anomaly detection based on the preprocessed data; s5, calculating a risk
score by adopting an evidence fusion strategy, and obtaining an output alarm packet; and S6, according to the output alarm packet, performing
traceability positioning and influence range evaluation based on the
traceability relation model. The agricultural product quality abnormity monitoring efficiency and reliability are effectively improved.