The invention belongs to the technical field of packaging jet printing, and discloses an external packaging
carton two-dimensional code jet printing quality online monitoring and error correction process, which comprises the following steps: building a vision,
laser, spectrum and pressure four-source sensing
positioning system in a preprocessing stage, comprehensively collecting
carton material and form related data, and forming a pre-calibration
closed loop in cooperation with microsample jet printing
verification; in the jet printing process, a high-speed imaging
system of self-adaptive light supplement works synchronously, after an image is preprocessed through a combinatorial
algorithm, various defects such as blooming and scratches are accurately recognized through a ResNet34-Transform-GAN fusion model, and then an LSTM neural network is used for predicting the development trend of the defects and conducting early warning; a three-level parameter
configuration system of cloud decision, edge execution and twin
simulation is utilized, an edge calculation unit uploads data such as materials, sizes and flatness collected in real time to the cloud, historical optimal parameters of the corresponding materials are called, and then dynamic adjustment is conducted in combination with the conveying speed.