Apparatus and method for soldering inspection using correlation neural network

摘要

In the present invention, cross-correlation and self-correlation between each color pattern are calculated using a color neural network obtained by acquiring a color image of a soldering portion under a tricolor circular illumination device and having a shape recognition algorithm by an artificial intelligence technique, It is possible to reliably recognize the shape of the soldering portion of the PCB and to perform the soldering process even when the target layers to be recognized are similar to each other when the new image is received and the position of the inspection target is not constant, The present invention relates to a brazing inspection apparatus and method using a correlation neural network capable of inspecting a brazing shape of a soldering portion under a tricolor circular illumination having a coaxial with a camera, And a system for accepting and classifying the image information extracted from the solder shape measuring means And a solder shape recognition means for recognizing a classification condition for the soldering portion image information by simulating the human classification ability with the sample learning data when a new image is received. A pre-processing step of converting the three-dimensional image information of the soldering part obtained under the three-color circular illumination having the coaxiality into more meaningful and simple information for the soldering part inspection, and comparing the result calculated in the pre- And the learning process of learning the synaptic weights of the neural network on the limited samples.