A medicine recognition equipment, named MedEye, which is based on AI edge computing techniques.  The medicine bag loaded with the medicine can be placed flat on the image sampling glass of the medicine recognition equipment. And then barcode reading module of the device will scan the QR code of the medicine bag. Reading the medicine content information, and drive the image lens module under the image sampling glass to shoot the medicine image in the medicine bag for the AI embedded system module in the device to perform AI edge computing to identify the medicine and its quantity. In terms of medicine recognition technology, this work uses the ssd_resnet_50_fpn model. Because the SSD recognition model has a fast operation speed and is matched with FPN, the image pyramid of the SSD is replaced with a feature pyramid, which combines deep and shallow features to recognize small objects. It is more sensitive to improve recognition accuracy, and it is used with the drug

This work uses the edge computing method to identify drugs in the AI embedded system module in the device. AI performs the image recognition results output by Object Detection and filters the frame selection area by 90% through the algorithm to filter the objects, so as to obtain more accurate Count the number of medicines. When the medicines are placed in a medicine bag to block each other, this work uses OpenCV to perform Hough conversion on the photos to calculate the correct amount.

The goal of this work is to integrate into the current standard drug dispensing process in the hospital. This work can assist the dispensing staff in checking whether the drug name, dose, and quantity in the medicine bag are consistent with the prescription of the medicine bag, thereby reducing the error rate of manual inspection and improving the efficiency of dispensing verification. A medical center in the south cooperated to verify the actual adjustment.