图像二值化分割在物流包装物标识图像识别中的应用OA

In response to the issues of poor segmentation effect and high identification error rates in existing recognition methods for identification images,this study proposes a logistics packaging identification image recognition method based on image binarization segmentation.The original logistics packaging identification image is preprocessed through grayscale conversion and Gaussian filtering to select the optimal threshold and reduce background interference.The anchor box position and identification area are determined via identification localization and loss function computation.The binary segmentation algorithm is further improved by integrating global and local information.Using this enhanced algorithm,the packaging identification image is segmented to extract the target identification region.This target-region image is then input into an improved deep convolutional neural network model,where the identification is successively localized,recognized,and decoded.Test results demonstrate that the logistics packaging label images segmented by the proposed method retain complete label information,free from interference such as noise points or light spots.The minimum misrecognition rate of packaging labels reaches 0.05%.