Research On The Method Of Cutting Pieces Contour Features Analysis Based On R~2CNN

The cut piece is an important middleware for the finished car seat,and its quality is directly related to the final quality of the seat.Among them,the feature of the eye knife of the cutting piece is an important index to measure whether the cutting piece is qualified.The category determination and size measurement of the eye knife are important inspection tasks that need to be completed before the piece is spliced.At present,the contour feature detection of cut pieces usually requires human interaction,high labor intensity,low efficiency,and depends on the experience of the workers.In recent years,detection systems based on machine vision and artificial intelligence technology have been gradually applied to the field of target detection and measurement due to their advantages of non-contact,high accuracy,and fast speed.Aiming at the detection of the contour features of the cut pieces,combining image processing and deep learning theory,we propose a feature detection method based on improved R~2CNN,on the basis of that,the analysis method of the contour features of the cut pieces is studied.Firstly,an image acquisition system consisting of a industrial camera,an LED backlight board,and an industrial computer was designed to obtain the cutting images.Secondly,in view of the difficulty of detection due to the complexity and diversity of eye-knife features,an R~2CNN network was used to detect the position of the eye-knife and return the angle information.At the same time,the ROI pooling strategy in the network structure was improved by designing a pool size that fits the contours of the piece to obtain a more accurate tilt target return frame as an accurate measurement area.Finally,according to the precise measurement requirements of features,based on sub-pixel edge extraction,a measurement method that incorporates the smallest circumscribed rectangle is proposed to obtain accurate feature parameters of eye-blade depth.To sum up,this thesis proposes a new method for contour feature detection and size measurement of cutting pieces,and realizes automatic analysis of different cutting contour features.The experimental results show that within the measurement area of 700mm?500mm,the accuracy of feature detection is 96%,the classification accuracy is 93.8%,and the average measurement accuracy of the size is 0.3 pixels.The method effectively completes the accurate measurement of contour features and has high accuracy.And the robustness is of great significance to the automatic analysis for the contour features of the cut pieces.