Research On Quality Detection Of Viscous Food Filling Based On Convolutional Neural Network

Filling quality detecting as an important link that is directly related to the quality of the products in the production of filling products,especially for the filling of viscous food.For a long time,the unique material properties of thick food have made the filling difficult to be very challenging.Therefore,this thesis carried out a research on the filling quality detection of viscous food based on convolutional neural network aiming at the existing manual detection and sensor detection methods with low accuracy,low efficiency and low ability.The specific research contents are as follows:(1)A method of collecting and processing the image of the filling quality of dense food was proposed to solve the problem that it is difficult to obtain the data samples of the filling quality of viscous food required by the model construction.Firstly,the image acquisition platform of filling quality was built.Secondly,image acquisition is carried out for filling liquid level state and finished product defects.Then,the collected images are sorted out and labeled,Finally,the image data set of filling quality is established,which provides data support for the subsequent design of the detection model of filling flow state of dense food and the detection model of filling product defects.(2)A method is proposed to detect the filling flow status for automatic filling of thick liquid food.The method is based on a convolutional neural network algorithm and it solves the problem of poor accuracy in traditional flow detection devices.An adaptive threshold segmentation algorithm was first used to extract the region of interest for the acquired level image.Next,normalization and augmentation treatment were performed on the extracted images to construct a flow status dataset.A VGG-16 network trained on an Image Net dataset was then used for isomorphic dataoriented feature migration and parameter tuning to automatically extract features and train the model.The identification accuracy and error rate of the network were verified and the advantages and disadvantages of the proposed method were compared to those of other methods.The experimental results demonstrated that the algorithm effectively detects multicategory flow status information and complies with the requirements for actual production.(3)A lightweight filling product defect detection method based on YOLOv4 object detection algorithm was proposed to solve the problem that the traditional filling product defect detection method is difficult to detect multiple targets at the same time.Firstly,the lightweight feature extraction was performed on the input samples through the Mobile Net V3 backbone feature extraction network.Secondly,the deep separable convolution strategy is used to reduce the computational cost of the enhance feature extraction network.Then,the full path aggregation network(FPANet)was designed and efficient channel attention(ECA)mechanism was introduced to improve the target feature expression of enhance feature extraction network.Finally,the model training and precision testing are carried out on the designed lightweight network,and the performance of other object detection algorithms is compared in the same dataset to reveal the superiority of the proposed method.The experimental results show that the proposed method can improve the detection speed while maintaining the accuracy,and the multi-objective high-speed detection of the filling product defects of viscous food was realized.