Research On Image Processing Technology And Its Application In Food Safety Traceability Scales

With the continuous progress of economy and society,people's dietary culture is becoming more and more diverse.Food hygiene and safety has become a hot topic.People buy fruits and vegetables and other foods in the supermarket,etc.They often expect to know the fruit and vegetable products of food origin,production status,inspection and quarantine information.In response to these circumstances,the scientists developed a traceability scale to solve the above situation.The traceability scale is used as a sales terminal,to achieve the input of food information,the transmission and recording of food safety information and transaction information,and print quality traceable certificate to consumers(Back yards).The traditional traceability scale is to achieve the recognition of fruit and vegetable products by two-dimensional code,but because fruit and vegetable products often need to package before affix the two-dimensional code labels,which consumes a lot of manpower and resources.In addition,there have a wide variety of fruits and vegetables and different k:inds of fruits and vegetables have different prices.All prices of fruits and vegetables mainly depend on artificial memory,which greatly increased the economy costs and the time costs of supermarket trainers.Based on this,this paper carries out the research on algorithms of fruits and vegetables recognition based on computer vision for traceability scales.For extracting features,Gabor wavelet transform is used to extract features from fruits and vegetables images.Considering the high dimension features,then two-dimensional PC A method is used for dimensionality reduction.In the classification stage,this paper uses sparse representation classifier and GMM model and deep neural network fusion classifier to recognize objects,Experimental results show that the proposed methods can get better recognition results.The main work of this paper is as follows:(1)The influence of Gabor wavelet transform and the size of different Gabor kernel window on recognition rate is discussed.Experimental results show that using Gabor wavelet transform to extract features has better spatial domain,frequency domain locality,and multi-directional selectivity.(2)On the basis of principal component analysis theory,2D-PCA is used to reduce the dimensions of Gabor features.Experimental results show that the use of 2D-PCA dimensionality reduction greatly improves the operation efficiency,compared to PCA method.(3)Combining compressed deep learning theory,the GMM model and Deep Neural Network fusion classifier is adopted for fruits and vegetable recognition.Experimental results show that the adopted methods achieve better recognition results,compared to GMM and sparse representation classifier.(4)Comparing Gabor +2 D-PCA and down sampling +2 D-PCA method,the experimental results showed that the using of Gabor feature extraction method get 94%-95%recognition rate,much higher than the latter.Based on the discussion of specific algorithms and results analysis,conclusion is obtained:Using Gabor +2 D-PCA feature extraction and dimensionality reduction method,combined with GMM and deep neural network classifier to classify fruits and vegetables can obtain a good recognition rate.