Automatic Detection Of Meat Based On Color Contrast And Feature Classification Algorithm And Its Hardware And Software Implementation

Meat is a necessity on the human table.As the population increases,the demand for meat is gradually increasing.At present,some processes in meat segmentation and inspection are still labor-intensive.With the development of technology,automated assembly lines are gradually replacing labor.The outbreak of the new crown epidemic has accelerated the unmanned process of processing and inspection in the meat segmentation industry.However,meat and other products are natural products with complex textures and significant individual differences.The defects such as fats and oils are close to the color of the body,which makes automated detection a certain challenge.This paper takes the wing tip,wing middle,wing root,chicken breast automatic classification and fat detection obtained by dividing the chicken carcass on the automated production line,and proposes color feature detection algorithm based on color contrast enhancement and feature classification algorithm.The algorithm is implemented by software and hardware integration,which realizes the automatic classification and defect detection of meat.This paper proposes a color feature rapid detection algorithm and detection process based on color contrast enhancement(CCE).The algorithm obtains a new color space XYZ by fusing and convolving the information of each dimension of the original RGB color space,so that the color noise that is similar in the RGB color space can be more distinguished in the XYZ color space.Through the detection and verification of the red and yellow series of similar colors,the results show that the algorithm can enhance the color contrast while suppressing the noise,solve the problem that the artificial color contrast algorithm(ACC)can only process the red series of features,and at the same time make up for the problem that the principal component analysis algorithm(PCA)cannot filter out the noise in a targeted manner.For meat with obvious characteristics to be classified,this paper proposes a meat classification method based on target area.The threshold is used to determine the number of pixels in the area of the object to be measured,so as to distinguish between wing tips and chicken breast.For sample objects with unobvious features,this paper proposes a classification method based on convolutional neural networks.This method uses the training set and test set obtained by data enhancement to train and test the convolutional neural network respectively,which can realize the classification of the wing middle and the wing root.This paper designs a software and hardware coordination system for automated inspection production lines,and implements the algorithm designed in the built software and hardware coordination system platform.Using Zynq chip as the development platform,using embedded Linux system as the control center,using hardware to achieve algorithm acceleration,Experimental results show that the system can process at least 3 pictures per second,meeting the production line speed requirement of 10,000 pieces/hour.In order to verify the effectiveness of the algorithm proposed in this article,this article carried out related experiments.The experimental results show that the accuracy of pre-classification of wing tips and chicken breasts based on the target area threshold can reach 100%,and the recognition accuracy of wing middles and wing roots based on convolutional neural networks is above 98%.In the detection of fat in chicken breast meat,compared with the20% average relative error rate obtained by the traditional principal component analysis method under the RGB color space,the detection process of the color contrast enhancement algorithm proposed in this paper can generally reduce the average relative error of the fat part detection to within 5%.The results show that the color contrast enhancement algorithm can effectively remove the interference of similar color features.