Research On Apple Defect Detection Based On Biological Speckl

Fresh agricultural products such as fruits and vegetables are the main source of People’s Daily diet.Due to their easy spoilage,it is difficult to preserve for a long time,resulting in huge waste and economic losses.Among them,physical damage is an unavoidable problem in the postharharvest chain of agricultural products.Problems can be found in the early stage of damage formation in time to ensure accurate quality classification of damaged fruits and reduce unnecessary losses.Laser biological speckle technology is a non-invasive,fast and non-destructive optical detection technology.Through the irradiation of highly coherent light on the object to be measured,the reflected light on the surface of the object and the scattered light through the inner surface will interfere with each other and superposition to form a random distribution of uneven light spots with alternating light and dark,which can reflect the different physical and chemical properties of the object to be measured.Such as internal particle motion state,temperature,composition and other important information.In this paper,the image processing algorithm based on laser biospeckle technology and different degree damage recognition and classification of apple were studied.Specific research contents are as follows:(1)The laser biological speckle experimental system was built,and the automatic detection software based on C# was developed to assist the equipment debugging and image acquisition of the experimental system.The influence of laser,lens,camera and other main equipment selection on the measurement of biological speckle is analyzed.(2)Based on wavelet transform and particle size distribution theory,the spectral characteristics and energy distribution characteristics of dynamic biological speckle are studied,and a speckle time-frequency energy grain size algorithm is proposed.The visual differences of different speckle image processing algorithms are compared,and the influence of different size time window on algorithm efficiency and experimental results is analyzed.The results show that the energy particle size distribution of different active samples is different.Compared with the traditional algorithm,the improved algorithm in this paper can not only provide more abundant time and frequency information of biological speckle,but also take into account damage discrimination and computational efficiency,and reduce the influence of irrelevant noise.(3)The quantitative relationship between the speckle activity of damaged apples and different evaluation indexes was explored,and the time-varying rule of biological speckle activity of damaged apples was characterized by recording the intensity of discrete indexes at different time points.The frequency characteristics of biological speckle activity in different damaged parts of apple were studied systematically.The results showed that the sensitivity of different frequencies to damage degree was different.(4)Based on the five statistical characteristics of moment of inertia,absolute difference,wavelet entropy,correlation coefficient and energy particle size,the feature extraction was carried out by Bayesian stepwise discriminant analysis,and the apple damage classification model was established by support vector machine and artificial neural network.The results show that the neural network model with 8 hidden layer neurons combined with 0 ～ 0.31 Hz and 2.18 ～ 2.5 Hz energy granularity features has the best performance,and the overall classification accuracy is 95%.The results of this study will provide references for the detection and grading of physical defects of apples,and is expected to be further popularized in the field of automatic fruit quality detection,which has certain significance for promoting the quality and safety of fresh fruits,improving the post-harvest economic value and international competitiveness of agricultural products.