Damage Detection Of Apple Based On Hyperspectral Imaging Technology

Apple is one of the common fruits in our country,and the quality of the apples would affected by the mechanical damage.The apple with damage need to be sort out immediately to avoid not only the entire fruit rot,but also the other common storage of fruit affected.Damage detection of apples can reduce the loss in the storage and transportation process.As the traditional method manual sorting has many disadvantages,such as heavy workload,poor repeatability,long measurement cycle,large environmental impact and easy damage to apples.In order to solve these problems,this paper presents a method for detecting apples damage based on hyperspectral imaging technology.The apples were chosen as the research object in this paper.Firstly,the mechanical damage detection of apples with hyperspectral imaging technology need to correct including black and white brightness correction,near sphere correction method based on the spectral image preprocessing,to reduce the apple brightness uneven distribution and make the damage easily to distinguish.The principal component analysis and ROC curve were used to analyze the hyperspectral data to select the effective wavelengths.The selected wavelengths were 529nm,622nm and 970nm,respectively.In order to test the effective of the wavelength in fruit damage detection,SVM was used to classify the apple spectral data with the effective wavelength,and the average accuracy was 91.25%.The establishment of SVM classification model of spectral information acquisition effective band of apple in the region of interest,the size of the sliding window using 10*10(step 5)across each apple within the region,with an average spectral values of the regional classification as model input,to judge whether the area exists and will judge for damage,damage to the visible region marking the optical image of apple,the damage detection model based on the SVM classification accuracy rate is 88.67%,can detect and mark the position of the damage of apple.Based on the above results,it can be proved that it is feasible to use the effective wavelength image to detect apple damage.