CN-116524494-B - Kiwi fruit internal defect nondestructive identification method, device, equipment and storage medium

CN116524494BCN 116524494 BCN116524494 BCN 116524494BCN-116524494-B

Abstract

The invention belongs to the technical field of quality detection, and discloses a method, a device, equipment and a storage medium for nondestructive identification of internal defects of kiwi fruits. The method comprises the steps of scanning a sample kiwi fruit to obtain an original scanning image, obtaining defect characteristic information according to the original scanning image, determining a defect evaluation data set according to the defect characteristic information, constructing a classification model according to the defect evaluation data set, obtaining the defect volume ratio of the kiwi fruit to be classified through the classification model, and quantitatively classifying the defects of the kiwi fruit to be classified according to the defect volume ratio. Through the mode, the automatic analysis of the kiwi fruits which are required to be subjected to defect identification and classification in the follow-up mode is realized based on scanning of the sample kiwi fruits, the quantitative classification of the defects of the kiwi fruits is realized, the manual damage visual inspection method is eliminated, the purposes of nondestructive automatic analysis and classification of the internal defects of the kiwi fruits are achieved, and the efficiency and the effect of kiwi fruit quality detection are improved.

Inventors

WANG JIAHUA
DAI HUANG
LIU XIAODAN
LIN YUQING
Li Qiaocong
GUO YUQING

Assignees

武汉轻工大学

Dates

Publication Date: 20260508
Application Date: 20230411

Claims (8)

1. The nondestructive identification method for the internal defects of the kiwi fruits is characterized by comprising the following steps of: scanning the sample kiwi fruits to obtain an original scanning image; establishing a mask according to the kiwi fruit boundary in the original scanned image to remove the image background in the original scanned image, so as to obtain a background removed image; Dividing the background removed image to obtain a sample area, and determining effective voxels according to the sample area; Dividing the effective voxels into gray scale abnormal regions according to a preset rule to obtain a plurality of divided regions; Removing the noise area in the partition area to obtain a feature screening area; Determining a defect distance of each feature defect in the feature screening area; Determining defect quantity information of the sample kiwi fruits according to the defect distance; determining sphericity information of the sample kiwi fruits according to the volume information and the area information of each characteristic defect; determining aspect ratio information of each characteristic defect according to the volume information; Determining the number of sample voxels and the number of defect voxels according to the sample region and the defect characteristic region; determining defect volume ratio information according to the number of the sample voxels and the number of the defect voxels; Determining defect characteristic information of the sample kiwi fruits according to the defect quantity information, the sphericity information, the length-width ratio information and the volume ratio information; determining a defect evaluation dataset according to the defect characteristic information; constructing a classification model according to the defect evaluation data set; Obtaining the defect volume ratio of the kiwi fruits to be classified through the classification model; and carrying out defect quantitative classification on the kiwi fruits to be classified according to the defect volume ratio.
2. The method of claim 1, wherein said determining a defect review dataset from said defect characterization information comprises: Performing destructive evaluation on the sample kiwi fruits to obtain real defect information of the sample kiwi fruits; the real defect information and the defect characteristic information are corresponding to obtain defect comparison information; and obtaining a defect evaluation data set according to the defect comparison information.
3. The method of claim 1, wherein said constructing a classification model from said defect review dataset comprises: obtaining patch configuration information, convolution configuration information and data input configuration information; Constructing an initial neural network according to the patch configuration information, the convolution configuration information and the data input configuration information; acquiring decision tree configuration information and leaf number configuration information; constructing an initial random forest model according to the decision tree configuration information and the leaf number configuration information; dividing the defect evaluation data into a training set and a testing set; And training the initial neural network and the initial random forest model according to the training set and the testing set to obtain a classification model.
4. The method of claim 1, wherein quantitatively classifying the defects of the kiwi fruit to be classified according to the defect volume ratio, comprises: Comparing the defect volume ratio with a first defect volume ratio and a second defect volume ratio, respectively; Determining defect degree information of the kiwi fruits to be classified according to the comparison result; And carrying out defect quantitative classification on the kiwi fruits to be classified according to the defect degree information.
5. The method of claim 4, wherein determining the defect level information of the kiwi fruit to be classified according to the comparison result comprises: Determining whether the defect volume ratio is greater than or equal to the first defect volume ratio according to the comparison result to obtain a first judgment result; Determining whether the defect volume ratio is greater than or equal to the second defect volume ratio according to the comparison result to obtain a second judgment result; Determining the defect classification of the kiwi fruits to be classified according to the first judging result and the second judging result; and determining the defect degree information of the kiwi fruits to be classified according to the defect classification.
6. The utility model provides a nondestructive recognition device of kiwi fruit internal defect, its characterized in that, the nondestructive recognition device of kiwi fruit internal defect includes: the sample scanning module is used for scanning the sample kiwi fruits to obtain an original scanning image; The information extraction module is used for establishing a mask according to the kiwi fruit boundary in the original scanning image to remove the image background in the original scanning image to obtain a background removal image, dividing the background removal image to obtain a sample area, determining the number of sample voxels and the number of defect voxels according to the sample area, dividing the effective voxels into a plurality of divided areas according to a preset rule, removing a noise area in the divided areas to obtain a feature screening area, determining the defect distance of each feature defect in the feature screening area, determining the defect number information of the sample kiwi fruit according to the defect distance, determining the sphericity information of the sample kiwi fruit according to the volume information and the area information of each feature defect, determining the length-width ratio information of each feature defect according to the volume information, determining the number of sample voxels and the number of defect voxels according to the sample area and the defect feature area, determining the defect volume ratio information according to the number of sample voxels and the defect voxel number, and determining the feature defect information of the sample kiwi fruit according to the defect number information, the sphericity information, the length-width ratio information and the volume ratio information; a data set preparation module for determining a defect evaluation data set according to the defect characteristic information; the model construction module is used for constructing a classification model according to the defect evaluation data set; the parameter calculation module is used for obtaining the defect volume ratio of the kiwi fruits to be classified through the classification model; And the quantitative classification module is used for quantitatively classifying the defects of the kiwi fruits to be classified according to the defect volume ratio.
7. A kiwi internal defect nondestructive identification apparatus comprising a memory, a processor and a kiwi internal defect nondestructive identification program stored on the memory and operable on the processor, the kiwi internal defect nondestructive identification program configured to implement the kiwi internal defect nondestructive identification method of any one of claims 1 to 5.
8. A storage medium, wherein a kiwi fruit internal defect nondestructive identification program is stored on the storage medium, and when the kiwi fruit internal defect nondestructive identification program is executed by a processor, the kiwi fruit internal defect nondestructive identification method according to any one of claims 1 to 5 is realized.

Description

Kiwi fruit internal defect nondestructive identification method, device, equipment and storage medium Technical Field The invention relates to the technical field of quality detection, in particular to a method, a device, equipment and a storage medium for nondestructive identification of internal defects of kiwi fruits. Background Post harvest quality of kiwi fruits is primarily maintained by cryopreservation, however, adverse effects of cryopreservation on kiwi fruits include lignification of subcutaneous tissues. Severe lignification greatly reduces the edibility of kiwi fruits, resulting in considerable economic losses. Bruising is unavoidable during kiwi fruit picking and transportation. Although cold injury and bruise can both have adverse effects on the storage characteristics of kiwi fruits, the bruised kiwi fruits are less storage-resistant, can release a large amount of ethylene, and influence other normal kiwi fruits in the ripening process. At present, the detection of the cold injury lignification and the bruise of the kiwi fruits needs to be visually detected by cutting, and in order to ensure the processing accuracy and efficiency of the kiwi fruits, the industry has started to search for a better technology for screening the bruise and the cold injury kiwi fruits. The foregoing is provided merely for the purpose of facilitating understanding of the technical solutions of the present invention and is not intended to represent an admission that the foregoing is prior art. Disclosure of Invention The invention mainly aims to provide a method, a device, equipment and a storage medium for nondestructive identification of internal defects of kiwi fruits, and aims to solve the technical problem that damage defects of kiwi fruits in the prior art are difficult to identify on the premise of not damaging fruits. In order to achieve the purpose, the invention provides a nondestructive identification method for internal defects of kiwi fruits, which comprises the following steps: scanning the sample kiwi fruits to obtain an original scanning image; obtaining defect characteristic information according to the original scanning image; determining a defect evaluation dataset according to the defect characteristic information; constructing a classification model according to the defect evaluation data set; Obtaining the defect volume ratio of the kiwi fruits to be classified through the classification model; and carrying out defect quantitative classification on the kiwi fruits to be classified according to the defect volume ratio. Optionally, the obtaining defect characteristic information according to the original scanned image includes: establishing a mask according to the kiwi fruit boundary in the original scanned image to remove the image background in the original scanned image, so as to obtain a background removed image; Dividing the background removed image to obtain a sample area, and determining effective voxels according to the sample area; dividing the effective acceleration gray abnormal region according to a preset rule to obtain a plurality of divided regions; Removing the noise area in the partition area to obtain a feature screening area; and determining defect characteristic information of the sample kiwi fruits according to the characteristic screening area. Optionally, the determining defect characteristic information of the sample kiwi fruit according to the characteristic screening area includes: Determining a defect distance of each feature defect in the feature screening area; Determining defect quantity information of the sample kiwi fruits according to the defect distance; determining sphericity information of the sample kiwi fruits according to the volume information and the area information of each characteristic defect; determining aspect ratio information of each characteristic defect according to the volume information; Determining the number of sample voxels and the number of defect voxels according to the sample region and the defect characteristic region; determining defect volume ratio information according to the number of the sample voxels and the number of the defect voxels; and determining defect characteristic information of the sample kiwi fruits according to the defect quantity information, the sphericity information, the length-width ratio information and the volume ratio information. Optionally, the determining a defect evaluation dataset according to the defect feature information includes: Performing destructive evaluation on the sample kiwi fruits to obtain real defect information of the sample kiwi fruits; the real defect information and the defect characteristic information are corresponding to obtain defect comparison information; and obtaining a defect evaluation data set according to the defect comparison information. Optionally, the constructing a classification model according to the defect evaluation dataset includes: obtaining patch configuration info