scispace - formally typeset
Search or ask a question
Patent

Potato surface defect detection method based on convolutional neural network

TL;DR: In this article, a potato surface defect detection method based on a convolutional neural network was proposed, which is characterized in that the CNN was adopted for detecting the surface defects of the potatoes, a training set serves as a sample for training the CNN, then a prediction set is input to the network after training, and a softmax classifier is adopted for achieving classification of the potato.
Abstract: The invention puts forward a potato surface defect detection method based on a convolutional neural network. Colored images of potatoes are obtained by industrial cameras, and an image processing module is used for conducting analysis processing on the obtained images. The method is characterized in that the convolutional neural network is adopted for detecting the surface defects of the potatoes,a training set serves as a sample for training the convolutional neural network, then a prediction set is input to the network after training, and a softmax classifier is adopted for achieving classification of the potatoes. According to the method, feature extraction is not required, there is no need to set specific detection algorithms according to defect types, the images are directly input tothe network, and therefore the potato surface defect detection method has the advantages of being high in accuracy, short in algorithm running time, high in feasibility and robustness and the like, and has a significant meaning in the potato processing industry.
Citations
More filters
Patent
24 May 2019
TL;DR: Wang et al. as mentioned in this paper presented a fruit quality detection system and method based on depth learning, which mainly judge the fruit quality according to the color, texture, shape and size of the fruit surface to complete the classification.
Abstract: The invention discloses a fruit quality detection system and method based on depth learning. The fruit quality detection system and method based on depth learning mainly judge the fruit quality according to the color, texture, shape and size of the fruit surface to complete the classification. The whole detection system is composed of a fruit transmission system, a computer vision recognition system and a classification system. Compared with a traditional fruit detection system, the system has the distinct characteristics of deep learning, and the characteristics are mainly reflected in the fruit image processing. Through the detailed analysis of a network structure of a convolution neural network algorithm and the training process of the convolution neural network, a fruit image recognition system based on the convolution neural network is constructed. Compared with a general neural network algorithm, the fruit image recognition system has the characteristics of simple structure, lesstraining parameters, high adaptability and the like.

7 citations

Patent
19 Oct 2018
TL;DR: In this article, a micro part quality detection system based on a convolutional neural network (CNN) is presented. But the system is not suitable for the detection of micro parts and can improve the automation degree and the efficiency of detection, and reduce the influence of human factors on the detection process.
Abstract: The invention discloses a micro part quality detection system based on a convolutional neural network. The micro part quality detection system comprises: A, collecting the surface image information ofa micro part by an image acquisition module formed by microscopic vision; B, detecting the image collected by the microscopic vision by using a convolutional neural network model, and classifying thedetected defect images; C, transmitting the classifying result into a main controller, and sending a control signal to a terminal actuator; and D, carrying out picking and classifying on the corresponding micro part by the terminal mechanical arm actuator according to the control signal transmitted by the controller so as to convey the part into the corresponding receiving box, such that the whole system completes the detection and defect classification on the surface quality of the micro part. According to the present invention, the system can effectively used for the detection of micro parts, and can improve the automation degree and the efficiency of detection, and reduce the influence of human factors on the detection process and the labor intensity of workers.

5 citations

References
More filters
Patent
21 Dec 2016
TL;DR: Wang et al. as mentioned in this paper proposed a panoramic-vision potato sorting and defect detection device which comprises a conveying device, a detection camera obscura, a sorting mechanism, an infrared sensor module, an image acquisition mechanism, a image processing-analyzing module, a data fusion module and a time sequence module.
Abstract: The invention discloses a panoramic-vision potato sorting and defect detection device which comprises a conveying device, a detection camera obscura, a sorting mechanism, an infrared sensor module, an image acquisition mechanism, an image processing-analyzing module, a data fusion module and a time sequence module, wherein a convolutional neural network and a support vector machine I (SVM I) are built in the image processing-analyzing module; an SVM II is built in the data fusion module; and the time sequence module is used for coordinating actions of all parts. The invention further discloses a potato sorting detection method using the above device. According to the device and the potato sorting detection method disclosed by the invention, all-around detection can be completed without turning over potatoes in a detection process. Accordingly, the potatoes can be protected from unnecessary damage on the one hand, and on the other hand, the instability during dynamic photo taking and detection can be avoided, and the image clarity and the detection accuracy can be improved. The device and the potato sorting detection method can be widely applied to real-time online detection of the external quality of potatoes, namely a kind of agricultural product, and has significance for promoting development of the Chinese potato industry.

13 citations

Patent
27 Nov 2001
TL;DR: In this paper, a bootstrap aggregating strategy is employed to combine the results of the classifications for each sample in the test data set made by each neural network model, which is indicative of the geographical origin of the commodity.
Abstract: The detection method includes generating a plurality of neural network models. Each model has as a training set a data set from a plurality of samples of a commodity of known origins. Each sample has been analyzed for a plurality of elemental concentrations. Each neural network model is presented for classification a test data set from a plurality of samples of a commodity of unknown origins. As with the training set, the samples have been analyzed for the same plurality of elemental concentrations. Next a bootstrap aggregating strategy is employed to combine the results of the classifications for each sample in the test data set made by each neural network model. Finally, a determination is made from the bootstrap aggregating strategy as to a final classification of each sample in the test data set. This final classification is indicative of the geographical origin of the commodity. The system includes software for generating the neural network models and a software routine for performing the bootstrap aggregating strategy.

8 citations

Patent
09 Oct 2013
TL;DR: In this paper, a potato defect detection method was proposed for rapidly detecting peeled potatoes which are damaged, rot and sprout, and the thermal image processing processes of a computer are as follows: carrying out median filtering or average filtering and noise reduction on thermal images, extracting the regions of the potato thermal images by adopting a threshold value segmentation method, calculating the average temperature values of pixels in the regions, then calculating the difference value between the temperature value of each pixel in the region, and calculating the areas S of real objects to which the pixels with the absolute values of the difference
Abstract: The invention discloses a potato defect detecting method. According to the potato defect detecting method, a thermal imaging technology is used for rapidly recognizing peeled potatoes which are damaged, rot and sprout, and the thermal image processing processes of a computer are as follows: carrying out median filtering or average filtering and noise reduction on thermal images, extracting the regions of the potato thermal images by adopting a threshold value segmentation method, calculating the average temperature values of pixels in the regions, then calculating the difference value between the temperature value of each pixel in the regions and the average temperature value, and calculating the areas S of real objects to which the pixels with the absolute values of the difference values being greater than 1 DEG C, if S is greater than 25mm , the potatoes are defective. The potato defect detecting method disclosed by the invention has high efficiency and good recognition accuracy and can be popularized and applied to the field of potato defect detection.

7 citations

Patent
20 Jun 2017
TL;DR: In this article, a potato defect detection and recognition system design based on machine vision is described, in which defected potatoes are identified and classified on a ZYNQ platform by utilizing a machine vision library Open CV of an embedded Linux system.
Abstract: The invention discloses a potato defect detection and recognition system design based on machine vision. The potato defect detection and recognition system design is characterized in that defected potatoes are identified and classified on a ZYNQ platform by utilizing a machine vision library Open CV of an embedded Linux system; characteristic factors of the defected potatoes with green peels, dry rot, crust and mechanical damages are extracted and R, G and B discrete degrees of variable defect factors are analyzed to realize detection and recognition of surface defects of the potatoes, and the algorithm precision is greatly improved. Wavelet transform is applied to analysis and detection of potato shapes of the potatoes, and ellipse radiuses of the potatoes are extracted and are subjected to normalization processing; grading is carried out through a RBF (Radial Basis Function) neural network, so that the efficiency and precision of recognizing the defected potatoes by grades are improved; potato images are pre-processed by utilizing an FPGA (Field Programmable Gate Array) and an algorithm in the Open CV is subjected to accelerated processing; a calculation speed and the algorithm efficiency are remarkably improved. A testing result shows that compared with an existing defected potato recognition and classification technology based on software image processing, an image processing algorithm is innovated and optimized by a novel method based on a hardware structure platform, and the processing speed and the algorithm efficiency are greatly improved; theories and experiments show that the design has relatively ideal detection efficiency and speed on the recognition and classification of the defected potatoes in an actual process. The design has very great significance on a potato processing industry.

3 citations