论文部分内容阅读
Abstract With the expansion of cultivation scale， pitaya diseases are gradually increasing. Traditionally relying on human observation to judge the disease is limited by the skills and experience of the observer， which cannot guarantee the accuracy and real-time of the judgment， and consumes much manpower and time. In this study， by collecting， segmenting， and labeling images of 4 main diseases of pitaya in the field， an image database of main diseases of pitaya in the field was constructed to provide a basis for computer image recognition of pitaya diseases. Thereby， it benefits reducing manual error and improving the accuracy and real-time of disease identification for agricultural production， but also lays a foundation for the future development of intelligent agriculture.
Key words Pitaya; Disease; Image; Recognition; Database building
Research Background and Significance
At present， human observation methods were mainly used to discover and identify pitaya diseases. Manual observations have a strong subjectivity， which is limited by the observers recognition skills and experience， and cannot guarantee the accuracy of the observation results. In actual production， the field of pitaya disease detection consumes more manpower， and it is difficult to check one by one， and it is impossible to have a more accurate real-time judgment on the incidence of the entire field. The discovery of pitaya diseases is not timely， or even wrong judgments， resulting in excessive or incorrect use of pesticides， causing unnecessary pollution to the environment， which is not conducive to the safety of pitaya food. The establishment of the pitaya disease image database helps to quickly and accurately identify the pitaya disease types and the incidence of the disease， which is conducive to the right strategies at the beginning， reduces the use of pesticides， and enhances the prevention and control of diseases to reduce production costs. Compared with traditional manual detection and identification methods， it has higher stability and superiority.
Research Status
Pitaya disease research
Pitaya is a perennial tropical fruit crop of the generus "Hylocereus" and "Selenicereus" in the family Cactaceae， which includes red pitaya （H. polyrhizus）， white pitaya （H. undatus） and yellow pitaya （. megalanthus）. It is native to Central America and northern South America and has a wide range of cultivation worldwide. In the early 1990s， domestic agricultural growers of China began to cultivate and gradually promoted cultivation in southern provinces such as Hainan， Guangdong， and Yunnan[1]. Red pitaya is the most common commercially cultivated variety in Hainan. The fruit is rich in nutritional value and is popular with people， and have high economic value. With the growing area of pitaya filed， the pitaya disease is gradually increasing， which reduces the quality of pitaya and reduces the yield， resulting in greater economic losses. The main diseases of pitaya in Hainan province include canker disease， stem rot disease and virus disease. According to the different symptoms of viral disease， pitaya virus disease can be divided into mosaic virus disease， round spot chlorosis symptom virus disease and so on. Pathogen of canker disease， Neoscytalidium dimidiatum. （Penz.） Crous&Slippers， mainly infect the stems and fruits of pitaya. In the early stage of disease， there are fading spots with round depressions of 3 to 4 mm in diameter. In the middle stage， crater-like reddish brown or black protrusions are gradually formed. In the later stage， crater-like ulcer spots are formed. At the same time， the internal decay of the susceptible tissue causes symptoms such as stem rot and fruit cracking.
The stem rot pathogen， Fusarium[Fusarium solani （Mart.） Sacc.]. The disease mainly harms the stem. At the early stage of infection， small brown spots are formed on the surface of the stem. In the middle stage of the infection， the tissue around the lesion gradually changes from green to light yellow and the range continues to expand. In the later period， the entire stem changes from green to dark yellow， translucent， soft rot tissue. In the end， except for the central vascular tissue， the stems rotted as a whole， and the disease spread up and down along the stems until the whole plant was necrotic.
Mosaic virus disease， the pathogen is a virus of the Potexvirus， such as the cactus X virus （CVX）. In the early stage， small chlorotic spots appear on the surface of the stem， and then the number increases. In the middle stage of the disease， the color of the chlorosis become yellow， and a yellow-green mosaic surface is formed in the later stage.
Round spot chlorotic virus disease， the pathogen is also a virus of the Potexvirus. The initial manifestation is the appearance of light green faded round small lesions on the surface of the stem. In the middle period of susceptibility， as the plant planting time prolongs， the area and size of the lesions increase， and the lesions continue to form into pieces in the later period of susceptibility. Viral diseases can cause decline of the plant resistance， which make the plant more susceptible to other diseases， and affect yield.
Research on automatic identification of plant diseases
With the development of computers and communication networks， people gradually apply computer image processing technology to crop disease identification. Computer image processing technology means that the computer digitizes the target picture， and then extracts， classifies， and calculates the obtained digital electrical signal， thereby improving the utilization rate of the image to meet peoples needs. Compared with visually recognizing images， computer image processing technology is more accurate and reliable. In other countries， computers have been applied to the judgment and classification of crop diseases since the 1970s[2-4]. In 2011， Cheng et al.[5] published a new image segmentation method. This method segments images based on the tests of significance. This segmentation method produces significance regions that can better exclude the surrounding complex environment interference in the image for disease recognition. In 2015， based on the surveillance video of rapeseed cultivation greenhouse， MA et al.[6] successfully identified rapeseed lesion types by segmenting and identifying the rapeseed lesion images in the video screenshot.
In China， although computer image processing technology has been developed late in the detection of plant diseases， many research results have also been achieved. In 2013， Peng et al.[7] relied on the size， texture and color characteristics of cucumber leaf lesions to identify disease pictures of cucumber leaves and achieved good experimental results. In 2014， Wang et al.[8] relied on the color features of HSV in wheat leaf disease images， and used binary classification to identify and identify wheat powdery mildew and rust images， and achieved good experimental results. In addition， many achievements and breakthroughs have been made in the identification of leaf diseases of plants such as corn[9]， rice[10]， cotton[11] and tomato[12]. No research report on image processing technology in pitaya diseases has been reported.
Research status of establishment of plant disease image database
Some universities in the United States have established an image library of plant disease specimens[13]. Some scholars have studied DAD （The digitally assisted diagnosis）. The DAD of plant diseases is to convert plant disease samples into images， and then upload the images to the network where it is identified and judged by relevant experts. And DAD has been used to some extent in China[14]. But DAD did not deviate from manual identification. The application of computer recognition to plant disease recognition has not established a large-scale plant disease image database based on computer image recognition technology to judge plant diseases.
In this study， the image of the main diseases of pitaya in the field was segmented， identified and annotated by the computer， and the image database of the main diseases of the pitaya in the field was established in order to apply the image processing technology to the identification of the pitaya diseases. The establishment of the field image database of pitaya diseases will significantly reduce the workload of pitaya disease prevention and control， and lay a foundation for the future development of smart agriculture. Design and Realization of Image Database of Pitaya Disease
Data collection
The field image is collected mainly in the main producing areas of pitaya in Hainan Province， such as Qionghai， Dongfang， Dingan， Ledong， Danzhou and so on. Among them， one of the most important image collection site is located at the base of Chinese Academy of Tropical Agricultural Sciences （19°29′N， 109°29′E， 139.962 m） in Danzhou， Hainan Province. The plant row is about 1.0 m×2.5 m. It was planted with upright cement column and steel cable frame. The pitaya variety is H. polyrhizus & H. costaricensis， planted in 2015. The images collected were all taken from 8：00 am to 10：00 am on a sunny day from January to December 2019. In this study， Canon EOS 800D was selected as the image acquisition device， with a resolution of 6 000 × 4 000 pixels and shooting under natural light. During the shooting process， a tripod was used to fix the camera， the height of the lens and the pitaya stem was adjusted to be consistent， and the distance between the lens and the stem surface was 0.5-1.0 m. The above four disease symptoms of pitaya stems were selected for picture collection.
Data preprocessing
The use of digital image preprocessing can greatly improve the manifestation of symptoms in the image. Before image segmentation， the image needs to be pre-processed with light and scale normalized. The image size was uniformly adjusted to 50 × 50 pixels. Different regions in the same image have different features. Graphic segmentation refers to dividing the image according to these features and selecting target regions. After the disease image was segmented， the disease category of the image was cyclically annotated. First， the disease category name of the image was annotated， and then the diseased grade was annotated （according to the early， middle， and late stages of the disease， the diseased grade is divided into three levels）. And then the disease image and the annotated label were stored in the database. Finally， the image categories and the labels in the image library were checked by the experts.
Data analysis
After sorting the images of the pitaya diseases collected in the experiment， 9 836 images and 89 video clips of 4 main diseases were finally obtained. The four main diseases are canker disease picture 2 （a）， stem rot disease picture 2 （b）， mosaic virus disease picture 2 （c） and round spot chlorotic virus disease picture 2 （d）.
Mosaic virus disease and round spot chlorotic virus disease both have chlorotic spots in the early stage， which is more difficult to distinguish each other， compared with other two diseases. In the field， the stretch direction of the pitaya stems is different， resulting in different angles between the plane of the stem and the sunlight. In particular， some stems are backlit， or some areas are shaded， which makes it difficult to identify the image. These pictures need to be processed to improve the accuracy of identification. Fig. 3 is a diagram of the different diseased grades of pitaya of canker disease. The first four pictures are the early stage， the middle four pictures are the middle stage， and the last four pictures are the late stage. It can be seen that at the beginning of the disease， smaller fading spots appear， In the middle stage， crater-like reddish brown or black protrusions are gradually formed. In the later stage， crater-like ulcer spots are formed.
Summary
The construction of the field image database of pitaya disease is an indispensable link in the research on the image recognition of different diseases of field pitaya. In this paper， computer image processing technology is used to segment and annotate images of pitaya of different diseases. An image database of the main diseases of pitaya in the field is established. Multi-labeling and hierarchical labeling of different diseases and different diseased grade of pitaya are realized to identify the categories of pitaya diseases， providing a basis for image recognition of the pitaya disease category and incidence. This database contains two types of data： the image of the pitaya disease and the video clip of the pitaya disease. The images and video clips of the database can provide training and test samples for the study of disease recognition algorithms. The establishment of a plant disease identification database not only greatly reduces the workload of disease prevention and control in agricultural production， but also lays a foundation for the future development of intelligent agriculture.
Acknowledgment
The authors thank Cui Yu， Fang Zongyu （Forestry College， Hainan University） for their help in image acquisition.
References
[1] LIU JW. Experimental research on introduction and cultivation of pitaya in solar greenhouse in Beijing area[D]. Yangling： Northwest A & F University， 2015. （In Chinese）
[2] BAURIEGEL E， GIEBEL A， HERPPICH WB. Rapid Fusarium head blight detection on winter wheat ears using chlorophyll fluorescence imaging[J]. Journal of Applied Botany and Food Quality， 2010， 83（2）： 196-203.
[3] ZHAO Y， HE Y， XU X. A novel algorithm for damage recognition on pest-infested oilseed rape leaves[J]. Computers and Electronics in Agriculture， 2012（89）： 41-50.
[4] MAHLEIN AK， STEINER U， HILLNHUTTER C， et al. Hyperspectral imaging for small scale analysis of symptoms caused by different sugar beet diseases[J]. Plant methods， 2012， 8（1）： 3-7. [5] CHENG MM， ZHANG GX， MITRA NJ， et al. Global contrast based salient region detection[C]//Proceedings of 2011 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）， Colorado Springs， CO， USA， 2011： 409-416.
[6] MA JC， LI XX， WEN HJ， et al. A key frame extraction method for processing greenhouse vegetables production monitoring video[J]. Computers and Electronics in Agriculture， 2015（111）： 92-102.
[7] PENG ZW， SI XL， WANG X， et al. Recognition of cucumber downy mildew based on image processing and fuzzy clustering method[J]. Chinese Journal of Agricultural Machinery Chemistry，2013， 34（2）： 212-216. （In Chinese）.
[8] WANG ML， NIU XJ， ZHANG HM， et al. Research on feature extraction and recognition technology of common diseases of wheat leaves[J]. Computer Engineering and Applications， 2014， 50（7）： 154-157. （In Chinese）.
[9] WANG N， WANG KR， XIE RZ， et al. Pattern recognition of corn leaf diseases based on Fisher discriminant analysis[J]. Chinese Journal of Agricultural Sciences， 2009， 42（11）： 3836-3842. （In Chinese）.
[10] GUAN ZX， TANG J， YANG BJ， et al. Research on rice disease recognition method based on image[J]. China Rice Science， 2010， 24（5）： 497-502. （In Chinese）.
[11] ZHANG JH， QI LJ， JI RH， et al. Cotton disease identification based on rough set and BP neural network[J]. Journal of Agricultural Engineering， 2012， 28（7）： 161-167. （In Chinese）.
[12] CHAI AL， LI BJ， SHI YX， et al. Tomato leaf disease recognition based on computer vision technology[J] .Chinese Journal of Horticulture， 2010， 37（9）： 1423-1430. （In Chinese）.
[13] HOLMES GJ， BROWN EA， RUHL G. What s a picture worth？ The use of modern telecommunications in diagnosing plant diseases[J]. Plant Disease， 2000（84）： 1256-1265.
[14] GONG GS， ZHANG M， QIN Y， et al. Construction of a digital standard library of plant diseases[J]. Journal of Yunnan Agricultural University， 2002， 17（4）： 361-362. （In Chinese）.
Key words Pitaya; Disease; Image; Recognition; Database building
Research Background and Significance
At present， human observation methods were mainly used to discover and identify pitaya diseases. Manual observations have a strong subjectivity， which is limited by the observers recognition skills and experience， and cannot guarantee the accuracy of the observation results. In actual production， the field of pitaya disease detection consumes more manpower， and it is difficult to check one by one， and it is impossible to have a more accurate real-time judgment on the incidence of the entire field. The discovery of pitaya diseases is not timely， or even wrong judgments， resulting in excessive or incorrect use of pesticides， causing unnecessary pollution to the environment， which is not conducive to the safety of pitaya food. The establishment of the pitaya disease image database helps to quickly and accurately identify the pitaya disease types and the incidence of the disease， which is conducive to the right strategies at the beginning， reduces the use of pesticides， and enhances the prevention and control of diseases to reduce production costs. Compared with traditional manual detection and identification methods， it has higher stability and superiority.
Research Status
Pitaya disease research
Pitaya is a perennial tropical fruit crop of the generus "Hylocereus" and "Selenicereus" in the family Cactaceae， which includes red pitaya （H. polyrhizus）， white pitaya （H. undatus） and yellow pitaya （. megalanthus）. It is native to Central America and northern South America and has a wide range of cultivation worldwide. In the early 1990s， domestic agricultural growers of China began to cultivate and gradually promoted cultivation in southern provinces such as Hainan， Guangdong， and Yunnan[1]. Red pitaya is the most common commercially cultivated variety in Hainan. The fruit is rich in nutritional value and is popular with people， and have high economic value. With the growing area of pitaya filed， the pitaya disease is gradually increasing， which reduces the quality of pitaya and reduces the yield， resulting in greater economic losses. The main diseases of pitaya in Hainan province include canker disease， stem rot disease and virus disease. According to the different symptoms of viral disease， pitaya virus disease can be divided into mosaic virus disease， round spot chlorosis symptom virus disease and so on. Pathogen of canker disease， Neoscytalidium dimidiatum. （Penz.） Crous&Slippers， mainly infect the stems and fruits of pitaya. In the early stage of disease， there are fading spots with round depressions of 3 to 4 mm in diameter. In the middle stage， crater-like reddish brown or black protrusions are gradually formed. In the later stage， crater-like ulcer spots are formed. At the same time， the internal decay of the susceptible tissue causes symptoms such as stem rot and fruit cracking.
The stem rot pathogen， Fusarium[Fusarium solani （Mart.） Sacc.]. The disease mainly harms the stem. At the early stage of infection， small brown spots are formed on the surface of the stem. In the middle stage of the infection， the tissue around the lesion gradually changes from green to light yellow and the range continues to expand. In the later period， the entire stem changes from green to dark yellow， translucent， soft rot tissue. In the end， except for the central vascular tissue， the stems rotted as a whole， and the disease spread up and down along the stems until the whole plant was necrotic.
Mosaic virus disease， the pathogen is a virus of the Potexvirus， such as the cactus X virus （CVX）. In the early stage， small chlorotic spots appear on the surface of the stem， and then the number increases. In the middle stage of the disease， the color of the chlorosis become yellow， and a yellow-green mosaic surface is formed in the later stage.
Round spot chlorotic virus disease， the pathogen is also a virus of the Potexvirus. The initial manifestation is the appearance of light green faded round small lesions on the surface of the stem. In the middle period of susceptibility， as the plant planting time prolongs， the area and size of the lesions increase， and the lesions continue to form into pieces in the later period of susceptibility. Viral diseases can cause decline of the plant resistance， which make the plant more susceptible to other diseases， and affect yield.
Research on automatic identification of plant diseases
With the development of computers and communication networks， people gradually apply computer image processing technology to crop disease identification. Computer image processing technology means that the computer digitizes the target picture， and then extracts， classifies， and calculates the obtained digital electrical signal， thereby improving the utilization rate of the image to meet peoples needs. Compared with visually recognizing images， computer image processing technology is more accurate and reliable. In other countries， computers have been applied to the judgment and classification of crop diseases since the 1970s[2-4]. In 2011， Cheng et al.[5] published a new image segmentation method. This method segments images based on the tests of significance. This segmentation method produces significance regions that can better exclude the surrounding complex environment interference in the image for disease recognition. In 2015， based on the surveillance video of rapeseed cultivation greenhouse， MA et al.[6] successfully identified rapeseed lesion types by segmenting and identifying the rapeseed lesion images in the video screenshot.
In China， although computer image processing technology has been developed late in the detection of plant diseases， many research results have also been achieved. In 2013， Peng et al.[7] relied on the size， texture and color characteristics of cucumber leaf lesions to identify disease pictures of cucumber leaves and achieved good experimental results. In 2014， Wang et al.[8] relied on the color features of HSV in wheat leaf disease images， and used binary classification to identify and identify wheat powdery mildew and rust images， and achieved good experimental results. In addition， many achievements and breakthroughs have been made in the identification of leaf diseases of plants such as corn[9]， rice[10]， cotton[11] and tomato[12]. No research report on image processing technology in pitaya diseases has been reported.
Research status of establishment of plant disease image database
Some universities in the United States have established an image library of plant disease specimens[13]. Some scholars have studied DAD （The digitally assisted diagnosis）. The DAD of plant diseases is to convert plant disease samples into images， and then upload the images to the network where it is identified and judged by relevant experts. And DAD has been used to some extent in China[14]. But DAD did not deviate from manual identification. The application of computer recognition to plant disease recognition has not established a large-scale plant disease image database based on computer image recognition technology to judge plant diseases.
In this study， the image of the main diseases of pitaya in the field was segmented， identified and annotated by the computer， and the image database of the main diseases of the pitaya in the field was established in order to apply the image processing technology to the identification of the pitaya diseases. The establishment of the field image database of pitaya diseases will significantly reduce the workload of pitaya disease prevention and control， and lay a foundation for the future development of smart agriculture. Design and Realization of Image Database of Pitaya Disease
Data collection
The field image is collected mainly in the main producing areas of pitaya in Hainan Province， such as Qionghai， Dongfang， Dingan， Ledong， Danzhou and so on. Among them， one of the most important image collection site is located at the base of Chinese Academy of Tropical Agricultural Sciences （19°29′N， 109°29′E， 139.962 m） in Danzhou， Hainan Province. The plant row is about 1.0 m×2.5 m. It was planted with upright cement column and steel cable frame. The pitaya variety is H. polyrhizus & H. costaricensis， planted in 2015. The images collected were all taken from 8：00 am to 10：00 am on a sunny day from January to December 2019. In this study， Canon EOS 800D was selected as the image acquisition device， with a resolution of 6 000 × 4 000 pixels and shooting under natural light. During the shooting process， a tripod was used to fix the camera， the height of the lens and the pitaya stem was adjusted to be consistent， and the distance between the lens and the stem surface was 0.5-1.0 m. The above four disease symptoms of pitaya stems were selected for picture collection.
Data preprocessing
The use of digital image preprocessing can greatly improve the manifestation of symptoms in the image. Before image segmentation， the image needs to be pre-processed with light and scale normalized. The image size was uniformly adjusted to 50 × 50 pixels. Different regions in the same image have different features. Graphic segmentation refers to dividing the image according to these features and selecting target regions. After the disease image was segmented， the disease category of the image was cyclically annotated. First， the disease category name of the image was annotated， and then the diseased grade was annotated （according to the early， middle， and late stages of the disease， the diseased grade is divided into three levels）. And then the disease image and the annotated label were stored in the database. Finally， the image categories and the labels in the image library were checked by the experts.
Data analysis
After sorting the images of the pitaya diseases collected in the experiment， 9 836 images and 89 video clips of 4 main diseases were finally obtained. The four main diseases are canker disease picture 2 （a）， stem rot disease picture 2 （b）， mosaic virus disease picture 2 （c） and round spot chlorotic virus disease picture 2 （d）.
Mosaic virus disease and round spot chlorotic virus disease both have chlorotic spots in the early stage， which is more difficult to distinguish each other， compared with other two diseases. In the field， the stretch direction of the pitaya stems is different， resulting in different angles between the plane of the stem and the sunlight. In particular， some stems are backlit， or some areas are shaded， which makes it difficult to identify the image. These pictures need to be processed to improve the accuracy of identification. Fig. 3 is a diagram of the different diseased grades of pitaya of canker disease. The first four pictures are the early stage， the middle four pictures are the middle stage， and the last four pictures are the late stage. It can be seen that at the beginning of the disease， smaller fading spots appear， In the middle stage， crater-like reddish brown or black protrusions are gradually formed. In the later stage， crater-like ulcer spots are formed.
Summary
The construction of the field image database of pitaya disease is an indispensable link in the research on the image recognition of different diseases of field pitaya. In this paper， computer image processing technology is used to segment and annotate images of pitaya of different diseases. An image database of the main diseases of pitaya in the field is established. Multi-labeling and hierarchical labeling of different diseases and different diseased grade of pitaya are realized to identify the categories of pitaya diseases， providing a basis for image recognition of the pitaya disease category and incidence. This database contains two types of data： the image of the pitaya disease and the video clip of the pitaya disease. The images and video clips of the database can provide training and test samples for the study of disease recognition algorithms. The establishment of a plant disease identification database not only greatly reduces the workload of disease prevention and control in agricultural production， but also lays a foundation for the future development of intelligent agriculture.
Acknowledgment
The authors thank Cui Yu， Fang Zongyu （Forestry College， Hainan University） for their help in image acquisition.
References
[1] LIU JW. Experimental research on introduction and cultivation of pitaya in solar greenhouse in Beijing area[D]. Yangling： Northwest A & F University， 2015. （In Chinese）
[2] BAURIEGEL E， GIEBEL A， HERPPICH WB. Rapid Fusarium head blight detection on winter wheat ears using chlorophyll fluorescence imaging[J]. Journal of Applied Botany and Food Quality， 2010， 83（2）： 196-203.
[3] ZHAO Y， HE Y， XU X. A novel algorithm for damage recognition on pest-infested oilseed rape leaves[J]. Computers and Electronics in Agriculture， 2012（89）： 41-50.
[4] MAHLEIN AK， STEINER U， HILLNHUTTER C， et al. Hyperspectral imaging for small scale analysis of symptoms caused by different sugar beet diseases[J]. Plant methods， 2012， 8（1）： 3-7. [5] CHENG MM， ZHANG GX， MITRA NJ， et al. Global contrast based salient region detection[C]//Proceedings of 2011 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）， Colorado Springs， CO， USA， 2011： 409-416.
[6] MA JC， LI XX， WEN HJ， et al. A key frame extraction method for processing greenhouse vegetables production monitoring video[J]. Computers and Electronics in Agriculture， 2015（111）： 92-102.
[7] PENG ZW， SI XL， WANG X， et al. Recognition of cucumber downy mildew based on image processing and fuzzy clustering method[J]. Chinese Journal of Agricultural Machinery Chemistry，2013， 34（2）： 212-216. （In Chinese）.
[8] WANG ML， NIU XJ， ZHANG HM， et al. Research on feature extraction and recognition technology of common diseases of wheat leaves[J]. Computer Engineering and Applications， 2014， 50（7）： 154-157. （In Chinese）.
[9] WANG N， WANG KR， XIE RZ， et al. Pattern recognition of corn leaf diseases based on Fisher discriminant analysis[J]. Chinese Journal of Agricultural Sciences， 2009， 42（11）： 3836-3842. （In Chinese）.
[10] GUAN ZX， TANG J， YANG BJ， et al. Research on rice disease recognition method based on image[J]. China Rice Science， 2010， 24（5）： 497-502. （In Chinese）.
[11] ZHANG JH， QI LJ， JI RH， et al. Cotton disease identification based on rough set and BP neural network[J]. Journal of Agricultural Engineering， 2012， 28（7）： 161-167. （In Chinese）.
[12] CHAI AL， LI BJ， SHI YX， et al. Tomato leaf disease recognition based on computer vision technology[J] .Chinese Journal of Horticulture， 2010， 37（9）： 1423-1430. （In Chinese）.
[13] HOLMES GJ， BROWN EA， RUHL G. What s a picture worth？ The use of modern telecommunications in diagnosing plant diseases[J]. Plant Disease， 2000（84）： 1256-1265.
[14] GONG GS， ZHANG M， QIN Y， et al. Construction of a digital standard library of plant diseases[J]. Journal of Yunnan Agricultural University， 2002， 17（4）： 361-362. （In Chinese）.