Abstract
At present, the quality of material life and richness of the Chinese residents have been improved. Some people have high requirements for quality of their personal lives and put forward the problem of health preservation. Honey, as a nutritious food, is deeply loved by people. There are a large number of trace elements in honey, such as VC, VA, VD, VB1, and VB2. The honey from the Chinese honeybee has a very high nutritional value and plays an important role in the pollination and reproduction of some plants. Therefore, the Chinese honeybee plays a very significant role in the ecological environment. Moreover, it is protected as the main species resource of the country, which also fully proves the importance of the Chinese honeybee. Chinese bees can survive in various ecological and geographical environments in China and have strong heat as well as cold resistance. They can survive in the hot environment in the south and withstand the dynamics of severe cold in the north. At the same time, they can make full use of a small number of honey sources and have strong resistance to a variety of diseases and pests. In fact, there will be a variety of insect invasion problems in the beehive culture of Chinese bees, and it is necessary to accurately detect various diseases and pests during the breeding of Chinese bees. However, there are a large amount of insect invasion and various disease sources in the breeding stage of the Chinese bees. Therefore, in this paper, we use a deep learning algorithm to detect the insect invasion of the Chinese beehive culture and analyze the bee colonies in six bee farms in the province of Sichuan. In addition, we measure the common insect and disease indexes of the Chinese bee and analyze the parasitism rate, microsporidia infection rate, virus infection rate, and virus infection titer of bee colonies in overwintering and spring breeding. The experimental results show that the anti-insect invasion situation of bees in the six bee farms is significantly different; however, the antimite ability is basically the same.
1. Introduction
The problem of insect invasion during beehive breeding has always been a problem faced by the beekeeping industries. Globally, this has also hindered the development of the beekeeping industries. There are four kinds of parasitic insects in bee diseases and insect pests, which are (i) fungi, (ii) viruses, (iii) microsporidia, and (iv) bacteria. According to relevant data, the annual loss of bee colony breeding caused by diseases and insect pests in China exceeds 30%, and the economic loss caused by honey production reduction reaches hundreds of millions of yuan [1]. Especially in recent years, China has paid attention to the quality and safety of honey products and gradually improved the standard of chemical drug detection materials. Chinese bee can survive in various ecological and geographical environments in China. It has strong heat resistance and cold resistance. It can survive in the hot environment in the south and withstand the dynamics of severe cold in the north. At the same time, it can make full use of a small number of honey sources and show strong resistance to a variety of diseases and pests. However, there will be a variety of insect invasion problems in the beehive culture of Chinese bees, and it is necessary to accurately detect various diseases and pests during the breeding of Chinese bees. However, there are a large amount of insect invasion and various disease sources in the breeding stage of Chinese bees that should be detected accurately. In terms of detection, machine learning-based prediction approaches have been well studied in the state-of-the-art literature. Various mechanisms and computational learning methods have been suggested to early detect similar diseases and reduce the losses of the bee industries [2, 3].
In this paper, we use the deep learning algorithm to detect the insect invasion of the Chinese beehive culture [4]. For the problem of insect invasion in Chinese beehive culture, this paper selects the deep learning algorithm for detection and analysis. It is one of the learning algorithms based on the human brain, such as the deep learning algorithm of the human brain, which uses the neural network to analyze and simulate the data. By combining with low-level features, it forms an abstract high-level representation of features and attributes. This paper uses a deep learning algorithm to simulate and analyze the problem of insect invasion in the process of Chinese beehive culture, summarizes various types of insect pests and diseases, analyzes the main types of insect pests and diseases attacking Chinese bees in hive culture in six farms, and studies the anti-insect and disease ability of Chinese bees in each stage.
n this paper, the process of intrusion algorithm based on deep learning is described in detail. Moreover, the operation process of the convolutional neural network (CNN) is introduced, and the intrusion detection technology is analyzed. Furthermore, the paper introduces the standard hive and ecological hive of Chinese beehive breeding and lists the detection technology of Chinese bee pest invasion. Through the analysis of six bee farms of Chinese beehive breeding in Sichuan province, the parasitic rate of bee mites and microsporidium spores is measured. Similarly, the virus infection rate and infection titer are calculated, and the invasion of Chinese bee pests in six bee farms in different periods are detected. The analysis shows that the virus infection is high. Furthermore, there are great differences in the invasion of insects such as microsporidia, and the parasitism of bee mites is basically the same. The main innovations of this paper are as follows:(i)The process of intrusion detection algorithm based on deep learning is described in detail(ii)The operation process of the convolutional neural network (CNN) model is introduced, and the intrusion detection technology is analyzed(iii)We introduce the standard hive and ecological hive of Chinese beehive breeding and lists the detection technology of Chinese bee pest invasion
The remaining of the paper is structured as follows: In Section 2, we discuss some related work from the existing literature. We provide a brief summary of each approach that has been studied to detect honeybee diseases. In Section 3, we suggest an intrusion detection algorithm based on the deep learning method. Section 4 is about the pest intrusion detection in Chinese beehive culture. Section 5 is about the experimental results. Moreover, an analysis of insect invasion detection in Chinese beehive culture is also provided in this section. Finally, Section 6 summarizes the paper and discusses directions for further research.
2. Related Work
Beehive is the main tool for breeding bees and is also the habitat for bees to reproduce. Moreover, various honey products produced by bees are also stored in beehives [2]. Various researchers have proposed different methods related to the beehive. Sun et al. pointed out that the style and size of beehives directly affect the survival and later reproduction of bees, which is conducive to the scientific management and breeding of beekeepers and greatly improves the economic benefits of the bee breeding [3]. Zhang pointed out that Chinese bees mostly live and grow in a cave nesting way in the natural environment. Other scholars pointed out that breeders use old beehives made of wickers, adobe, wooden boxes, and hollow trees to domesticate Chinese bees and make them adapt to life in beehives [4]. However, these kinds of artificial beehive also incur problems in the process of breeding Chinese bees. For example, most of them are made by themselves, and there is no standardized basis for reference. In addition, it is difficult to build nests and take honey, which is not conducive to the large-scale breeding of the Chinese bees, which hinders the development of the Chinese bee breeding industry and hits the enthusiasm of beekeepers [5].
Xu et al. used the Italian bee Langs live frame hive method to breed the Chinese bees, and remarkable results were achieved. At the same time, more than ten kinds of Chinese bee live frame hives were invented to maintain the stable development of China’s beekeeping industry [6]. Wang et al. pointed out that beehives directly affect the honey production of bees, and the number of bee colonies and honey production are different with different types of beehives [7]. For example, the superimposed beehive is based on reducing the capacity of a single Lang’s beehive and adding the number of relay boxes, so that the bee colony has a very strong swarm potential and is convenient to form a mature bee colony. Su et al. readjusted the nest frame, bee path, and quantity of Lang’s hive according to the breeding mode of the Chinese bee in Guangxi. After improvement, it has a wide range of applications in the market and is also loved by the majority of farms, especially for commercial breeding of the Chinese bee [8].
It is well-understood that the honeybee is the main economic insect. The main disease of a bee colony in the world is microcystis. Sener’s research data show that the two honeybee microbroods are very similar in morphology, and there are only differences in size. Moreover, N. apis is approximately 6.0 × 3.0 μm and N. ceranae is basically 4.7 × 2.7 μm or so [9]. Chemurot et al. used a transmission electron microscope to identify N. apis and N. ceranae based on the polar coil, and the authors believe that a scanning electron microscope can accurately measure the length and diameter of the microbrood [10]. Yudakhina measured that the size of honeybee microbrood in the Oriental Honeybee colony was approximately 4.68 × 2.19 μm. According to the data, it is speculated that the microsporidium in the Chinese bee may be N. ceranae [11].
3. Intrusion Detection Algorithm Based on Deep Learning
3.1. Deep Learning Algorithm Concept
Deep learning algorithms are widely used in computational and machine learning methods. These algorithms are used to deal with the problem of deep extraction of the data features [12]. When using artificial neural networks (ANNs) to deal with some very simple problems, there are only two key neurons, namely, (i) input signal neurons and (ii) output signal neurons. However, deep learning algorithms are not a double-layer structure. In fact, they try to hide a large number of linear or nonlinear elements in the double-layer structure, which can better process the input data (as shown in Figure 1).
Deep learning is an important part of the machine learning methods. In fact, deep learning is used to represent the multilayer network structure model. It simulates the basic characteristics of neurons processing and transmitting information through neural learning, which is shown in Figure 1 [13]. In general, multiple nodes are connected in the neural network calculation model, which shows the following two basic characteristics:(1)It effectively deals with the problem of weighted input value between each node and adjacent nodes(2)The weight has a direct impact on the transmission data strength between nodes, and the algorithm weight is adjusted by self-learning
The essence of deep learning is to represent some features on abstract data or very complex functions by using hierarchical and multilayer linear networks. During image processing, it is shown to train a large number of experimental samples to obtain learning image features. These can be obtained in the process known as feature extraction.
3.2. Convolutional Neural Network (CNN)
Generally, represents the convolution process. Formula (1) is a continuous one-dimensional convolution solution equation, and formula (2) is a discrete one-dimensional convolution solution equation.
In the previous formulas, f represents input and G represents the kernel function. In most scenarios, convolution uses mathematical language to explain the signals obtained from the real world. People collect signals in the real world by superposition or coupling, and the essence of superposition or coupling is convolution. That is, represents the signal collected by people from U1 and U2, whereas the UN original signal is obtained by the convolution process. For example, the sound heard can be regarded as the result of convolution between the student source and the propagation medium. Similarly, the seismic information received by the seismic sensor can be regarded as the result of convolution between the shock wave formed by the seismic source and the propagation medium of the array element. In addition, the image obtained by a remote sensing sensor is obtained by convolution of the ground object and the light source. The signals generated after collecting and convoluting ground object signals, microwave signals, and light sources are remote sensing images [14]. Based on the theory of convolution between signals and ground objects, people can get the required signal deconvolution from the ground objects [15].
3.3. Intrusion Detection Technology
The intrusion detection system can monitor the traffic on the network and the log data stored in the computer and judge whether there is a malicious attack through analyzing various characteristics of the data [16]. The main components of the proposed intrusion detection system are hardware and software. Figure 2 shows the deployment model of the intrusion detection system.
4. Pest Intrusion Detection in the Chinese Beehive Culture
4.1. Chinese Beehive
4.1.1. Standard Beehive
The standard beehive uses a standard 7-frame box with a thickness of about 2.5 cm. During the breeding of the Chinese bees, the relay box can be added according to the demand. The standard nest foundation, nest frame, and partition board are installed in the middle of the box. The top is covered with a box cover and cloth [17]. Finally, asbestos tile, straw, and wheat straw are covered on the cover, which are used to prevent rain and to shade from the natural disasters. Figure 3 shows a typical view of the standard box of the Chinese bee.
4.1.2. Ecological Beehive
According to the nature of the beehive, it can be called an ecological beehive. The basic principle of the beehive is to save space, produce honey with high maturity and quality, and cultivate strong bee colonies [18]. The components of the ecological beehive include multiple boxes, box covers, and box bottoms. Generally, the height, thickness, and width of the box are 10 cm, 3 cm, and 30 cm, respectively. It is a square wooden box with no cover at the top and bottom. After fixing and overlapping, a complete box is formed [19]. The box cover is made of a wooden board, with a length and width of 40 cm and a thickness of 3 cm. It is fixed on the top box, and asbestos tiles, straw, and wheat straw are covered on the box cover to prevent rain and to shade from the natural disasters. The length, width, and thickness of the box bottom formwork are 50 cm, 40 cm, and 3 cm, respectively. Four wooden strips with a width of 3 cm are used to nail into a square frame with an outer 30 cm. The number of ecological hives required is determined according to the number of cultured bees. Usually, an ecological hive is composed of three boxes, which is shown in Figure 4.
4.2. Techniques and Tools for Rapid Diagnosis of Chinese Bee Pest Invasion
In fact, each bee is small and most bees have the same external symptoms after diseases and insect pests. At present, the methods for diagnosing bee diseases are only used in the laboratory, which are described in detail, as illustrated in Table 1. The detection methods used include serological detection, microscopic examination, and molecular biology examination. In particular, the honeybee virus did not cause obvious symptoms for a long time after the invasion of diseases and pests, resulting in the formation of low-concentration recessive infection in the honeybee population. If the external environment meets the requirements, it will activate a large number of insect invasions, help them replicate rapidly in the host cells, improve the pathogenicity, and form a large-scale lethal honeybee virus. In most environments, there will be multiple compound infections or cross-infection transmission of different pathogens in a bee colony. The researchers proposed that Varroa mites can spread and carry the KBV virus [12, 20]. The harm caused by parasitic Varroa mites is serious after the number reaches medium scale, becoming a virus with high lethality. Therefore, once the honeybee epidemic appears on a large scale, due to the lack of directly applied systems or diagnostic and prediction tools, it is difficult for honeybee farmers to use effective measures to diagnose and control, so as to form the epidemic in various regions. At the same time, bee farmers will use various drugs to reduce the number of bee deaths, hoping that a drug can play a therapeutic role [21]. These drugs will remain in honey products, resulting in safety and quality problems of the bee products.
4.3. Pest Intrusion Detection in Chinese Beehive Culture
4.3.1. Test Colony
In this paper, six bee farms in six counties and cities, including Ya’an, Shimian, Xichang, Baoxing, Pingchang, and Mingshan, were selected as test colonies. The bee colonies and honey samples collected were transported to the Sichuan livestock and Poultry Quality Inspection Center in October 2019. After the selected test, colonies were sent to the Sichuan Inspection Center, where they were replaced with new beehives prepared before, and the test sites were placed alternately at random, in terms of number, and each colony was managed.
4.3.2. Measurement Index
In this paper, six bee colonies were sampled and measured on December 12, 2019, and March 12, 2020. The basic indicators selected in the determination of the bee colonies are (i) the number of microsporidia spores, (ii) the parasitic rate of bee mites, (iii) virus infection titer, and (iv) virus infection rate [22].
4.3.3. Test Method
(i)Parasitism Rate of Bee Mites. Approximately, 50 adult worker bees were randomly taken out from the outer nest spleen and partition of each bee colony, and the worker bees were frozen paralyzed. The parasitism rate of the bee mites at different sampling points can be obtained by checking the number of bee mites on them and calculating the ratio between the total number of bee mites and the total number of seals. After calculating the average value of the two sampling points, the parasitism rate of the bee mites in this bee colony can be obtained. During the overwintering period, all bee colonies were treated with mites to ensure that the bee colonies can survive the winter safely. The effect of bee colony mite control during overwintering was checked, and all bee colonies were treated during this time period [23].(ii)Microsporidium Spores. A total of 15 adult worker bees were randomly taken from the outer nest spleen and partition of each bee colony, the abdomen of 30 bees was cut off, and 30 ml of purified water was put into it for grinding. After mixing evenly, we dropped the ground liquid in the middle of the blood cell counting plate. Moreover, we selected 10 × in order to calculate the number of microsporidia spores with a 40x microscope. Finally, we counted each bee colony three times and averaged the three times to obtain the number of microsporidia spores of each bee in each bee colony [24].(iii)Virus Infection Rate and Titer. A total of 15 adult worker bees were randomly taken out from the outer nest spleen and partition of each colony. The number of bees in each colony was 30. They were frozen with liquid nitrogen and then ground. Then, 100 mg powder was taken out, and the total RNA was extracted with RNA Extraction Kit and passed through ReverTra Ace@qPCR RT Master Mix reverse transcription kit which can obtain cDNA. Next, we use SYBR@Green detection of infection titer and infection rate of viruses in honeybees with the real-time PCR MasterMix fluorescent quantitative PCR reagent. The viruses are CBPV, BQCV, DWV, IAPV, and SBV, respectively [25]. According to the plasmid standard curve, initially established by the Sichuan bee laboratory, the absolute quantification of the virus was judged and the calculation formulas are as follows:
Here, y is the CT value obtained through detection, X represents the number of virus log copies, and the virus infection titer is represented by 10xcopies.
4.3.4. Statistical Analysis
In this paper, the number of adult bees, microsporidium spores, daughter spleen, and virus infection titer in bee colony were analyzed by factor variance, and multiple comparisons were made by the Tukey method in one-way ANOVA. Moreover, the Chi square test is used to study the virus infection rate and bee mite infection rate. Similarly, SPSS 20.0 software is used for statistical analysis. From the statistical test and 95% confidence interval point of view, if the value is less than 0.05, there will be a very significant difference [20].
5. Analysis of Insect Invasion Detection in Chinese Beehive Culture
5.1. Swarm Mobilization
In this paper, six bee breeding bases in the Sichuan Province are selected to analyze the pest intrusion detection in the Chinese beehive breeding based on the deep learning algorithm, and the data on the day when the test bee colony enters the site are listed in Table 2. Among the 30 bee colonies, only four bee colonies in Ya’an bee farm have 11 nests and all the remaining other bee colonies have 10 nests [26]. The number of daughter spleens of each bee colony in different bee farms is very different from that of bee rats. The nest spleen of bee colonies in Ya’an, Shaoxing, Changxing, and Lanxi bee farms is symmetrical. Because there are a large number of nest spleens in the Ya’an bee farm, the number of daughter spleens and bees is significantly higher than that in other bee farms. However, on the Chunan bee farm, the spleen of the bee colony is less [27]. The number of bees on the Jiangshan bee farm is less, and the number of daughter spleens is also less than that in other bee farms.
To sum up, there is no significant difference in the number of daughter spleens and brush collars among bees on different bee farms (). Furthermore, there is no significant difference between the number of daughter spleens () and the number of adult bees () in each bee farm after entering the site. The colony potential in different bee farms during the spring breeding and overwintering is basically the same, which is shown in Figure 5.
5.2. Analysis of the Parasitism Rate of Bee Mites
The overwintering period is in the month of December every year. The bee colony in each bee farm checks the number of bee mites of adult bees and calculates the parasitism rate of bee mites. The result is that the average parasitism rate of bee mites on the Jiangshan bee farm is higher than that on other bee farms, but the difference is quite small (). The specific data are shown in Figure 6. Since honeybees can thoroughly control mites in honeybee colonies during the overwintering periods, the number of bee mites in honeybees is small before sampling in March. The surface inspection results of adult bees show that there are no bee mites, indicating that the parasitic rate of bee mites in all honeybee colonies is 0.
5.3. Analysis of the Microsporidian Infection
During the overwintering period in December, the number of microsporidia spores in the bee colony ranged from 0 to 7.0 × 105/bee intervals. It was observed that five groups of the bees in Xichang bee colony were not infected with microsporidia. The detection of microsporidia infection in each group of bees in Ya’an bee farm is very serious. After calculation, the average number of spores infected by each bee is 2.7 × 105, which was higher than that of other bee farms (). Figure 7 shows the number of infected spores of the bees in other bee farms. In March of the next year, the number of microsporidia infected spores in the bee colony in spring was 0 to 2.3 × 106/bee. Excluding Ya’an bee farm, the number of microsporidia in the body of bees in the remaining five bee farms remained elevated. In particular, the average number of spores in each bee body in Baoxing bee farm increased rapidly from 0.2 × 105 in the overwintering period to 9.1 × 105 in the spring breeding period, which was higher than that in Shaoxing and Ya’an bee farms. Moreover, there was no significant difference in the remaining bee farms.
5.4. Analysis of the Virus Infection Rate
During the overwintering period in December and the spring breeding period in March, five common viruses appear in each Chinese beehive farm, namely, bqcv, DWV, CBPV, SBV, and IAPV. The infection rate of the bqcv virus in each bee farm is shown in Figure 8. During the overwintering period of bees in each bee farm, the infection rate of bee colonies infected with the bqcv virus is high, which can reach as high as 60% to 100%. Similarly, the infection rate of the CBPV virus is 80% to 100%, the infection rate of the DWV virus is 100%, and the infection rate of the IAPV virus is 20% to 80%. This should be noted that the lowest infection rate is by the SBV virus. From the overwintering period to the spring breeding period, the bee farms did not contain the bqcv virus and the infection rate of the other four viruses continued to decrease, significantly. In the spring breeding period, the infection rate of the bee colony infected with the IAPV virus in the Mingshan farm is approximately 60%, which is higher than that in ordinary farms (0%). The infection rate of bee DWV virus in the Xichang farm is 100%, which is higher than that in the Baoxing farm.
5.5. Virus Infection Titer Analysis
By measuring the virus titer of infected species in six bee farms, selected in this paper, the infection titer infected with the bqcv virus is shown in Figure 9. When each bee farm is in the overwintering period, only the SBV virus infection titer varies greatly. It could be easily observed that only the Ya’an farm is not infected with the SBV virus, which is lower than the other five farms. The SBV virus infection titer of bees in the farms is much higher than that in Changxing and Lanxi farms [28]. During the breeding period, only five bees in Mingshan could not be detected to be infected with the SBV virus.
6. Conclusions and Future Work
Bees are social insects, and the external environment directly affects the health status of bee breeding groups. Therefore, when detecting the invasion of Chinese beehive breeding pests, we should select the best area and complete it based on the deep learning algorithm. Through the analysis of virus infection and microsporidia infection in six Chinese beehive breeding plants in the province of Sichuan, we observed that the parasitism of bee mites is basically the same. During the study, the bees in each bee farm have no strong natural antibee mite ability. At the same time, the degree of microsporidia infection in each bee colony is also very different. Microsporidia infection in some bee farms is very serious. In addition, each bee farm also has a serious phenomenon of virus infection; however, the antivirus ability is basically the same. This phenomenon is directly related to the parasites such as microsporidia and bee mites. In this paper, we analyzed the virus infection and insect invasion of bees in six Chinese bee farms by using the intrusion detection of the deep learning algorithm, so as to provide insect protection methods for bee breeding and management in the future. This will help to better prevent and control the infection degree of microsporidia and bee mites and improve the ability of bee breeding to resist diseases and insect pests.
In the future, we will propose other intrusion detection algorithms using more sophisticated deep learning methods such as graphical convolution network (GCN) and attention based networks. This is also well-understood that learning based prediction mechanisms are typically grounded on the dataset. In other words, the deep learning based prediction algorithm training may take expressively longer periods if the size of the data set is huge and vice versa. Therefore, it is essential to keep the performance aspect in mind when designing similar algorithms. In the future, we will consider reducing the computational time needed to train the model either through aggregation methods or using more robust algorithms. Finally, the impacts of hot, cold, winter, and spring weathers should be further investigated.
Data Availability
The data used to support the study are included in the paper.
Conflicts of Interest
The authors declare that there are no conflicts of interest for the publication of this paper.