Research on classification and identification of library based on artificial intelligence

Abstract

With the popularization of open borrowing, it is urgent for major libraries to explore new collection classification and identification systems to assist or replace librarians in their daily book cataloguing work. The messy books can be marked according to invisible colors, which puts forward a new method of library classification. In the algorithm used to detect image edges, by looking for the traditional sensitivity measurement system, the best parameters are applied to the wrong book image edges, and the test results are verified to find the best parameter range. A two-dimensional algorithm is used to segment the information element matrix to reduce sound. The results show that the image edge test is effective and reduces the influence of noise on the image test. The self-adjustment control algorithm of a single neuron is used to control the robot orbit, which makes the robot not only react quickly, but also reduce a small part of the error in meeting the actual needs of the robot. The effect on the result is getting smaller and smaller.

Keywords

Artificial intelligence ant colony algorithm library classification recognition

1 Introduction

With the continuous development of our economy, people’s pursuit of life is becoming higher and higher, which promotes the continuous development of open reading. With the popularity of this form of reading and borrowing and the constant recognition of this kind of borrowing, open simplicity has become a phenomenon. Of course, the emergence of this phenomenon is of great benefit, but it is inevitable that there will also be some chaos. For example, many books in libraries are placed in the wrong place. This not only affects the effective circulation of books, but also brings management problems to the effective circulation of books. The digitization and automation of traditional work is a priority for the library sector, and many libraries are close to being able to return books through computers, which greatly facilitates users. However, in fact, the author finds that the current digitization process of libraries raises an unsolved problem, that is, the collection has not been classified and identified. The classification of digital collection also requires librarians to manually classify and identify books every day, and to classify books returned by readers, which is not only inefficient, but also easy to lead to physical and mental fatigue and mental stress. In addition, libraries must classify and identify books manually. Borrowing makes it difficult to prevent the clutter of books on shelves, which not only prevents readers from finding the required books in time, but also hinders the provision of library services. To this end, it is urgent to study the new classification and identification system of library collections in order to assist or replace librarians in dealing with daily errors and books. The implementation of the system will be important and useful in assisting librarians to improve their management, reduce workload, improve efficiency and improve the overall quality of library services.

This paper proposes a new method of library classification and identification according to invisible color marking. Improve the efficiency and accuracy of library classification and identification, and maintain the traditional library borrowing methods. The invisible color marking technology is very simple, and it is cost-effective and easy to introduce automatic robots. at the same time, the test image border is based on ant algorithm. these algorithms seek the best parameter collocation range by understanding the traditional TPS system, applying the best parameters to the wrong book image border to detect and verify the test results, and dividing the model into plasma. Information elements use two-dimensional algorithm, the results show that image edge testing is effective, reducing the impact of noise on image testing. The self-adjustment control algorithm of a single neuron is used to control the robot orbit, which makes the robot not only react quickly, but also reduce a small part of the error in meeting the actual needs of the robot. The effect on the result is getting smaller and smaller.

2 Related work

As Wu Yunxia and Xia Ming mentioned in the literature [1], nearly 25 libraries have used this classification system by 2014. Although these libraries are all in the United States, it is not difficult to find that there are many other libraries in other countries using these systems. Some libraries in Australia and Canada, for example, are already in use [2]. The literature, many scientists have demonstrated the application of robot technology on library management line, which can use bar code reader to scan the specific book information collected and compare the scanning information with the information entered by the user. It uses this method to find books, wrong books can be rearranged, and robot data is used to keep books in order. Literature [3], some scientists in Singapore have designed a fully automatic robot that can travel freely through the library during the day or at night, and can help librarians sort. This feature is achieved by scanning FRID labels to get lost and misplaced books, which is very accurate, according to some data, sometimes as accurate as 99%. As described in the literature [4], the Humboldt University Library in Germany designed a mobile robot that can travel to and from warehouses, transport and retrieve publications, when cross-storey transportation is required, the robot moves to the elevator entrance, elevator doors have metal detectors, Elevator automatically opens, once the robot leaves the elevator, the elevator automatically closes, its basic function is to automatically transmit between the office and the functional book area that requires bookshelves and classifications. The literature [5] shows that Johns Hopkins University SuthakornJ and others have designed a robot system, which is a robot that searches the books that readers need on one bookshelf and automatically puts them on another device. It can automatically flip the web page and enable readers to read from afar. Literature [6] shows that the TomizawaT of Tsukuba University in Japan designed a remote operating system to enable people to read books remotely with mobile manipulators and to consult library books. According to the literature [7], Professor Chen Jinyuan of the School of computer and Technology of Nanjing University and his team have successfully developed a new intelligent robot for library management. So that books and their inventory can be quickly and accurately classified and help readers find the right books. Intelligent robots, including scanner equipment, intelligent computer online query system and intelligent bookshelf, can help to find information obtained and stored through robot scanners in the library and can be shared in real time. Readers can obtain accurate information about the required books through the search system. Literature [8] examples of libraries that use science and technology to promote change, such as the Shanghai Library, the Hubei Library, the Hangzhou Xinhua Library, the Chengdu University of Technology Library and the Beijing Library; innovative advice to readers, such as the Shanghai Library Tooling “, The library can also help readers understand time, routes, etc., specially designed to meet the needs of children’s reading areas, a robot that can advise readers and tell stories for children; Ningbo University Park Library has introduced a robot that can also provide readers with advice. While waiting for a long time, readers say hello to readers and introduce the basic information of the library. Chengdu University of Technology Library launched a commercial version of the “small” robot to provide consulting services, effectively extending the time and space of library consulting services.

Document [9] introduces the book catalogue robot developed by Shanghai Science and Technology Company. The robot can classify books by some techniques, that is, manually labeling each book, and then identifying it as a reader robot running according to a predetermined route. After finding the bookshelf for cataloguing, it automatically identifies each book on the bookshelf and catalogs it. During cataloguing, the robot sends real-time information to the server, and librarians can display data through various models such as computer or telephone. Document [10] introduces the robotics technology developed by Guangzhou Robot Co., Ltd., which can be used in a variety of fields, including consulting, interpretation, information collection, wireless navigation, language interaction, interpersonal interaction, tactile interaction and automatic replenishment, in order to facilitate the management of library classification and identification [11 –13].

3 Design of bookmark based on TRIZ principle

The necessity of innovative icons proposed in this paper can solve the following problems: displaying information about book logos, facilitating readers’ reading, facilitating librarians’ organization and classification, and allowing them to be applied to the intuitive identification of machines [14 –16]. Make it have reliable recognition accuracy and ensure the economy of the object.

3.1 Rules for the arrangement of invisible colour labels

Guided by the TRIZ principles, this document proposes a new model of invisible bookmarks, printing bookmarks with ink that is not visible in color. in normal light, the ink is opaque, you can only see the cable number, and in ultraviolet light, you can see the color sequence corresponding to the book number. thus, the color of the cover is avoided in relation to the book label sequence [17, 18]. This marking not only makes it easier to classify and borrow books, but also makes it possible to borrow books by asking for book numbers. It also helps staff and machines identify books and update colors through electronic data. The ultraviolet fluorescent ink selected in this paper is a kind of luminous substance, which absorbs the energy under the sun, stores it, and finally displays it in the form of visible light [19 –21]. Fluorescent ink can only glow under ultraviolet light, and its length is 365 nm wavelength, which can not only display its color, but also absorb the energy of light and enhance its own light output [22, 23]. We printed colorless ultraviolet fluorescent ink with a total length of 5 cm per unit on the original sealed unmoved monochromatic label. The color label structure is shown in Fig. 1.

Fig. 1

Hidden Color Bookmark.

A rectangular label of 5 cm ×2.5 cm is determined according to the size of the paper label. There are five colors. From top to bottom, it represents the classification of 1, 2, 3, 4 and 5. The first five characters of the book number are color and alphabet and color and number comparison table, respectively. Comparison of colors and letters as show in Table 1, Comparison of colors and figures as show in Table 2.

Table 1

Comparison of colors and letters

Letters	Color name	Letters	Color name
A	Big Red	N	Big Red/
B	Rose red	O	Rose red/
C	Orange	P	Orange/
D	Lemon yellow	Q	Lemon yellow/
E	Dark green	R	Dark green/
F	Olive green	S	Olive green/
G	Medium Blue	T	Medium Blue/
H	Sky Blue	U	Sky Blue/
I	Blue	V	Blue/
J	Yellow	W	Yellow/
K	Black	X	Black/
L	Grey	Y	Grey/
M	Purple	Z	Purple/

Table 2

Comparison of colors and figures

Figures	Color name
0	Big Red
1	Rose red
2	Orange
3	Lemon yellow
4	Dark green
5	Olive green
6	Medium Blue
7	Sky Blue
8	Blue
9	Yellow

3.2 Production process of invisible colour book labels

The specific operational measures for the book registration process listed in Fig. 2 include: the entry of the basic library materials into the database by the librarian; the classification of books; and the choice of the first, second, third, fourth and fifth catalogues of the library, the finalization of the numbering according to the organizational rules for the application number, and the listing of the accounts for each category (including the alphabet and the account number), the finalization of all the necessary document numbers and the entry of the book numbers into the database unit [24 –26].

Fig. 2

Block diagram of book registration module.

3.3 Identification and classification of books based on invisible colors

First of all, the robot starts from its initial position, uses the electromagnetic guidance robot installed on the ground, starts from the starting point signal, tracks the electromagnetic installed on the ground, makes a judgment at any time during the movement, and stops the guidance immediately if there is an end signal on the electromagnetic conductor.

Before entering each bookshelf, the robot determines the number in front of the bookshelf in a fixed position and, by adjusting the number, draws the book from the book number corresponding to the number on the bookshelf, then moves the robot to a fixed position in front of the next bookshelf, then stops to adjust automatically. The robot camera looks at the books on the bookshelf, records the color sequence of the books on the bookshelf, detects and extracts the characteristics of the book from the color series image of the library number using the image processing algorithm, and, in addition, as shown in Fig. 3, the standard color series obtained by the identification number, uses the extracted color distribution, The standard color of the book section listed in the figure fills the image systematically.

Comparing the book number in the book collection with the color sequence in the standard color series, if the two colors are the same, it will be concluded that there is no incorrect book sequence and the robot will be guided along the electromagnetic direction to the next position; If the scale of the RGB is different, it is preliminarily concluded that the order of the book memory in this area is disordered. In this case, comparing the color series with the standard image series, we can see the edge characteristic colors of different series, that is, incorrect sequence, edge distribution of error series color, calculation of error series, width of error classification, and location in this area, and analyzing the color sequence of error classification, comparing color series with standard database and its classification.

According to the wrong size and order of the books, put the books in the wrong order, adjust the machine to take the sequence of the wrong books from the bookshelf, arrange them according to the color order of the books, and put them on the corresponding bookshelves. Each time an error is found on the bookshelf or book order, the attached bookshelf is judged to be full. The robot wants to enter the fast input electromagnetic boot system on the ground, record the currently determined shelf number and return node location, retreat and take books from the bookshelf, and then return the node through the fast output electromagnetic boot system. And continue to classify the shelves of the program. If the bookshelf is not full, keep working until the next working node gets the book back.

When the robot detects the terminal signal in the ground electromagnetic conduction system, they download the relevant books from the shelf, then return to the starting point and stop working.

Fig. 3

Schematic illustration of the standard color sequence.

The specific flow of book identification and sorting is shown in Fig. 4 below.

Fig. 4

Flowchart of book detection method.

4 Image Edge detection and extraction based on ant colony algorithm

4.1 Image edge extraction method for books

Image marginalization is a very important link in the process of image processing. It embodies an important feature of computer and visual processing. Image marginalization is mainly used in image processing. It is worth mentioning that the most important feature of image edge is a series of extraction, recognition and measurement of image through computer vision algorithm, and the boundary of image is extracted by computer vision. By extracting the boundary to further identify and measure the image, there are many common image edge methods, the most important is the differential operation method, which can directly extract the edge of the image from the high component image.

The most classical image edge detection can be divided into two main types, the first type is edge detection, this edge detection is based on the first order differential edge detection operator, the second type is based on the second order differential operator edge detection.

The first order differential operator mainly uses the difference approximate differential method to carry on the first order derivation to the image which wants to marginalize, through some derivation further obtains the edge image matrix, through the edge image gradient matrix, Gradually maximize the gradient edge and find out that the maximum value is the edge of the image. For example, the Roberts operator was proposed by famous scientists in the 1960 s. The basic principle of this operator is to derive the object. The edge of the target can be found more accurately by finding the maximum value of the gradient matrix. By establishing an image template of 2×2, in which two diagonals are found, the pixel difference (such as formula 1) is obtained by gradient difference between the two diagonals, the gradient of the image is calculated on each diagonal.

$\nabla f = (f (x, y) - f (χ - 1, y), f (x, y) - f (x, y - 1))$ (1)

The edge of the image is mainly calculated by the second order differential equation. The second derivative is obtained by calculation, and then the second derivative is further analyzed. The second derivative zero crossing is an important proof of analysis. It can detect the edge of the image. The zero crossing of the second derivative not only reflects the zero value, but also reflects a limit value of the second derivative. For example, the Log operator was proposed by famous scientists in the 1980 s. This operation is mainly carried out by Gao Si filter. After the operation, the peak value of the first derivative in the operation is retained, the limit value of the first derivative is found by the first derivative, and finally the zero point is used to determine its exact position.

LOG operators are as follows:

$Δ^{2} G = \frac{\partial^{2} G}{\partial^{2} x} + \frac{\partial^{2} G}{\partial^{2} y} = - \frac{2 \partial^{2} + x^{2} + y^{2}}{2 π σ^{6}} e^{- (x^{2} + y^{2}) / 2 σ^{2}}$ (2)

The detection results of the classical edge detection operator are greatly affected by the surrounding noise. The smaller the influence of the surrounding noise, the better the detection effect of the edge detection operator. When the surrounding noise is more affected, the effect of edge detection is worse, so that the effect of edge detection can not reach the previously estimated level, so that the image can not be accurately detected, so the image edge can not be accurately detected. Pixel point four neighborhood diagram as show in Fig. 5, Pixel point eight neighborhood schematic as show in Fig. 6, Plane projection of a two-dimensional rectangular graph as show in Fig. 7.

Fig. 5

Pixel point four neighborhood diagram.

Fig. 6

Pixel point eight neighborhood schematic.

Fig. 7

Plane projection of a two-dimensional rectangular graph.

4.2 An ant colony algorithm model for image edge detection

In the process of processing image edges, ant colony algorithm can be fully applied. The application principle of ant colony algorithm is to regard the applied image as a picture without boundary and direction. In this picture, the specific pixel nodes of each picture can be transformed into nodes of ant motion. This translates into a very large-scale image problem. It is worth noting that in the process of image edge detection and extraction, the noise is also affected by noise. Therefore, the factor can be changed into a gray gradient in the image by expectation, which can further reduce the effect of noise on image edge extraction. The image can also be removed first, which can further reduce noise pollution and continuously enhance the accuracy and accuracy of ant colony algorithm. Firstly, the detected images are classified one by one by filtering, and divided into several sizes of images. Several ants are randomly placed in each image, and the number of randomly placed ants must be the same as the pixel content in each image. that is, the number of ants represents the number of pixels in this image. the distance between ants in each image is shown in Fig. 1. the distance and probability of ant movement are further calculated and evaluated by computer algorithm, and the next image node to be moved is also predicted. Through the probabilistic algorithm, more ants want to choose the moving path and image. Because ants walk different image nodes to reflect different pixel moving processes, they have to further pass the two-dimensional algorithm. The pixel information in the image is calculated. Through further calculation, the edge points of the image are gradually found out, and then each edge point of the image is connected to form a more accurate edge image.

The application steps of ant colony algorithm are as follows:

First of all, the image should be treated with noise removal, which can further reduce the effect of noise on image edge extraction. The image can be photographed by camera. Because the camera also contains many complex sensors, it is easy to mix some noise even if it is quiet during shooting. The maximum possible impact of noise on the image is minimized and the error is continuously reduced.

The image can then be classified by filtering, dividing the image into several small image regions, placing the same number of ants in each small image region as the number of pixels, that is, representing the number of pixels through the number of ants. The information matrix of the image is established by the algorithm. At the same time, the limit value should be set for the information matrix of the image. The limit value includes the maximum value and the minimum value, and the number of updates of the pheromone matrix should be set. Ant colony algorithm time is also increasing, which greatly reduces the image processing rate.

Next, the ants placed in each image area are calculated to estimate their moving direction and distance, and the moving probability of each ant is calculated. If an image area is selected by more ants, the more pheromones are left in the image area. The specific algorithm formula is shown in formula 3.

$P_{(i, m)}^{(n)} τ_{(i, j)} = \frac{{(τ_{i, j}^{(n - 1)})}^{α} {(η_{i, j})}^{β}}{\sum_{(i, j) ɛ Ω_{(i, ω)}} {(τ_{i, j}^{(n - 1)})}^{α} {(η_{i, j})}^{β}}$ (3)

Information elements are gradually produced with the movement of ant colonies. α represents the important value of various information elements, the larger his value, the more the accumulated value of information elements produced by ants, that is, the stronger the search ability of the ant, Ant population will move by low concentration of information elements, such as high concentration of information elements.

The gradient value of the image is calculated by formula 4: $η_{i, j} = \frac{1}{z} v_{c} (I_{i, j})$ (4)

The movements of ants can also be further recorded by setting tabu tables, which prevent ants from taking the same path and from making repeated path explorations. To improve the efficiency of ant population search, tabu should also be combined with image processing. When the image processing loop is over, the tabu should be reset again.

After the end of the graph processing cycle, the information elements left by the ant population should be recorded, and the information elements should be systematically updated before the next graph processing cycle is opened. After updating, the information element content of pixel is further calculated according to the gradient value of information element, mainly by formula 5. $τ_{i, j}^{(n)} = (1 - ρ) \cdot τ_{i, j}^{(n - 1)} + ρ \cdot Δ_{i, j}^{(k)}$ (5)

After completing a series of algorithms, the information value needs to be re-set, the image is re-classified, and the target edge is segmented continuously. In this paper, a two-dimensional algorithm is used to segment the image edge systematically, so that the influence of noise on image processing can be reduced by resetting threshold and pheromone matrix. The calculation efficiency of this method is greatly improved. The calculation process is greatly optimized.

4.3 Threshold selection based on two-dimensional OTSU algorithm

Compared with the traditional ant colony algorithm, the two-dimensional OTSU algorithm is more concise, has a very powerful adaptability, and can minimize the impact of noise on the edge of the image. The ability to resist noise is very high.

Although the gray value of the pixel in the target region is very similar to that in the adjacent region, the gray value of the pixel at the edge of the image is very different from that of the other gray values. The probability is calculated by formula 6 and formula 7.

$p_{背景} = \sum_{t = 0}^{s} \sum_{j = 0}^{t} P_{ij}$ (6) $P_{目标} = \sum_{i = s}^{L - 1} \sum_{j = t}^{L - 1} P_{ij}$ (7)

Background Mean Vector u₀: $u_{0} = {({\bar{u}}_{0 i}, u_{0 j})}^{T} = {(\sum_{i = 0}^{s} \sum_{j = 0}^{t} \frac{i p_{ij}}{w_{0}}, \sum_{i = 0}^{s} \sum_{j = 0}^{t} \frac{{jp}_{ij}}{w_{0}})}^{\bar{T}}$ (8)

u of target mean vector₁:

$u_{1} = {({\bar{u}}_{1 i}, {\bar{u}}_{1 j})}^{T} = {(\sum_{i = s + 1}^{L - 1} \sum_{j = t + 1}^{L - 1} \frac{i p_{ij}}{w_{1}}, \sum_{i = s + 1}^{s} \sum_{j = t + 1}^{t} \frac{{jp}_{ij}}{w_{1}})}^{\bar{T}}$ (9)

Total mean vector of 2D histogram: $u = {({\bar{u}}_{i}, {\bar{u}}_{j})}^{T} = {(\sum_{i = 0}^{L - 1} \sum_{j = 0}^{L - 1} i p_{ij}, \sum_{i = 0}^{L - 1} \sum_{j = 0}^{L - 1} i p_{ij})}^{\bar{T}}$ (10)

Definition of discrete degree matrix between target and background:

$S_{B} (s, t) = w_{0} {(u_{0} - u)}^{T} (u_{0} - u) + w_{1} (u_{1} - u)^{T} (u_{1} - u)$ (11)

$(S, T) = arg max_{1 < s < L - 1} max_{1 < t < L - 1} {Trac (S_{B} (s, t))}$ (12)

4.4 Simulation results and analysis

By extracting and detecting the edge of the image, the recorded results are shown in Table 3.

Table 3
Experimental results of optimal parameters

Ant colony algorithm edge extraction time (s)

Number of experiments One II III IV V Average time

Traditional Ant Colony Algorithm 1369.0 1410.6 1525.4 1439.6 1460.3 1441.0

Ant colony algorithm 424.13 425.16 430.54 423.32 425.87 425.80

	Ant colony algorithm edge extraction time (s)
Traditional Ant Colony Algorithm	1369.0	1410.6	1525.4	1439.6	1460.3	1441.0
Ant colony algorithm	424.13	425.16	430.54	423.32	425.87	425.80

Through the further analysis of the experimental results, it is found that the ant colony algorithm can find out the edge of the image very accurately, and accurately measure the distance of the wrong frame book, and accurately clip the wrong frame book.

5 Simulation experiment and result analysis

5.1 Kinematics simulation of book sorting robot

Manipulators have the function of simulation, but this model may require them to move in only three directions, so the study of these models may be relatively simple. Because they only need to meet the needs of the more basic, not too many multi-directional complex motion. This software has relatively few functions compared with other software, and its function is relatively weak in modeling. Of course, if it is compared with other 3D software, we can design the model of manipulator by some data, then model by some software, build the basic model, add some software architecture and model to make the robot model more real, then save it into a certain format, and finally import it into the environment they should exist for running test. Table 4 records the motion pairs of the mechanism model.

Table 4
Motion pairs of mechanism models

Serial number Component Constraints

1 and 2 Rotary disc and X axis guide Mobile pair

1 and 6 Rotating disc and fixing base Rotary pair

2 and 3 Y shaft guide rail and Y shaft slider Mobile pair

3 and 4 Y axis slider and Z axis Fixed pair

4 and 5 Z shaft and end actuators Mobile pair

6 Base and earth Fixed pair

Serial number	Component	Constraints
1 and 2	Rotary disc and X axis guide	Mobile pair
1 and 6	Rotating disc and fixing base	Rotary pair
2 and 3	Y shaft guide rail and Y shaft slider	Mobile pair
3 and 4	Y axis slider and Z axis	Fixed pair
4 and 5	Z shaft and end actuators	Mobile pair
6	Base and earth	Fixed pair

During the whole process of imitation, we need all kinds of coordinated operation, and we need to ensure that each function can run correctly. In this paper, we use step function. $St e p (χ, χ_{0}, h_{0}, x_{1}, h_{1})$ (13)

The x in the formula is the independent variable of the function, x₀ h value of independent variable representing the starting point of step function₀x representing the value of the function at the starting point of the step₁Represents the value of the independent variable at the end of the step, h₁Represents the value of the function at the end of the step.

This simulation is a complete robot process of retrieving and placing books, that is, when the robot finds the target book, the machine begins to move, the book returns to its original position, and the simulation time is set to 80 seconds.

A displacement function of three sliding joints and one rotating joint of the manipulator is set, in which the driving function of the X axis is:

$\begin{matrix} STEP (time, 0, 0, 5, 110) + STEP (time, 30, 0, 35, - 110) \\ + STEP (time, 45, 0, 50, - 110) + STEP (time, 60, 0, 65, 110) \end{matrix}$ (14)

Y axis drive function is:

$\begin{matrix} STEP (time, 0, 0, 5, 110) + STEP (time, 30, 0, 35, - 110) \\ + STEP (time, 45, 0, 50, - 110) + STEP (time, 60, 0, 65, 110) \end{matrix}$ (15)

Z axis drive function is: $STEP (time, 0, 20, - 800)$ (16)

The rotation drive function is:

$STEP (time, 35, 0, 45, 90 d) + STEP (time, 65, 0, 75, - 90 d)$ (17)

After the simulation, the simulation results can be clearly viewed in the ADAMS/PostProcessor of the post-processing module, and the simulation results can be transformed into tables or animation forms to accurately reflect the relevant characteristics of the model.

The simulation results show that the components of the manipulator can run smoothly without sudden disturbance, and for books on library shelves at different heights, the machine can reach the predetermined book collection place to meet the requirements of the library. This is the real work of the robot.

PID control and Matlab simulation based on single neuron network

The proportion (P), integral (I) and differential (D) of deviation form the control quantity through linear combination, and control the controlled object, so it is called PID controller. PID controller can adjust the whole system signal deviation through the output port to make the actual output value consistent with the theoretical value, which is a negative feedback control, so that the actual value of the output of the accused object is consistent with the value. The control method is the most widely used method in closed circuit control system. The control performance is stable, the control accuracy is high, the parameters are easy to adjust, but it is relatively rough. X Axial displacement and velocity variation curve as show in Fig. 8, Y Axial displacement and velocity variation curve as show in Fig. 9, Z Axial displacement and velocity curves as show in Fig. 10.

Fig. 8

X Axial displacement and velocity variation curve.

Fig. 9

Y Axial displacement and velocity variation curve.

Fig. 10

Z Axial displacement and velocity curves.

PID control is a linear controller that can be x according to input values_in(t) y) to actual output value_out(t) Constitute control deviations: $e (t) = x_{m} (t) - y_{0 m} (t)$ (18)

PID control law is: $u (t) = k_{P} (e (t) + \frac{1}{T_{i}} \int_{0}^{t} e (t) d t + \frac{T_{d} de (t)}{d t})$ (19)

T in the formula_iAn integral time constant; a T; a_dIt is a differential time constant.

Form 19 can be further written: $u (t) = k_{P} e (t) + k_{i} \int_{0}^{t} e (t) d t + k_{d} \frac{de (t)}{d t}$ (20)

Considering the stability, response speed, overshoot and steady-state precision of the control system, the three parameters are as follows:

k of proportional factors_pThe function is to speed up the response speed of the whole system and improve the adjustment accuracy of the system.

k of integral coefficients_iThe function is to eliminate the steady-state error of the system.

k of differential coefficients_d Its effect is to improve the dynamic of the system, its function is to control the deviation change in any direction in the process of coping, and to predict the deviation change.

Computer control is a kind of sampling control, which can only be calculated according to the deviation of sampling, not according to the deviation of sampling. Sample. Therefore, the continuous software control algorithm can not be used directly, it must be separated KT the sampling points represent continuous time T, the digital integration points replace the time integration points, and the linear approximation points replace the time difference. namely:

${\begin{matrix} t \approx kTk = 0, 1, 2, \dots \\ \int_{0}^{t} e (t) dt \approx \sum_{j = 0}^{k} e (jT) \\ \frac{de (t)}{dt} \approx \frac{e (KT) - e ((k - 1) T)}{T} \end{matrix}$ (21)

The T is the sampling period, and the discrete expression can be obtained by normalizing PID sampling period:

$\begin{matrix} u (k) = & k_{p} [e (k) + \frac{1}{T} \sum_{j = 0}^{k} e (j) + T_{d} (e (k) - e (k - 1))] \\ = k_{p} e (k) + k_{i} \sum_{j = 0}^{k} e (j) + k_{p} (e (k) - e (k - 1)) \end{matrix}$ (22)

According to the recursive principle, the incremental PID control expression can be obtained:

$\begin{matrix} Δ u (k) = k_{p} (e (k) - e (k - 1)) + k_{i} e (k) \\ + k_{d} (e (k) - 2 e (k - 1) + e (k - 2)) \end{matrix}$ (23)

Personal neural regulator is a linear neural adaptive neural network with simple structure and strong control structure. the crude three neuron rights into the neural network are the three coefficients of the PID algorithm, while the SYNAPSE rights adjusted according to the network learning algorithm come from the automatic organization of adjustable control PID proportion, integration and calculation until the controller stabilizes.

The active action of single neural network is linear, that is, the output of nerve and the induction of local control are linear, and the output of nerve is: $Δ u (n) = K \cdot v (n) = K (\sum_{j = 1}^{3} ω_{j} (n) x_{j} (n))$ (24)

In practical engineering application, the online correction part of the weight coefficient of neural network learning principle is trimmed with practical experience. The process is as follows:

${\begin{matrix} ω_{1} (k) = ω_{1} (k - 1) + η_{ρ} e (k) u (k) (e (k) + Δ e (k)) \\ ω_{2} (k) = ω_{2} (k - 1) + η_{i} e (k) u (k) (e (k) + Δ e (k)) \\ ω_{3} (k) = ω_{3} (k - 1) + η_{d} e (k) u (k) (e (k) + Δ e (k)) \end{matrix}$ (25)

5.3 Analysis of simulation results

The simulation shows that the response time of PID localization algorithm is about 22 seconds, while that of PID incremental algorithm is about 1.5 seconds, while that of single nerve to PID control algorithm is about 0.15 seconds. This shows that the response time of the customized algorithm for individual applications is my point. When different learning rules are applied to the PID adaptive algorithm control of a single neuron, the control errors of the whole system may be slightly different. The experimental results show that the best way to control the nervous system is to use LNS learning rules to adjust the right value of the nervous system.

6 Conclusion

This document classifies books according to color marks, reduces the burden of librarians classifying and classifying incorrect books, improves the efficiency of book circulation, and designs an invisible color marking program. On this basis, the robot of automatic identification and classification of incorrect books is studied, and the workflow is provided. Improving the traditional ant algorithm is helpful to understand the best combination of ant algorithm parameters. Combining the best parameters with the two-dimensional Otosu algorithm, it has been used to identify the wrong image edges of a book. The test results are good and the noise pair detection is effectively reduced. This paper designs a robot trajectory tracking control algorithm on a small space curve. The standard PID control based on single neural network is compared with PID control. The results show that PID individual adaptive control algorithm not only responds quickly. Although it is also defective, it is relatively well controlled. The approach proposed in this paper, which does not change the traditional habit of readers searching and ordering books with catalog numbers, while allowing for the rapid identification and classification of erroneous books, provides viable solutions for improving management and library services and reducing the workload of staff.

References

Rajab

K.D.

, The Effectiveness and Potential of E-learning in War Zones: An Empirical Comparison of Face-to-Face and Online Education in Saudi Arabia, IEEE Access 21(1) (2018), 11–21.

Baldwin

and Trespalacios

J.H.

, Evaluation Instruments and Good Practices in Online Education, Online Learning 21(2) (2017), 52–58.

Krause

, Hall

, Williams

J.J.

, et al., Connecting Online Work and Online Education at Scale, Social Science Electronic Publishing 43(4) (2017), 3536–3541.

Wolf

R.C.

, Romer

A.K.

and Nord

, New Science, New Media: An Assessment of the Online Education and Public Outreach Initiatives of The Dark Energy Survey 77(5) (2018), 12–19.

Funes

and Mackness

, When inclusion excludes: a counter narrative of open online education, Learning Media and Technology 55(10) (2018), 1–20.

Welch

, Virtual Mentoring Program within an Online Doctoral Nursing Education Program: A Phenomenological Study, International Journal of Nursing Education Scholarship 14(1) (2017), 71–88.

Bialous

S.A.

, Sarna

, Wells

M.J.

, et al., Impact of Online Education on Nurses’ Delivery of Smoking Cessation Interventions With Implications for Evidence-Based Practice, Worldviews on Evidence-Based Nursing 42(10) (2017), 78–84.

Cook

P.J.

, Online Education and the Emotional Experience of the Teacher, New Directions for Teaching and Learning 201(153) (2018), 67–75.

Lasheng

Y.U.

, Xu

W.U.

and Yu

, An Online Education Data Classification Model Based on Tr_MAdaBoost Algorithm, Chinese Journal of Electronics 28(01) (2019), 25–32.

10.

Ovetz

, Click to save and return to course: online education, adjunctification, and the disciplining of academic labour, Work Organisation Labour & Globalisation 11(1) (2017), 48–70.

11.

Yang

Y.F.

and Tsai

C.C.

, Exploring in-service preschool teachers’ conceptions of and approaches to online education, Australasian Journal of Educational Technology 33(1) (2017), 134–147.

12.

Kolluru

and Varughese

J.T.

, Structured academic discussions through an online education-specific platform to improve Pharm.D. students learning outcomes, Currents in Pharmacy Teaching and Learning 9(2) (2017), 230–236.

13.

Amant

, Kirk. Of Friction Points and Infrastructures: Rethinking the Dynamics of Offering Online Education in Technical Communication in Global Contexts, Technical Communication Quarterly 43(2) (2017), 1–19.

14.

Okoh

, Makame

M.H.

and Awad

A.I.

, Toward online education for fingerprint recognition: A proof-of-concept web platform, Information Security Journal: A Global Perspective 55(8) (2017), 1–12.

15.

Steele

J.R.

, Jones

A.K.

, Clarke

R.K.

, et al., Use of an Online Education Platform to Enhance Patients’ Knowledge About Radiation in Diagnostic Imaging, Journal of the American College of Radiology 14(3) (2017), 386–392.

16.

Tse

M.M.Y.

, Tang

, Budnick

, et al., Pain and Pain Management Among University Students: Online Survey and Web-Based Education, Cyberpsychology Behavior and Social Networking 20(5) (2017), 305–313.

17.

Ruolan

, The Preliminary Exploration of the Application of MOOC in College Oral English Teaching, 2016 8th International Conference on Information Technology in Medicine and Education (ITME). IEEE 78(4) (2016), 32–43.

18.

Choi

, Empirical profiles of academic oral English proficiency from an international teaching assistant screening test, Language Testing 34(1) (2017), 247–266.

19.

Suthakorn

, Lee

, Zhou

, et al., A enhanced robotic library system for an off-site shelving facility, Springer Tracts in Advanced Robotics 24 (2006), 437–446.

20.

Kuilin

, Bo

and Lijun

, Research and Practice of the Wisdom Library Based on UHF RFID, Modern Intelligence 36(8) (2016), 88–92.

21.

Burton

and Kattau

, Out of Sight but not Lost to View: Macquarie University Library’ s Storage Print Collection, Australian Academic & Research Libraries 44(2) (2013), 102–113.

22.

Behan

and Keeffe

D.T. O

, The development of an autonomous service robot. Implementation: Lucas The library assistant robot, Intelligent Service Robotics 1(1) (2008), 73–89.

23.

Rai

and Dutta

, Image Edge Detection using Modified Ant Colony Optimization Algorithm based on Weighted Heuristics, International Journal of Computer Applications 68(15) (2014), 5–9.

24.

Dorigo

, Bonabeau

and Theraulaz

, Ant algorithms and stigmergy, Future Generation Computer Systems 16(8) (2000), 851–871.

25.

Rada-Vilela

, Chica

, Cordón

Ó.

, et al., A comparative study of Multi-Objective Ant Colony Optimization algorithms for the Time and Space Assembly Line Balancing Problem, Applied Soft Computing Journal 13(11) (2013), 4370–4382.

26.

Wang

, Long

, Lu

, et al., The segmentation of wear particles in ferrograph images based on an improved ant colony algorithm, Wear 311(1-2) (2014), 123–129.

	Ant colony algorithm edge extraction time (s)
Number of experiments	One	II	III	IV	V	Average time
Traditional Ant Colony Algorithm	1369.0	1410.6	1525.4	1439.6	1460.3	1441.0
Ant colony algorithm	424.13	425.16	430.54	423.32	425.87	425.80