postal automation system for mail sorting

International Journal of Emerging Technology and Advanced Engineering

Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 5, Issue 3, March 2015)

118

Postal Automation System for Mail Sorting Manjunatha V.G

1

1Student, Dept of Instrumentation & Technology, Dayananda Sagar College of Engineering, Bengaluru, India

Abstract— Our proposed work is to design the mail sorting

automation system using Optical Character Recognition

technique. It can sort the mails using the pincode, without the

human intervention. It removes the hassle of two-step process

of barcode encoding and decoding, resulting in a clean,

effective postal system. Hence we propose to design a sorting

machine that will remove the above mentioned constraints

and thus sort the mails easily. The principle used for sorting is

the Optical Character Recognition using MATLAB software.

Camera, placed over the slide unit captures the image of the

address. The pin code is selected and compared with a set of

characters in the data base. On finding a positive match,

based on the pin code, the mail is segregated by the MATLAB

program involving OCR technique. The processed data is sent

to the microcontroller, which activates the actuating arm

(servo motor) to allow the letters to move to the respective

stack (zone) and thus sorting the mails automatically,

reducing the human effort and errors..

Keywords—Mail Sorting, MATLAB, Optical Character

Recognition, PIC microcontroller, Postal Automation.

I. INTRODUCTION

Indian post is the most widely distributed postal network

in the world. With its service for over 150 years old, it has

been the backbone of the nation‟s communication and has

played a crucial role in the country‟s socio-economic

development. It touches the lives of Indian citizens in many

ways: delivering mails, accepting deposits under Small

Savings Schemes, providing life insurance cover under

Postal Life Insurance (PLI) and Rural Postal Life Insurance

(RPLI) and providing retail services like bill collection,

sale of forms, etc.

India has the largest Postal Network in the world with

over 1, 55,015 Post Offices (as on 31.03.2009) of which 1,

39,144 (89.76%) are in the rural areas. At the time of

independence, there were 23,344 Post Offices, which were

primarily in urban areas. Thus, the network has registered a

seven-fold growth since Independence, with the focus of

this expansion primarily in rural areas. On an average, a

Post Office serves an area of 21.21 Sq. Km and a

population of 7175 people.

Figure 1:- Current sorting method

The manual labor used in this wide and intense network

of services is very less, for example in Chennai where they

receive lakhs of mails every single day. To sort these mails,

on an average Chennai has only about 30 people to do the

job. 30 people to sort numerous mails is a very tedious job

to do. The figure 1 shows the current manual sorting

method. It consumes a lot of time and has an adverse effect

on the health of the people sorting the letters. With the

responsibility of delivering the letters on time, comes the

duty of working fast.

The sorting has to be done as fast as possible to be

received by the recipient on time. This method can

include some human errors like misreading the pincodes,

misplacing the letters in the wrong stacks, etc. It can also

affect the eyes due to constant vigilance on the pincode. To

meet the demands, a huge number of people should be

employed by the post offices. However, the numbers of

people for doing the job are very few. The Indian postal

system lacks in the human labor. Hundreds of human

beings are required for the task but in reality, only thirty

percent of that are available for the job. This increases the

labor, fatigue, errors, etc.

II. LITERATURE REVIEW

Mani, et.al [1] proposed many artificial neural network

models have been proposed to mimic the human brain in

solving problems involving human-like intelligence.



119

An application of an artificial neural network approach

for optical character recognition (OCR) is discussed in this

paper. We examine a simple pattern- recognition system

using an artificial neural network to simulate character

recognition.

Young-Mo Kim, et.al [2] proposed many practical

factors that need to be considered for a reliable character

recognition system in 3D space, the location of the visual

angle of a camera might play a crucial role. Different

viewpoints in 3D space produces distorted license plate

images in a camera. For this reason, a method is developed

to segment and to recognize characters of license plate

objects undergoing variant perspective view. A method for

segmenting license plate characters on a moving vehicle in

actual outdoor environment is based upon object contours.

And this method for recognizing is constructed from a

feature-based approach, parameterized by an affine

invariant parameters and the affine invariant features.

Experimental results show that the performance of the

proposed method is simple and robust, particularly when

objects are heavily distorted with strong perspective view.

Birch, et.al [3] proposed an algorithmic architecture for a

high-performance OCR system for hand-printed and

handwritten addresses is proposed. The architecture

integrates syntactic and contextual post processing with

character recognition to optimize postcode recognition

performance and verifies the postcode against simple

features extracted from the remainder of the address to

ensure a low error rate. An initial implementation of all

parts of the proposed system is reported, showing an

overall postcode recognition rate of 44% and correct

extraction of verification information for 24% of upper-

case addresses and 27% of mixed-case addresses.

Kavallieratou, et.al [7] proposed this paper deals with

the discrimination between machine-printed and

handwritten text, a prerequisite for many OCR applications.

An easy-to-follow approach is proposed based on an

integrated system able to localize text areas and split them

in text-lines. A set of simple structural characteristics that

capture the differences between machine-printed and

handwritten text-lines is presented and preliminary

experiments on document images taken from databases of

different languages and characteristics show a remarkable

performance.

Hidayatullah, et.al [8] proposed automatic license plate

recognition which is useful in a lot of applications. There

are a lot of methods used for that purpose and one of which

is Optical Character Recognition (OCR).

OCR in this paper is used to recognize optically

processed printed character in license plate number which

is based on template matching. We found some factors that

affect the effectiveness of template matching based Optical

Character Recognition in recognizing the license plate

number. Those factors are font type, template resolution,

noise, tilting and the painting. We have done some

improvement for the Indonesian condition by using a more

relevant font type, construction a new template with higher

resolution, and removing more noise.

III. FLOW CHART AND FUNCTIONAL BLOCK DIAGRAM

Figure 2:- Flow Chart of Mail Sorting System

The Postal Automation System for Mail Sorting works

as the flow chart is shown in figure 2, Initially image of

letter is captured using camera and processed individually.

The captured image is sent to MATLAB for feature

extraction and pattern recognition. After recognition is

carried out, it is compared with the data base to identify the

alphabets and numbers. Here we are only using the

extracted six digit number i.e. pincode for further

implementation.



120

Once the pincode matches with the data base, the motors

are actuated the hardware unit using PIC microcontroller.

Figure 3:- Functional Block Diagram

The block diagram consists of the following

components:

A. Envelope

The inland letters of dimensions 6x4” are designed such

that the State name and pincode are written together on the

fifth line of the address. It is placed on the movable sliding

plate at a predefined position.

B. Camera

12 Mega pixel USB camera fixed to the Sliding

Structure takes an image of the envelope

C. Sliding Unit

After a delay of 5 seconds, movable plate lifts up and the

envelope falls onto the conveyor belt. This is run by a 12V

DC Motor of 30 rpm

D. Conveyor Belt

Run by a 12V DC Motor of 60 rpm, through a pulley

system, the belt transports the envelope to the collection

unit. By this time the software code would have processed

the image and detected the zone it belongs to.

E. Collection Unit

It is operated by a high torque VS5 Servo Motor. Based

on the zone detected, it rotates to align the respective box

to the conveyor

F. PIC Microcontroller

PIC is a modified Harvard architecture microcontroller‟s

family made by Microchip Technology, derived from

the PIC1650 originally developed by General

Instrument's Microelectronics Division.

The name PIC initially referred to "Peripheral Interface

Controller". PICs are popular with both industrial

developers and hobbyists alike due to their low cost, wide

availability, large user base, extensive collection of

application notes, availability of low cost or free

development tools, and serial programming (re-

programming with flash memory) capability.

G. Computer

A computer is used for programming for the

computation of arithmetic and logical operations and run

many applications compatible with the application platform

and its processor. MATLAB is one such application which

runs on MICROSOFT WINDOWS environment. The

memory unit and the processor of the computer used here is

choose to meet the specifications and minimum

requirements of MATLAB such that the application runs

smoothly.

IV. SOFTWARE WORKING

The hardware works as explained later in the hardware

implementation. The main program for the algorithm is

written in MATLAB which analyze the pattern comparing

with the different pre-loaded images of the fonts in

greyscale JPEG format as shown in figure 4.

Figure 4:- Grayscale Images store in database

Using these images, the patterns of the scanned image or

the image captured from the camera is matched to get the

required characters present in the image. For an optimum

result the captured image in first oriented in such a way that

the characters are aligned on to zero degree line.

The detailed software working is explained below.

The input image is given to the MATLAB and is stored

in terms of matrix in it.

The image is then processed to align the characters to

zero degree line using the page alignment algorithm.



121

After the alignment process, each character is segmented

in the image by analyzing the intensity levels between

the characters and also by comparing with the images of

the characters loaded in the data base.

After recognition of all similar patterns and assigning the

characters to it, the extracted text is stored in a text file.

This can be also seen in the Command Window of

MATLAB.

Each Pincode is assigned with a value based on the sub

divisional zone it belongs. The sub divisional zones are

Bangalore North, Bangalore south, Bangalore East,

Bangalore West, and Bangalore Rural. If the Pin code

doesn‟t match with the database of the mentioned

subdivisions, it is considered that the pincode belongs to

other region.

The target device is programmed to read the transmitted

data and control the motors attached to it to sort the

mails.

Here the target device used is PIC 18f452

microcontroller and the IDE tool used to program is

MPLAB.

The microcontroller controls the rotation of the

servomotor connected to the sorter by giving a required

pulse width modulated (PWM) signal so that the servo

rotates to the particular angle and the conveyer belt

drops the mail to the box located at that particular angle.

Here each box is used for each zone in Bangalore. When

the data is received from MATLAB, the controllers

generates a pulse signal to rotate the motor present

below the rotating plate to align to the particular angle.

For each value received the sorter positions to the

assigned value.

The complete system has 2 more separate models, a

conveyer belt and a sliding unit. The sliding unit is

where the letters or mails are placed. Here the images of

mails are captured and they are mad to side down to the

conveyer belt. The conveyer belt drops each mail to the

sorter. During this movement on the belt, the servo

aligns.

V. HARDWARE DESIGN AND IMPLEMENTATION

The detailed description of hardware that has and its

working. The hardware includes the electronic components,

the electromechanical instruments and the mechanical

hardware model.

A. Electronic Components

The electronic components consists of the PIC 18F452

microcontroller, the development board which supports the

integration of microcontroller with other hardware

components, the motor driver circuit for interfacing dc

motors with the microcontroller. The description of each

electronic component is described below

1) PIC18F452 Microcontroller: The PIC 18F452

Microcontroller is a 40 pin, high performance RISC CPU.

The Microcontroller has 32 kilobytes of flash memory,

1536 bytes of on-board RAM and 256 bytes of data

EEPROM. The source code is backward compatible with

PIC16 and PIC 17 instruction sets. This microcontroller has

5 I/O ports and 4 Timers. This microcontroller has one

serial communication port and one parallel slave port. Its

instruction set consists of 75 pre-defined instructions. It

also has a 10-bit A/D module with 8 input channels. The

PIC Processor has a Harvard architecture, i.e. separate

instruction memory and data memory. The figure 5, shows

the pin diagram of PIC 18F452.

Figure 5:- PIC 18F452 Pin Diagram

2) Motor Driver Circuit: The motor driver circuit is a

board consisting of a L293D IC used to provide the

necessary current required for a DC motor connected to

it. This circuit is used as the output current of a

microcontroller board is not sufficient to drive a motor.

The figure 6 shows the circuit connections of a motor

drive board.



122

Figure 6:- L293D Board and its Connections

B. Electromechanical Instruments

The electromechanical instruments are the ones used for

the purpose of obtaining mechanical motions and

actuations by providing electrical impulses. Here in our

prototype, we use DC motor and the Servo motor

controlled by the Microcontroller. The detailed descriptions

of these elec1tromechanical instruments are explained

below.

1) Metal Geared Servo Motor: Servomotor has three

inputs Vcc (red wire), GROUND (brown wire- GND) and

PWM control coming from the Microcontroller (orange

wire). We connect the 7V DC power supply to the Vcc,

ground pin of the servomotor to ground of the controller

board and the remaining PWM input pin is connected to the

Port D.4 of the microcontroller. The figure 7, represents the

servo motor. The system runs inside a timed loop because

we need to keep generating a pulse at least every 20ms to

keep the servomotor in position. For a servomotor to move

to a particular position we need to give the appropriate duty

cycle and frequency

2) Dc motor: A DC motor is a two terminal

electromechanical device that converts electrical energy

into useful mechanical energy like a mechanical

movement. The figure 8 represents the High torque DC

Motor. This motor works on the basis of Faraday‟s Law.

Figure 7:- Servo Motor Diagram

Figure 8:- DC Motor

C. Mechanical Hardware Model

The hardware model consists of mainly three parts, they

are

1) Sliding Unit: The Sliding unit is a structure in which

the postal letters are placed in the horizontal sliding tray.

The Figure 9(a) and Figure 9(b) shows the basic hardware

structure of the sliding unit. A horizontal sliding plate is

attached to this structure so that it can freely move with one

end soldered to the model using hinges. Sliding unit is a

structure constructed using Aluminum sheet of thickness

1mm and aluminum bars and beams to hold the structure.

Sliding unit comprises of two plates. A movable Base

plate made of metal sheet of length 11.5” and width 5.5”.

Fixed Slide metal plate is cut of length 13.5” and width

5.5”.



123

A guide of 1” is made on either side of the sheet by

folding it twice by half, i.e. 0.5”. This serves the purpose of

guiding the letters and preventing them from slipping off

the slide unit. A hinge is welded between the two plates. A

base plate made of reinforced iron of length 19” and width

9” is used as the primary base of the structure. Rectangular

aluminum rods are cut to lengths of 11.5”, 18”, 18” and 18”

and attached to the base at distances of 3”, 8.9”, 12” and

18” respectively. With respect to these offset rods are cut.

Aluminum rectangular rod of 13” is cut and bent at 1.5”

and 5” using a HBM. It is then fixed to the base plate and

the fixed slide plate using 6mm bolts. Aluminum

rectangular rod of 19” is cut and bent at 2.75” and 6.8”

using a HBM. It is then attached to fixed slide plate and

base plate using 6mm bolts.

Two Aluminum rectangular rod of 25” is cut and bent at

7” and 10” using a HBM. It is then attached to fixed slide

plate and base plate using 6mm bolts. Aluminum L-Bend

are then attached across the above rods, thus forming a base

for the movable slide base plate to rest and move freely up

to an angle of 45 degree upwards. Two horizontal rods are

attached across the structure to provide it more stability

using 6mm bolts and drilling machine. The 12V 30 rpm DC

motor is then attached to a sheet metal plate using 1mm

screws. This plate is then bent across the edges to form a U

bend plate which is then attached to the rods supporting the

movable plates. A T bend is attached to the underside of

the movable base plate. A lever system is then made by

using 3 slotted rods linked to each other at the ends by

using 4mm screws fixed to the T bend and the motor shaft.

The assembly is shown in figure 9 (c).

Figure 9(a):- 3-D Model structure of the Sliding Unit

Figure 9(b):- Basic Hardware Structure of Sliding Unit

Figure 9 (c):- Front View and Side View of Lever Structure

A cylindrical rod is used across the two slotted angles

and fixed using stoppers to maintain the angle and

alignment of the lever system. A normally closed switch is

fixed on the underside of the movable slide plate close to

the hinge to restrict the movement of the plate beyond 45

degrees, failing which the motor might be pushed the

restrictions and can fail.



124

The motor wire is connected to the „C‟ (common)

terminal of the switch. The other wire from the motor and

the „NC‟ (normally closed) terminal of the switch is given

to the two ends of 12v adapter. 10 mega pixel camera is

then mounted above the sliding plate from where it can

take images of the envelope address.

Sliding unit is used to push the mails placed on it to the

conveyor belt. A digital camera is placed on the sliding unit

for image capturing. The captured image is processed the

DC motor in this unit is actuated using the microcontroller.

Now the lever structure push the horizontal plate upwards

and the postal letter slides to the conveyor belt unit, this

mechanism is shown in figure 9(d).

Figure 9(d): Sliding Unit Mechanism

2) Conveyor Belt Unit: Conveyor belt unit is used for the

task of delivering the letter into the collection unit. This

runs by a 12V DC Motor of 60 rpm, through a pulley

system, the belt transports the envelope to the collection

unit. The figure 10 shows the conveyor belt used in our

prototype.

This is constructed by aluminum sheet cut into L-Bends

of length 26” (2pieces) and 7” (2 pieces) and V-shape on

shorter pieces at distances 1” and 5” and make U- Bends.

Attach all the pieces to make a base. Fix the wheels onto

the cylindrical rods at distance of 3.5” and put the

conveyor belt around them. Now place the circular

cardboards at all 4 corners to avoid slipping of the

conveyor belt. Sheet metal guides of length 27” on both

sides are placed to help guide the letters along. Attach the

belt system to the base using the 6.5” rods at distances with

respect to the length of the belt system. Use the slotted

angles to provide stability to the structure. Use sheet metal

of length 6” to fix the DC Motor between the 6.5” rod and

the 4” rod. Attach one gear to the cylindrical rod and one to

the shaft of the motor. Make a dip at the middle of the

gears circumference using a file. Now using a rubber belt

make a pulley system.

Figure 10:- Conveyor Belt Arrangement

3) Collection Unit: The Collection unit is a sorter which

collects the letters from the conveyor belt. It is operated by

a high torque VS5 Servo Motor. This consists of 4 boxes

indicating North, South, and East, West based on the

subdivision of the district and others for other regions

outside the district. This unit rotates on a servo motor

depending on the zone the envelope is recognized to align

the respective box to the conveyor belt unit. The figure 11

shows the side view and the top view of the collection unit.

The collection unit is built by cutting a slot in the metal

plate such that it can accommodate the servomotor and

provide the whole structure stability and grip. The spikes of

the servomotor are then fixed on to the chassis which

provides the primary base to mount the thermocol. After

the thermocol is mounted on the chassis, the four zonal

boxes are then placed at angles 45, 90, 135, 180 degrees for

west, east, south and north zones respectively.

Figure 11:- Collection Unit

D. Functionality of the Hardware

The figure 12 shows the hardware model setup which

has mainly three separate sub designs namely identification

unit, the conveyor belt and the collection unit.

Letters are placed on the movable slide plate. Image of

the address is taken and MATLAB software determines

which zone it belongs to using Optical Character

Recognition technique and matching with the data form the

database of pincodes divides based on zones.



125

Now the counter output from the microcontroller to the

DC Motor (12V 30rpm) based on the serial communication

from the software results in the movement of the movable

sliding plate in upwards direction up to 45 degrees and

returns back to its initial position after a small delay of

time. After the sliding plate reaches 45 degree, it can no

longer move. Hence the push switch is used to breaks the

circuit and avoids damage to the motor.

Figure 12: Complete Hardware Module

The letter slips onto the conveyor belt. Here another DC

motor is used for its continuous rotation. The belt transfers

the letter to the collection unit. Depending on the digital

output sent to the servo motor from the software through

microcontroller, it rotates by the corresponding duty cycle

and falls into the particular zone box.

E. Hardware Circuit diagram

The circuit is constructed as show in the figure 13. The

input signal is given from the computer to PIC

Microcontroller through UART communication. Based on

the input, the motors are actuated to control the hardware.

Here we are using L293D modules for controlling speed

and direction of DC motors.

Figure 13: Circuit Diagram of PIC controller with DC and Servo

Motor

VI. RESULTS AND CONCLUSION

A. Results

The Postal Automation System was tested with different

envelops for different localities. The system was tested for

efficient running for different pincode and also envelops

placed at different orientations. The output is obtained as

shown on Command Window of MATLAB as shown in

figure 14.

Figure 14:- Output in MATLAB



126

The approximate time consumed is about 4 seconds by

MATLAB and the complete time consumed is less than 7

seconds so that the mails can be sorted on an average of

500 to 800 mails per hour depending on the feeder. The

model works on the principle of Optical Character

Recognition hence the requirement of barcodes or RFID

tags are not needed. Depending on the zip code detected by

MATLAB, the servo motor gets actuated by PIC 18F452 to

rotate the collection tray to the required collection box.

Figure 14(b):- Mail Sorting in Collection Unit

We have also found out the exact pulse width of the

servo motor for its positioning to a particular value. The

values for the pulse width and its corresponding angle is

mentioned in Table I.

TABLE I

Positive Pulse Width For Each Degree Of Motion Of Servo Motor

Angle (in

degrees)

Pulse width

(in ms)

Angle ( in

degrees)

Pulse width

(in ms)

0 0.44 100 1.44

10 0.55 110 1.545

20 0.64 120 1.65

30 0.74 130 1.76

40 0.84 140 1.86

50 0.93 150 1.97

60 1.02 160 2.08

70 1.12 170 2.18

80 1.23 180 2.28

90 1.33

B. Conclusion

Application of modern control mechanism on mail

sorting system may be an efficient approach of reducing the

delays and mismanagements, experienced in our traditional

postal system. Since the existing postal system is time

consuming and prone to human errors, it is high time that

the Indian Postal Department is upgraded to new

technology as soon as possible. Adoption of the evolving

technology will find India a place in the list of fast

developing countries, with a highly efficient and fast postal

system in India.

Hence Postal Automation System for Mail Sorting is the

first step towards such technological advancement. „Mail

Sorter using LABVIEW‟ segregates printed letters on zonal

division basis by using the highly advanced technology of

OCR and software MATLAB powered by MATHWORKS.

It is gaining popularity because of its flexibility in

programming approach and hardware simulation. This

method is highly efficient in sorting printed mails.

Hardware that comprises this project is extremely sturdy,

occupies very little space and is one time investment that

provides invaluable future returns. Thus we have devised

an efficient and smart mail sorting machine using OCR and

MATLAB technology along with a well-equipped and

modernized hardware structure consisting of Sliding Unit,

Conveyor belt and Collection Unit. Integration of above

software and hardware provides us with the “Postal

Automation System for Mail Sorting”.

The major importance and advantage of this system is

that it is small and compact and space occupied by the

setup is very less of few meters square feet, unlike the

existing automatic sorting machines. This also avoids

human intervention since the sorting is done by the

program, thus preventing damages to the human eye. It also

errors like misreading the pincode, misplacing the letters

are impossible. The major constraint is time which is

overcome using this as machine oriented tasks are always

fast, faster than the man can think of. Unlike the bar code

method, there is no 2-step procedure: bar code generation

and sorting according to the generated bar code are the 2

steps in bar code process, which is absent in this operation.

Any color envelope can be used: unlike the bar code

method, envelope color requirement is not an issue.

VII. FUTURE SCOPE

Automatic mail sorting operation opens a wide scope for

future development. The technique can be extended to read

handwritten mails using neural networks and fuzzy logic.



127

Owing to the fact that every individual in the world has

unique handwriting, setting the character training for every

handwriting is beyond our imagination. Neural network

and fuzzy logic works on artificial intelligence. The

algorithm behind this will identify any type of handwriting.

So this method can be extended from printed official letters

to any handwritten unofficial letters. If trained further,

fuzzy logic can help identify and read different languages

so that language will not be a barrier in receiving the

messages from the loved ones.

One can further automate, variable masks by installing

and programming variable ROI by scanning. Further

sorting can be carried out up local district and main road

level by updating and widening the data base.

REFERENCES

[1] Mani, Nallasamy, Srinivasan, B., "Application of artificial neural

network model for optical character recognition," Systems, Man, and Cybernetics, 1997. Computational Cybernetics and Simulation.,

1997 IEEE International Conference on , vol.3, no., pp.2517,2520

vol.3, 12-15 Oct 1997, doi: 10.1109/ICSMC.1997.635312.

[2] Mi-Ae Ko, Young-Mo Kim, "A simple OCR method from strong

perspective view," Information Theory, 2004. ISIT 2004. Proceedings. International Symposium on , vol., no., pp.235,240, 13-

15 Oct. 2004, doi: 10.1109/AIPR.2004.8.

[3] Kabir, E., Downton, A.C., Birch, R., "Recognition and verification

of postcodes in handwritten and hand-printed addresses," Pattern Recognition, 1990. Proceedings. 10th International Conference on ,

vol.i, no., pp.469,473 vol.1, 16-21 Jun 1990, doi:

10.1109/ICPR.1990.118148

[4] Ahmadi, A., Ritonga, M.A., Abedin, M.A., Mattausch, H.J., Koide,

T., "A Learning OCR System Using Short/Long-term Memory

Approach and Hardware Implementation in FPGA," Evolutionary Computation, 2006. CEC 2006. IEEE Congress on , vol., no.,

pp.687,693, 0-0 0, doi: 10.1109/CEC.2006.1688378

[5] Cannon, M., Fugate, M., Hush, D.R., Scovel, C., "Selecting a

restoration technique to minimize OCR error," Neural Networks,

IEEE Transactions on , vol.14, no.3, pp.478,490, May 2003, doi: 10.1109/TNN.2003.811711

[6] Alon, Jonathan, Athitsos, V., Sclaroff, S., "Online and offline

character recognition using alignment to prototypes," Document Analysis and Recognition, 2005. Proceedings. Eighth International

Conference on , vol., no., pp.839,843 Vol. 2, 29 Aug.-1 Sept. 2005, doi: 10.1109/ICDAR.2005.177

[7] Kavallieratou, E., Stamatatos, S., Antonopoulou, H., "Machine-

printed from handwritten text discrimination," Frontiers in Handwriting Recognition, 2004. IWFHR-9 2004. Ninth International

Workshop on , vol., no., pp.312,316, 26-29 Oct. 2004, doi:

10.1109/IWFHR.2004.65

postal automation system for mail sorting

Documents