Keywords

1 Introduction

License plate detection plays a major role in the world because vehicles are increasing day-by-day and pollution is also increasing due to increase of vehicle number. In this paper, we introduce the pollution fine control; in this, we impose fine on the vehicle owner whose pollution check date is expired. In commercial cities, the commercial places or near the buildings, parking areas are built, but the problem with those areas is that it does not have enough space. So to manage space of these parking lots, proper management and proper use of parking space are necessary, as this will improve the efficiency and reduces management cost. A single person in the parking areas is not enough to manage all kind of activities such as charging persons to place vehicle, how to control vehicle’s in and out space. This has made necessary to use the system of automatic plate recognition which in turn provide benefit like reduces management cost required to control and manage parking areas. Also to reduce road accidents and stolen vehicles cases, the number plate recognition system is highly effective which can not only help police (by reducing their man power), but also can track vehicles whenever needed.

The images of various vehicles have been acquired, and thereafter submitted to the MATLAB software where they are first resize the image and then convert it into gray scale images. Then, we use median filter to remove the noise in the image. Brightness, contrast, and intensity adjustments are made to optimum values to enhance the number plate and its digits. The output is saved in a notepad. Objective of proposed work is to create the modules for traffic regulation using a system of number plate detection. The input of the module is a part of the captured image, and output is a number which is a editable form of license plate. Basically, the proposed system recognizes all kinds of number plates of different varieties.

The rest of the paper is organized as the follows: Sect. 2 presents the background and related research of the proposed work. Section 3 describes the architectural design of proposed work. Section 4 elaborates the methodologies of proposed system. Section 5 highlights the experimental set up and results, and finally, Sect. 6 concludes the proposed work.

2 Backgrounds and Related Work

Previously, different work had already been carried out in the area of license plate recognition. There are five divisions in this system, which are input image, preprocessing of image, localization of the license plate in vehicle image, character segmentation from the license plate, and character recognition from the segmented license plate. The work of first two divisions should be done in an accurate way so that license plate can be easily searched and localized. Thus detection will greatly affect the efficiency and the speed of the whole license plate recognition system.

In the earlier works, the edge detection techniques was found to be an important technique that works by changing brightness of the plates, but this technique could not be used for images. But they can be applied to complex images due to their sensitivity to the unwanted edges.

Seyed Hamidreza Mohades Kasaei et al. (2011) have presented a real-time and robust technique of license plate detection and recognition; it uses the morphology and template matching. The performance of corrected plate’s identification and localization was 98.2%, and correctly recognized characters were 92% [1].

Abdul Mutholib et al. (2013) have presented the optimization of ANPR algorithm on limited hardware of Android mobile phone. Proposed algorithm was based on Tesseract library. The template matching-based OCR will be compared to artificial neural network (ANN)-based OCR. Results on 30 images showed that the recognition rate of license plate was 97.46%, and the processing time in recognition of license plate was 1.13 [2].

Farajian et al. [3] studied different techniques and compared them. Author finds out that every technique has own limitations, and some method gives better results under some conditions. This paper will lead to select the best technique for detection just according to circumstance. After this survey, author concludes that there was no ideal technique which is suitable for all circumstances [3].

Anagnostopoulos et al. (2008) have applied a method to locate the license plates, which consist of three stages: initially, used the Sobel operator that will fetch the vertical edges of the image of vehicle. After that applied the integral image and HSV color space [4].

Pratik Madhukar Manwatkar et al. (2014) In this paper, author has reviewed and analyzed different methods to find out the text characters from the images. Author has discussed some application of text recognition system [5].

3 Architectural Design of Proposed Work

Figure 1 presents the architectural design of our proposed system of traffic regulation using number plate detection. In this work, we have performed data preprocessing as the first step of number plate detection.

Fig. 1
figure 1

Process flow diagram

Full size image

The overall description of the design specification as shown in Fig. 1 is explained below.

3.1 Input Image

The initial step of license plate extraction system is to select an input vehicle image. This image contains license plate of vehicle. After the selection of the image, preprocessing of the image is performed.

3.2 Preprocessing of Image

This is the second step of license plate recognition system. Preprocessing is very essential for the good quality performance of character segmentation.

Preprocessing steps are:

  1. 1.

    Resizing image

  2. 2.

    Rgb-to-gray conversion

  3. 3.

    Noise removal (median filter)

Preprocessing of an image means “preparation” of the sample/image to introduce it to an algorithm for specified task such as recognition of image, feature extraction.

3.3 Brightness and Intensity of Image

After preprocessing of the image, we adjust the brightness and intensity of the image. To scale the intensity in MATLAB, we use imadjust() function. Intensity of the image is the average of the intensities of all pixels in image. Brightness can be increased or decreased by simple addition or subtraction to the image matrix.

3.4 Fill Holes in the Binary Image

A general use of the flood-fill process is to fill “holes” in image. Holes fill in the binary image are performed because of representation of foreground objects in the image. In our system, we use imfill() function to fill the holes in the binary image.

3.5 Display Registration Number in the Text Box

After filling the holes in the image, we obtain a number from the image. This number is displayed in the text box. This is the registration number of the vehicle.

3.6 Local Database

We create a local database which contains the information of vehicle owner. When registration number is obtain, then our system compared this registration number to the database. When match is found in the database, then all the details of the vehicle owner are obtained.

3.7 Pollution Fine

This system mainly works on the pollution fine control. In this, if pollution check date is expired, then fine is imposed on the vehicle’s owner. This fine is impose by sending the mail to the owner of the vehicle, and when pollution check is not expired, then no mail is sent to the vehicle owner and no fine is imposed on the owner.

4 Methodologies

The main goal of this system is to focus on the localization and recognition of the numbers present in license plate of a vehicle. We locate the license plate part of the vehicle from every images first and then register them to a reference image by template matching in same precision. We use a simple templates matching to recognize the letters in the estimated trained images. Before the processing of two proposed phase for the solution, input images need to preprocess of accurate output. The following sections describe the process of these methods step-by-step. This paper, based on the image processing methods, gives the image pretreatment, for the position of license plate on the whole image of vehicle.

To make the vehicle license plate recognition, multiple small phases must implement such as image acquisition, preprocessing of image, localization of license plate, segmentation of alphanumeric code, recognition of license plate code.

4.1 Image Preprocessing

Image preprocessing considerably increase the consistency of an optical examination for the specific feature detection. In the license plate, we generally use resizing of image and then convert into gray scale image. To resize the image, we are using imresize(); in this, width of the image is 400 and height of the image is adjusted accordingly to maintain the aspect ratio. After that we use rgb2gray() to convert the resized image into gray scale image.

4.2 License Plate Localization

In license plate localization, first we must have to find the location of the number plate in whole vehicle image for identification of registration number. To execute the system from input image to the output characters as in editable form, localization process takes important role. For the localization, basically, we use median filter to filter out the noise in the image; in this, noise type is pepper and salt. Dilation is performed after the noise removal from the image; then, we locate the license plate in the input image.

4.3 Segmentation of Alphanumeric Code

Segmentation of alphanumeric code is a key step in license plate extraction system. Many difficulties are in this step, which are noise presence in the plate area, alphanumeric codes are not properly aligned. If the alphanumeric codes in license plate are in ideal condition, that is, codes are sufficiently separated, code segmentation may be accomplished directly from vertical and horizontal segmentation. In this, it detects the horizontal lines in the image with a pixel value zero. After that, the image is converted into gray scale image, and then, we simply use ‘for’ loop to detect the portion of the image that had connected objects with pixel value of zero.

4.4 Recognition of Alphanumeric Code

Recognition of alphanumeric code is a process for the automatic conversion of alphanumeric code into editable text. After segmentation, alphanumeric codes of the license plate system need to carry on recognition for all isolated alphanumeric code. The template matching technique is an appropriate technique for the recognition of particular alphanumeric code. This technique is used in binary images, properly built templates, and it also obtained fine results for gray scale images. In this character recognition, we adjust the brightness and intensity of the image. After that, we fill the holes present in the binary image, and in the end, we use template matching technique to recognize the characters from the image. Finally, the output is displayed in the text box that is registration number of the vehicle.

4.5 Pollution Fine Controller and Driver Information Fetcher

After the character recognition, we checked the owner name, city, address, registration number of vehicle, and contact number. After recognition of the vehicle registration number, we compare the registration number to the local database. If the match is found, then we get the detail of the vehicle owner. Then, we compare the registration number to pollution control table, then we get the expiry date of the pollution check of vehicle owner. If current date is in the valid range, then all are OK, otherwise the fine is to be calculated and the fine can be send on manual basis or automatically on e-mail, and the details will be send by fetching the details required from the local database record using the registration number. Structure of the database of the local database will look like (Fig. 2).

Fig. 2
figure 2

Structure of the database of the local database

Full size image

Structure of the database of the pollution control will look like (Fig. 3).

Fig. 3
figure 3

Structure of the database of the pollution control

Full size image

The main objective of this work is to present more proficient way of pollution control testing for vehicle by using alphanumeric code segmentation and recognition. In proposed system, we manually created the database of vehicle license plate image. The license plate image is converted into text image by using alphanumeric code recognition.

5 Experiment Results

The proposed system was evaluated by taking vehicle input image manually, and the result obtained is shown in the following steps

  • Input image

  • Resize the input image

  • Convert image into gray scale image

  • Apply median filter to remove the ‘pepper and salt’ noise

  • Scale brightness and intensity of the image

  • Fill holes in the binary image

  • Display the license plate number

  • Compare with database

  • Display the Vehicle owner detail

  • If pollution check is expired, then mail is send to vehicle owner (Figs. 4, 5, 6, 7, 8, 9 and 10).

    Fig. 4
    figure 4

    Input image

    Full size image
    Fig. 5
    figure 5

    Resize input image

    Full size image
    Fig. 6
    figure 6

    Gray scale conversion

    Full size image
    Fig. 7
    figure 7

    Scale brightness and intensity

    Full size image
    Fig. 8
    figure 8

    Fill holes in binary image

    Full size image
    Fig. 9
    figure 9

    Final image

    Full size image
    Fig. 10
    figure 10

    Registration number

    Full size image

The experiments have been performed to test the performance of proposed system and to measure the processing time and recognition rate. Totally, 22 images were taken under various time, conditions, and angle as shown in Tables 1 and 2.

Table 1 Results of license plate detection
Full size table
Table 2 Performance of proposed license plate detection
Full size table

After the character recognition, registration number of the license plate is recognized; then, we compare the registration number to the manually created local database of vehicle owner detail. If vehicle registration number is found in the local database, then detail of the vehicle owner is displayed as shown in Fig. 11.

Fig. 11
figure 11

Display vehicle owner detail

Full size image

The registration number will be matched with the local database record of pollution check control. If the previous registration number is found in with the pollution control database, we get the expiry date and the date of pollution check shown in Fig. 12.

Fig. 12
figure 12

Pollution check control database

Full size image

Case 1: If pollution check expiry date is valid, then an e-mail is sent to vehicle owner, which contains the information of pollution check expiry date as shown in Fig. 13.

Fig. 13
figure 13

E-mail sent to the vehicle owner regarding expiry date of pollution check

Full size image

Case 2: If pollution check date is not valid, then e-mail is sent to the vehicle owner regarding fine imposed,. as shown in Fig. 14.

Fig. 14
figure 14

E-mail sent to the vehicle owner regarding fine

Full size image

6 Conclusion

The main aim of this work is to improve security of registered number plate in India. The proposed approach finds pollution check control of vehicle and extracts the driver information using license plate detection method. In proposed work, if pollution check date is expired, then mail is sent to the vehicle owner regarding fine imposed, and if pollution check date is not expired, then mail is sent to the vehicle owner regarding pollution check expiry date. We can also restrict the vehicles which are 10-year old or more than 10-year old in our work. The experimental results show that this approach can achieve outstanding and robust detection performance in pollution fine controller and vehicle restriction area. For the experimentation, totally 22 images of license plate were tested. The overall system performance of the system is measured on the basis of accuracy and recognition rate; recognition rate is 92.57% and accuracy is 50%. The limitation of the proposed system is that it fails to recognize some characters, which are ‘1’, ‘2’, and ‘R’. Future work has also been discussed to recognize the double line license plate and multiline license plate. We will also increase the template matching database.