KR101965294B1 - Method, apparatus and system for detecting light car - Google Patents

Abstract

Disclosed are a method for determining a compact car, an apparatus thereof, and a system thereof. The apparatus for determining a compact car uses a deep learning algorithm to determine whether a photographed car is a compact car. The apparatus performs preprocessing for an input image of the photographed car to generate a preprocessed image in order to increase accuracy of the determination. By precisely determining whether a car is a compact car using the preprocessed image, a service for the compact car such as a compact car discount or the like can be automatically performed even if car driver information or car information are not provided from the outside.

Description

[0001] METHOD, APPARATUS AND SYSTEM FOR DETECTING LIGHT CAR,

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The following embodiments relate to a method, apparatus and system for image recognition using a deep-running algorithm, and more particularly to a method, apparatus and system for determining lightness.

Various benefits such as rate discounts are applied to vehicles that use parking lots. However, in order to discount such a rate, discount processing is performed using most people's eyes or a separate coupon or discount coupon.

For automatic discount processing by the system, it is necessary for the local government or the government to share the vehicle information, but it is impossible to share the vehicle information for the purpose of protecting the personal information.

In the reality that the burden on parking lot management and administration expenses is increasing due to the continuous increase of labor costs, it is urgently required to develop a technology that can automatically perform discount processing by the system.

Of the vehicles using the parking lot, the share of light vehicles is around 10% ~ 20%. If a discount is automatically applied to a light car, the cost of labor will be reduced and it will be a great help in parking lot operation.

Korean Patent Laid-Open No. 10-2008-0022719, published on March 12, 2008 (name: light vehicle discrimination device and method thereof)

One embodiment provides a method, apparatus, and system for determining whether a light vehicle is a light vehicle using a deep running algorithm.

One embodiment provides a method, apparatus, and system for providing a pre-processed image to a deep-running algorithm to improve the accuracy of determining whether a light or dark is present.

A communication unit, at one side, for receiving an input image from which a vehicle is photographed; And a processor for generating a preprocessed image by performing preprocessing on the input image and generating a result of a deep learning algorithm indicating whether the taken vehicle is a light vehicle by inputting a preprocessed image in a deep learning algorithm Device is provided.

The light-vehicle discriminating apparatus may further include a photographing unit photographing the vehicle and generating the input image.

The processing unit may adjust the size of the image by performing cropping on the input image.

The processing unit may perform histogram equalization on the input image to adjust the pixels of the image to have values within a predetermined range.

The processing unit may perform an average value subtraction on the input image to adjust an average of the values of the pixels of the input image.

The processor may input the preprocessed image to the deep learning algorithm to determine a light vehicle score and a non-light vehicle score, and calculate a probability that the vehicle is a light vehicle using the light vehicle score and the non-light vehicle score.

The processor may adjust the input values so that input values for a predefined number of layers of the deep learning algorithm have a constant distribution.

The input value may include values of pixels of the input image.

The processor can improve the accuracy of the determination as to whether the vehicle is a light vehicle by applying the preprocessing for the input image to a linear model where there is no bias in the first layer of the depth learning algorithm.

On another side, receiving a captured input image of a vehicle; Performing preprocessing on the input image to generate a preprocessed image; And generating a result of a deep learning algorithm indicating whether the taken vehicle is a light vehicle by inputting a preprocessed image into a deep learning algorithm.

Wherein the generating the preprocessed image comprises: generating a cropped image by performing cropping on the input image; Performing a histogram equalization on the cropped image to generate a histogram equalized image; And performing an average value subtraction on the histogram equalized image to generate an average value adjusted image.

The preprocessed image may be the average value adjusted image.

Generating the result comprises: inputting the preprocessed image to the deep learning algorithm to determine a light vehicle score and a non-light vehicle score; And calculating the probability that the vehicle is a light vehicle using the light vehicle score and the non-light vehicle score.

If the probability is a value within a specified range, the depth learning algorithm may be performed by inputting the pre-processed image and the lightness information indicating whether the photographed image is a light vehicle or not.

The light vehicle information may indicate that the vehicle is a light vehicle when the service provided when the vehicle is a light vehicle and the light vehicle information may indicate that the vehicle is a light vehicle, The information may indicate that the vehicle is not a light vehicle.

The input values may be adjusted such that the input values for the predefined number of layers of the deep learning algorithm have a constant distribution.

The input value may include values of pixels of the input image.

The accuracy of the determination as to whether the vehicle is a light vehicle can be improved by applying the preprocessing for the input image to a linear model in which there is no bias in the first layer of the deep learning algorithm.

At least one of the size of the vehicle, the length of the vehicle, the height of the vehicle and the shape of the vehicle can be derived from the captured image of the vehicle.

At least one of the length of the vehicle, the height of the vehicle, and the shape of the vehicle can be used as an input value of the deep learning algorithm.

At least one of a mark representing the brand of the vehicle, a character string representing the manufacturer of the vehicle, and a character string representing the vehicle type of the vehicle may be detected in the captured image of the vehicle.

At least one of a mark representing the brand of the vehicle, a character string representing the manufacturer of the vehicle, and a character string representing the vehicle type of the vehicle may be used as an input value of the deep learning algorithm.

On another aspect, there is provided a computer-readable recording medium containing a program for performing a light-discern discrimination method.

A method, apparatus and system for determining whether a light vehicle is a light vehicle using a deep running algorithm are provided.

A method, apparatus and system are provided for providing a preprocessed image to a deep-running algorithm to improve the accuracy of determining whether a light vehicle is a light vehicle.

FIG. 1 shows a structure of a light vehicle discriminating apparatus according to an embodiment.
2 is a flowchart of a method of discriminating a light vehicle according to an embodiment.
Figure 3 shows a zero centered correction according to an example.
FIG. 4 illustrates normalization of an input value according to an absence of a bias value according to an example.
5 shows a pre-processing for an input image according to an example.
6 shows the effect of the pretreatment according to an example.
Fig. 7 shows the result of the discrimination of the minor car according to an example.

The following detailed description of exemplary embodiments refers to the accompanying drawings, which illustrate, by way of illustration, specific embodiments. These embodiments are described in sufficient detail to enable those skilled in the art to practice the embodiments. It should be understood that the various embodiments are different, but need not be mutually exclusive. For example, certain features, structures, and characteristics described herein may be implemented in other embodiments without departing from the spirit and scope of the invention in connection with an embodiment. It is also to be understood that the location or arrangement of the individual components within each disclosed embodiment may be varied without departing from the spirit and scope of the embodiments. The following detailed description is, therefore, not to be taken in a limiting sense, and the scope of the exemplary embodiments is to be limited only by the appended claims, along with the full scope of equivalents to which such claims are entitled, if properly explained.

In the drawings, like reference numerals refer to the same or similar functions throughout the several views. The shape and size of the elements in the figures may be exaggerated for clarity.

The terms used in the examples are intended to illustrate the embodiments and are not intended to limit the invention. In the examples, the singular includes the plural unless otherwise stated in the specification. It is noted that the terms "comprises" and / or "comprising" used in the specification are intended to be inclusive in a manner similar to the components, steps, operations, and / And that additional configurations may be encompassed within the scope of the embodiments of the exemplary embodiments or the technical ideas of the exemplary embodiments. When it is mentioned that a component is "connected" or "connected" to another component, the two components may be directly connected or connected to each other, It is to be understood that other components may be present in the middle of the components.

The terms first and second, etc. may be used to describe various components, but the components should not be limited by the terms above. The above terms are used to distinguish one component from another. For example, without departing from the scope of the right, the first component may be referred to as a second component, and similarly, the second component may also be referred to as a first component.

In addition, the elements shown in the embodiments are shown independently to represent different characteristic functions, and do not mean that each element is composed of separate hardware or a single software constituent unit. That is, each component is listed as each component for convenience of explanation. For example, at least two of the components may be combined into a single component. Also, one component can be divided into a plurality of components. The integrated embodiments and the separate embodiments of each of these components are also included in the scope of the right without departing from the essence.

Also, some components are not essential components to perform essential functions, but may be optional components only to improve performance. Embodiments may be implemented only with components that are essential to implementing the essentials of the embodiments, and structures within which the optional components are excluded, such as, for example, components used only for performance enhancement, are also included in the scope of the right.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings in order to facilitate embodiments of the present invention by those skilled in the art. In the following description of the embodiments, detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present disclosure rather unclear.

When entering the vehicle, the license plates on the front and rear sides of the entering vehicle are photographed as images, and the license plate number of the photographed license plates is recognized. At this time, as the photographed images are input to the process for discriminating the light cars, The result of whether or not it can be determined. In order to accurately determine whether the vehicle is a light vehicle, a preprocessing operation is required on the photographed image.

FIG. 1 shows a structure of a light vehicle discriminating apparatus according to an embodiment.

The light discriminating apparatus 100 may include at least a part of the processing unit 110, the communication unit 120, and the storage unit 130 as components. The components may communicate with each other via one or more communication buses or signal lines.

The components shown with respect to the minor discrimination device 100 in Fig. 1 may be only one example. All of the illustrated components may not be essential to the light discriminating apparatus 100. [ The light discriminating apparatus 100 may have more or fewer components than those shown in Fig. Further, two or more components shown in Fig. 1 may be combined. In addition, the components may be configured or arranged differently than shown in FIG. Each component may be implemented in hardware, including one or more signal processing and / or application specific integrated circuits (ASICs), or in software, or in a combination of hardware and software.

The processing unit 110 can process a job required for the operation of the light-dark discriminating apparatus 100. [ The processing unit 110 may execute the code of the operation or step of the processing unit 110 described in the embodiments.

The processing unit 110 can generate and process signals, data, or information generated by the light discriminating apparatus 100, which are input to the light discriminating apparatus 100, outputted from the light discriminating apparatus 100, Compare, and judge related to data or information. In other words, in the embodiment, the generation and processing of data or information and the inspection, comparison and judgment related to data or information can be performed by the processing unit 110. [

For example, the processing unit 110 may be at least one processor.

The processor may be a hardware processor and may be a central processing unit (CPU). The processor may be plural. Alternatively, the processor may include a plurality of cores and may provide multi-tasking that simultaneously executes a plurality of processes and / or a plurality of threads. At least some of the steps of the embodiments may be performed in parallel for a plurality of objects through a plurality of processors, a plurality of cores, a plurality of processes, and / or a plurality of threads.

For example, the processing unit 110 may execute the code of the operation or step of the light-discriminating apparatus 100 described in the embodiments.

For example, the processing unit 110 may run a program. The processing unit 110 can execute a code (code) constituting a program. The program may include an operating system (OS), a system program, an application, and an app of the light vehicle discrimination apparatus 100.

In addition, the processing unit 110 may control other components of the light discriminating apparatus 100 for the function of the processing unit 110 described above.

The communication unit 120 may receive data or information used for the operation of the light discriminating apparatus 100 and may transmit data or information used for the operation of the light discriminating apparatus 100. [

The communication unit 120 may transmit data or information to another device in the network to which the light-vehicle discrimination device 100 is connected, and may receive data or information from another device. In other words, in the embodiment, the transmission or reception of data or information can be performed by the communication unit 120. [

For example, the communication unit 120 may be a networking chip, a networking interface, or a communication port.

The storage unit 130 may store data or information used for the operation of the light-scene discrimination apparatus 100. [ In the embodiment, the data or information of the light discrimination apparatus 100 may be stored in the storage unit 130. [

For example, the storage unit 130 may be a memory. The storage unit 130 may include a built-in storage medium such as a RAM and a flash memory, and may include a removable storage medium such as a memory card or the like.

The storage unit 130 may store at least one program. The processing unit 110 may execute at least one program. The processing unit 110 can read the code of at least one program from the storage unit 130 and execute the readout code.

Operations, functions, and characteristics of the processing unit 110, the communication unit 120, and the storage unit 130 of the light vehicle discrimination apparatus 100 will be described in detail below with reference to embodiments.

The light-scene discrimination apparatus 100 may further include a photographing unit 140. The photographing unit 140 can generate an image of the object by photographing the object.

2 is a flowchart of a method of discriminating a light vehicle according to an embodiment.

In step 210, the photographing unit 140 may photograph the vehicle to generate an input image from which the vehicle is photographed.

The photographing unit 140 may be a camera. The photographing unit 140 can photograph a vehicle entering the parking lot or a vehicle advancing from the parking lot.

In step 220, the communication unit 120 may receive an input image.

In step 230, the processing unit 110 may perform preprocessing on the input image to produce a preprocessed image.

Step 230 may include at least a portion of steps 231, 232, and 233.

As will be described below, the processing unit 210 may generate a cropped image, a histogram equalized image, and / or an averaged-value adjusted image in generating the preprocessed image.

In step 231, the processing unit 110 may crop the input image to adjust the size of the input image.

The processing unit 110 may perform cropping on an input image to generate a cropped image.

The cropped image may have a predetermined size.

For example, the dimension of the pixels of the cropped image may be (224, 224, 3). In the dimension, the first number is the width of the image, the second number is the height of the image, and the third number is the number of channels.

In step 232, the processing unit 110 may apply Histogram Equalization (HE) on the input image to adjust the pixels of the input image to have values within a predetermined range. The default range can be from 0 to 1.

The processing unit 110 applies histogram equalization to the input image to convert the pixels of the input image to have values within the first predetermined range and then to correct the pixels of the input image again to have values within the second predetermined range can do. The first defined range may be from 0 to 255. < RTI ID = 0.0 > The second predetermined range may be from 0 to 1. For example, the processing unit 110 can apply the histogram equalization to the input image to convert the pixels of the input image to have values of 0 to 255, and then correct the pixels of the input image to have values of 0 to 1 have.

Alternatively, the processing unit 110 may perform histogram equalization on the cropped image to generate a histogram equalized image. The pixels of the histogram equalized image may have values within a predetermined range.

In step 233, the processing unit 110 may perform an average value subtraction (MS) on the input image to adjust an average of the values of the pixels of the input image.

Alternatively, the processing unit 110 may perform an average value subtraction on the histogram equalized image to produce an average value adjusted image.

The values of the pixels of the average value adjusted image may be closer to the predetermined value than the values of the pixels of the histogram equalized image. The default value may be 0.5.

The preprocessed image may be an image to which at least one of 1) cropping, 2) histogram equalization and 3) mean value subtraction is applied to the input image. Alternatively, the preprocessed image may be the averaged-value adjusted image described above.

At step 240, the processing unit 110 may input the preprocessed image to a Deep Learning algorithm to produce a result of the deep-running algorithm. The result of the deep-running algorithm may indicate whether the photographed vehicle is a light vehicle.

The deep running algorithm may be the VGG Net deep running algorithm or the ResNet 34 deep running algorithm.

For example, the deep-running algorithm may be an algorithm for discriminating a light that determines whether the photographed vehicle is a light vehicle in an image input by a deep-running algorithm.

Step 240 may include the steps of steps 241 and 242. [

In step 241, the processing unit 110 may input the preprocessed image to the deep learning algorithm to determine the light and light non-light points.

In step 242, the processing unit 110 may calculate the probability that the vehicle is a light vehicle using the light vehicle score and the non-light vehicle score.

The probability that the vehicle is a light vehicle can be calculated according to Equation 1 below.

[Equation 1]

1 / (1 + e) ^{(? -?)}

beta may be a non-lightweight score.

α may be a minor score.

The above-described miniscule discrimination method can also be used for learning for a deep learning algorithm for discriminating a miniscule. For example, the above-described step 240 may be replaced with a step for learning. In step 240, the processing unit 110 may perform the learning of the deep learning algorithm by inputting the pre-processed image in the deep learning algorithm and the light intensity information indicating whether the photographed light train is a light vehicle.

In addition, such learning may be performed selectively. For example, if the probability that the vehicle is a light vehicle is a value within a specified range or a value outside a specified range, which is calculated in step 242, it is considered uncertain whether the light vehicle is a light vehicle, Learning of the deep learning algorithm can be performed by inputting the preprocessed image and the lightness information indicating whether the photographed image is a light vehicle or not.

In such a case, the lightness information may be inputted by an operator of the system such as a parking lot using the light discriminating apparatus 100 or the light discriminating apparatus 100, or may be determined by other information of the system using the light discriminating apparatus.

For example, the lightness information may be input by an operator who identifies an input image or a preprocessed image. For example, the existing learning of the deep learning algorithm can involve operator intervention if it is not clear whether the vehicle is a light vehicle.

For example, the lightness information may be determined based on whether or not the service provided when the vehicle is a light vehicle by the system is applied to the photographed vehicle. The service provided when the vehicle is a light vehicle may be a light fare discount for the fare. Here, the charge may be a parking charge or a toll.

For example, when a service provided when the vehicle is a miniature vehicle is applied to the photographed vehicle, the miniature car information may indicate that the car is a miniature car. If the service provided when the vehicle is a light vehicle is not applied to the photographed vehicle, the light vehicle information may indicate that the vehicle is not a light vehicle.

In the learning for discriminating the light and the light, other information than the preprocessed image may be input to the deep learning algorithm. Such other information can improve the accuracy of the minor difference discrimination.

1) The processing unit 110 can derive the size of the vehicle, the length of the vehicle, the height of the vehicle, and the shape of the vehicle by using the image processing technique in the captured image of the vehicle.

The processing unit 110 calculates at least one of the size of the derived vehicle, the length of the vehicle, the height of the vehicle and the shape of the vehicle as an input value for learning the deep learning algorithm and a deep learning algorithm for determining whether the vehicle is a light vehicle It can be used as an input value.

2) The processing unit 110 can detect, from the photographed image of the vehicle, a mark indicating the brand of the vehicle, a character string indicating the manufacturer of the vehicle, and a character string indicating the vehicle type of the vehicle.

The processing unit 110 receives at least one of a mark representing the brand of the vehicle, a character string representing the manufacturer of the vehicle, and a character string representing the vehicle type of the vehicle from the captured image of the detected vehicle, an input value for learning of the deep learning algorithm, And can be used as an input value of the deep learning algorithm when it is determined whether or not it is.

3) In addition, a number string or the like detected from other photographed images can be used as the learning value of the deep learning algorithm and the input value of the learned deep learning algorithm.

Figure 3 shows a zero centered correction according to an example.

FIG. 3 shows the distribution of the values of the input vector of the deep learning algorithm.

In order to determine whether a light vehicle is a light vehicle using the deep learning algorithm, it may be desirable that the values of the input vector are distributed such that the average of the input vector values of the deep learning algorithm is zero.

In analyzing an input image using a deep running algorithm, it may be impossible or difficult to correct so that the average of the values of the input vectors is zero, since the deviation of the first input values for the deep running algorithm is too large. Therefore, correction is required for the input values.

FIG. 4 illustrates normalization of an input value according to an absence of a bias value according to an example.

There may be no bias in the convolution layer of the deep-running algorithm. Thus, it may be important that the input values (i.e., the values of the pixels of the input image) are normalized.

However, the values of the pixels of the input image may vary in average depending on factors such as the time at which the input image is taken or the field. Thus, it may not be desirable to normalize the values of the pixels of the input image to a predetermined average.

5 shows a pre-processing for an input image according to an example.

Deep-running algorithms such as ResNet34 can have a predefined number of layers. For example, ResNet34 can have 34 layers.

The processing unit 110 may adjust the input values so that the input values for the predetermined number of layers of such a deep running algorithm have a uniform distribution. The input values may include values of pixels of the input image.

2, the input image may be cropped to a predetermined size, such as 224, 224, 3, and the pixels of the cropped image may be converted to have values of 0 to 255, It can be corrected to have values of 0 to 1 again. After this correction, the values of the pixels of the input image may be averaged at the values of the pixels of the input image such that the values of the pixels of the input image have a constant distribution.

Through the preprocessing including such histogram equalization, the average of the pixels of the image can be made constant and can become a state suitable for normalization as the average becomes constant.

That is to say, the processing unit 110 applies a preprocessing including histogram equalization to the input image to a linear model (i.e., convolution layer) where no bias exists in the first layer of the deep-running algorithm, The accuracy of the determination as to whether the vehicle is a light vehicle can be improved.

6 shows the effect of the pretreatment according to an example.

6 shows an input image on the left side and a preprocessed image on the right side.

As shown in FIG. 6, the dark image can be enhanced by the preprocessing. Also, as the feature of the image is maintained by the preprocessing, the average of the values of the pixels of the image becomes constant, which may be suitable for normalization.

Fig. 7 shows the result of the discrimination of the minor car according to an example.

In Fig. 7, "value" may indicate the probability that the vehicle is a light vehicle. The closer the value of "value" is to 1, the more likely the vehicle is a light vehicle.

The apparatus described above may be implemented as a hardware component, a software component, and / or a combination of hardware components and software components. For example, the apparatus and components described in the embodiments may be implemented within a computer system, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable array (FPA) A programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to execution of the software. For ease of understanding, the processing apparatus may be described as being used singly, but those skilled in the art will recognize that the processing apparatus may have a plurality of processing elements and / As shown in FIG. For example, the processing unit may comprise a plurality of processors or one processor and one controller. Other processing configurations are also possible, such as a parallel processor.

The software may include a computer program, code, instructions, or a combination of one or more of the foregoing, and may be configured to configure the processing device to operate as desired or to process it collectively or collectively Device can be commanded. The software and / or data may be in the form of any type of machine, component, physical device, virtual equipment, computer storage media, or device , Or may be permanently or temporarily embodied in a transmitted signal wave. The software may be distributed over a networked computer system and stored or executed in a distributed manner. The software and data may be stored on one or more computer readable recording media.

The method according to an embodiment may be implemented in the form of a program command that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The program instructions to be recorded on the medium may be those specially designed and configured for the embodiments or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Magneto-optical media, and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions include machine language code such as those produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments. For example, it is to be understood that the techniques described may be performed in a different order than the described methods, and / or that components of the described systems, structures, devices, circuits, Lt; / RTI > or equivalents, even if it is replaced or replaced.

100: Light vehicle discrimination device
110:
120:
130:
140:

Claims (18)

A communication unit for receiving an input image of a vehicle;
A preprocessing unit for performing preprocessing on the input image to generate a preprocessed image, inputting a preprocessed image to a deep-running algorithm, and generating a result of the deep-running algorithm indicating whether the photographed vehicle is a light vehicle
Lt; / RTI >
Wherein the processor inputs the preprocessed image to the deep learning algorithm to determine a light vehicle score and a non-light vehicle score, calculates a probability that the vehicle is a light vehicle using the light vehicle score and the non-light vehicle score,
If the probability is a value within a specified range, the processing unit performs learning of the deep learning algorithm by inputting the pre-processed image and the lightness information indicating whether the photographed image is a light vehicle in the deep learning algorithm,
When the service provided when the vehicle is a light vehicle is applied to the vehicle, the light-weight information indicates that the vehicle is a light vehicle, and when the service provided when the vehicle is a light vehicle is not applied to the vehicle, Wherein the vehicle is not a light vehicle. The method according to claim 1,
A photographing unit for photographing the vehicle and generating the input image,
Further comprising: The method according to claim 1,
Wherein the processing unit crops the input image and adjusts the size of the image. The method according to claim 1,
Wherein the processing unit performs histogram equalization on the input image to adjust the pixels of the image to have values within a predetermined range. The method according to claim 1,
Wherein the processor performs an average value subtraction on the input image to adjust an average of values of pixels of the input image. delete The method according to claim 1,
Wherein the processing unit adjusts the input values so that input values for a predefined number of layers of the deep learning algorithm have a constant distribution,
Wherein the input value includes values of pixels of the input image. The method according to claim 1,
Wherein the processor enhances the accuracy of the determination as to whether the vehicle is a light vehicle by applying the pre-processing on the input image to a linear model in which there is no bias in the first layer of the deep-running algorithm. Receiving a captured input image of a vehicle;
Performing preprocessing on the input image to generate a preprocessed image; And
Generating a result of a deep-running algorithm indicating whether the taken vehicle is a light vehicle by inputting a preprocessed image to a deep-running algorithm
Lt; / RTI >
Wherein generating the result comprises:
Inputting the preprocessed image to the deep learning algorithm to determine a light vehicle score and a non-light vehicle score; And
Calculating a probability that the vehicle is a light vehicle using the light vehicle score and the non-light vehicle score
Lt; / RTI >
If the probability is a value within a specified range, then the depth learning algorithm is performed by inputting the preprocessed image and lightness information indicating whether the photographed image is a light vehicle,
When the service provided when the vehicle is a light vehicle is applied to the vehicle, the light-weight information indicates that the vehicle is a light vehicle, and when the service provided when the vehicle is a light vehicle is not applied to the vehicle, Wherein the vehicle is not a light vehicle. 10. The method of claim 9,
Wherein the generating the preprocessed image comprises:
Performing cropping on the input image to generate a cropped image;
Performing a histogram equalization on the cropped image to generate a histogram equalized image; And
Performing an average value subtraction on the histogram equalized image to produce an average value adjusted image
Lt; / RTI >
Wherein the preprocessed image is the average value adjusted image. delete delete delete 10. The method of claim 9,
Wherein the input values are adjusted such that input values for a predefined number of layers of the deep learning algorithm have a constant distribution,
Wherein the input value includes values of pixels of the input image. 10. The method of claim 9,
Wherein the accuracy of the determination as to whether the vehicle is a light vehicle is improved by applying the preprocessing for the input image to a linear model in which no bias exists in the first layer of the deep learning algorithm. 10. The method of claim 9,
Wherein at least one of a size of the vehicle, a length of the vehicle, a height of the vehicle, and a shape of the vehicle is derived from the captured image of the vehicle, and at least one of the length of the vehicle, Is used as an input value of the depth learning algorithm. 10. The method of claim 9,
Wherein at least one of a mark representing the brand of the vehicle, a character string representing the manufacturer of the vehicle, and a character string representing the vehicle type of the vehicle is detected in the captured image of the vehicle, a mark indicating the brand of the vehicle, Wherein at least one of a character string representing a car type and a character string representing a car type of the vehicle is used as an input value of the deep learning algorithm. A computer-readable recording medium embodying a program for carrying out the method according to any one of claims 9, 10, 14, 15, 16 and 17.

Images