KR102180916B1 - Artificial Intelligence image analysis system and method thereof - Google Patents

Abstract

The present invention relates to a system and a method for AI image analysis. The system is installed at a parking lot entrance, a self-driving vehicle, or the like. According to the present invention, a change in ambient illuminance is analyzed by means of AI, the illuminance is adjusted for non-error occurrence, and then a vehicle entering a parking lot is imaged or a self-driving vehicle images the surrounding area. The AI image analysis system includes: a basic illuminance register unit storing basic illuminance for adjusting the shutter, gain, and aperture levels of an imaging device; an illuminance analysis unit selecting a daytime illuminance rectangle area or a nighttime illuminance rectangle area in accordance with the entering direction of the vehicle by analyzing an image for illuminance analysis in accordance with an AI automatic illuminance adjustment algorithm; a mode-specific zone range analysis unit providing a threshold illuminance value by selecting a illuminance level range added or subtracted within a mode-specific zone range based on the illuminance value of the start mode through the AI automatic illuminance adjustment algorithm with respect to the selected rectangle area; and a central processing unit (CPU) performing control such that the provided threshold illuminance value is applied to the imaging device and the imaging device acquires an image for object recognition by imaging the vehicle or an object based on the threshold illuminance value.

Description

Artificial Intelligence image analysis system and method thereof

The present invention relates to an artificial intelligence image analysis system and method, and in more detail, it is installed on a parking lot entrance and an autonomous vehicle, etc., before photographing a vehicle entering a parking lot, or before an autonomous vehicle photographs the surroundings, by artificial intelligence. By analyzing changes in ambient light intensity and automatically adjusting the light intensity so as not to cause an error, and then photographing an image, the vehicle number entering the parking lot is recognized, or an autonomous vehicle is taking an image while moving, and various data necessary from the image are recorded in real time. The present invention relates to an artificial intelligence video analysis system and method that allows the user to safely drive to the destination by analyzing it with an unmanned driving method.

In general, the license plate of a vehicle is a unique number assigned to each vehicle to be identified differently, and the manufacturer, production period, use purpose, displacement, prime mover type, owner, residence registered illegal vehicle, stolen vehicle inquiry, subtitle violation vehicle search, and wanted vehicle It is used to search information such as speeding violation vehicle, non-paid vehicle, toll collection, vehicle passage violation, and special vehicle classification.

Therefore, it is very important to recognize the vehicle number quickly and accurately in any speed and inquiry environment.

The vehicle number recognition technology is an image recognition technology that recognizes the vehicle number by taking an image from a vehicle entering, leaving, moving, or parking.

When applied to the parking control system, it analyzes visited vehicles and various regular vehicles through the non-ticketing entry and exit method, and automatically controls the entry and exit status, and in the case of visiting vehicles, the parking time is automatically set by a paid/unmanned fare calculator. Calculate.

When applied to the parking location control system, it performs functions such as video storage and monitoring with automatic announcement function and video security control suitable for the violation to prevent inside car search control, guidance control, and illegal parking violation. You can easily and automatically search for images stored in the NVR with only the vehicle number.

In addition, image recognition technology is widely used to crack down on various illegal parking violations and road traffic violations such as various parking violations, designated parking violations, special parking area violations, parking and stopping violations.

In addition, the image recognition technology makes it possible to know whether there is a helper through a pop-up of an existing image including images of a person in an emergency situation and a three-dimensional pop-up on the surrounding four screens when linked with emergency control.

When applied to various image analysis technologies of autonomous driving, the eyes and intelligence play a role in allowing the vehicle to safely autonomously drive by taking images of surrounding objects while driving at high speed and analyzing the speed, direction of travel, and types of objects. Do it.

However, in the image recognition technology, when backlight, reflected light, or metering occurs due to sunlight or light from surrounding light sources, an image that cannot detect the speed or driving direction of surrounding objects and cars may be captured. It disappears outside.

Therefore, in order to take an image that can analyze the type and speed direction of moving objects in any case before taking the image for analyzing various images, the illumination analysis to prevent the occurrence of errors in advance before taking the image analysis image For each moving object to be able to analyze the image of the moving object by first photographing the image for use, the high-intelligence illuminance adjustment that can be adjusted automatically for each moving object is then taken, and then the analysis image to be used for various image analysis is taken. It should be possible to analyze various video data.

Vehicle number recognition technology is one of the image recognition technologies, and in order to quickly and accurately recognize the number recognition image, even if any natural phenomenon error occurs, an image based on the threshold value for image analysis must be taken.

Even if an image with a threshold value is taken, this means that if the entire image range is taken as a threshold value, the vehicle number cannot be analyzed so that the license plate part of the vehicle cannot be recognized, such as reflected light, backlight, metering, some shadows, bad car headlights, It is recorded as an error-prone image due to fog, snow, or rain.

Therefore, only the surrounding illuminance can be photographed with a partial area threshold in which number analysis is possible only when the threshold illuminance is changed based on the illuminance around the license plate.

All existing cameras have evolved with the same function, a technology that shoots all parts of the captured image with a high contrast ratio.

A technology that shoots with a high contrast ratio can increase the sharpness of real-time image quality that occurs in real natural phenomena, but in terms of image analysis, in order to analyze the image, objects included in every corner of the screen within the entire shooting range are better than the contrast ratio of the image quality. One of the important factors is to take an image with a lower contrast ratio and higher definition for analysis.

This is to analyze the part where the license plate is located or the moving object and speed, the type of surrounding objects, etc. for video shooting in which only the contrast ratio is increased due to the angle, location, amount of illumination, and shadow that the sun or light source illuminates. There is a need to take an image for image analysis after automatically adjusting the illuminance through artificial intelligence so that it can be performed.

In addition, the conventional vehicle number recognition system installed outdoors generates a shadow by the object when there is an object in the surrounding environment where the sun shines, and when a phenomenon in which a part of the vehicle license plate is covered by the shadow occurs, the vehicle license plate is As number recognition became impossible because it could not be found, there was a need for a technology that could automatically adjust the illuminance through artificial intelligence.

Conventional vehicle number recognition systems installed outdoors or image analysis systems installed in autonomous vehicles do not normally occur at all depending on the angle and time of light shining from sunlight or strong light sources around them, but Images with errors such as backlight, reflected light, and metering are being taken due to reflection angles and shadows from objects. As a result, since number recognition may become impossible, there is a need for a technology that automatically adjusts the illuminance so that an error does not occur prior to photographing an image for image analysis.

Conventional vehicle number recognition systems installed outdoors or image analysis systems installed in autonomous vehicles are used to determine the shape, size, and thickness of clouds, the angle of the sun, and the amount of light when large amounts of small clouds are continuously moved. Due to the complex action, there is a phenomenon in which the variation in illuminance changes rapidly and frequently.

At this time, the current threshold around the license plate is accurately analyzed and changed to the threshold illuminance based on the license plate, and in the meantime, if an image is taken outside the threshold illumination value range due to the occurrence of infinite natural phenomena, misrecognition or non-recognition occurs. In particular, in the case of the occurrence of false recognition, as the deviation of the threshold value increases, the misrecognition phenomenon increases. Therefore, there is a need for a technology capable of reducing false recognition errors.

The conventional vehicle number recognition system installed outdoors or the image analysis system installed in autonomous vehicle equipment, etc., when diffuse reflection occurs due to poor installation of car headlights, or the occurrence of standing water due to snow or rain, the vehicle is within the video recording range. Diffuse reflection occurred only at the moment of entry, and it was analyzed as a different illuminance value, which was the main cause of the occurrence of the phenomenon that deviated from the illuminance analysis range. As a result, there has been a phenomenon in which the vehicle number cannot be recognized at night, so the necessity of a technology that analyzes the range of day and night illumination analysis is required.

Conventional vehicle number recognition systems installed outdoors or image analysis systems installed in autonomous vehicles, when a phenomenon in which the front part of the camera image photographing is covered by objects moving close to the surroundings occurs, blinds and images are captured. A phenomenon similar to the shooting phenomenon occurs, and it is impossible to analyze various image data unless images of objects around the rear license plate or the overlapping camera are taken together.

The conventional vehicle number recognition system installed outdoors is the vehicle in the rear license plate of the vehicle in the event that the front license plate cannot be read, such as dirt is thrown on the front license plate of the vehicle due to snow or rain, or snow is accumulated on the running vehicle. Be able to recognize the number.

The conventional vehicle number recognition system installed outdoors does not recognize the vehicle number when the visible distance due to fog is within 3m or when the daytime illumination is dark, etc., especially in the case of electric vehicle license plates, only the license plate is dark. Because it is color, the illuminance is daytime, but the number plate is dark and the infrared light is not turned on, so it is often not possible to recognize the number when recognizing the vehicle number.

In the image analysis equipment for blocking various types of error occurrences as described above, the conventional vehicle number recognition system or the image analysis system installed in an autonomous vehicle, etc., do not know when and how the climate change of natural phenomena will change rapidly and from time to time. As it is changing, two methods are largely used to solve this.

The first is a method of recognizing only the vehicle number with only the black and white images taken after turning on only the infrared light without adjusting the illuminance after completely blocking the visible light itself because the illumination control method cannot block infinite generation.

Second, since artificial intelligence illuminance adjustment occurs indefinitely, it analyzes the roughly predicted illuminance around the license plate, and creates three virtual images made of illuminance values such as the maximum, minimum, and equivalent values expected to move relative to the previous illuminance, and sequentially the vehicle number. It is a virtual illuminance control method that recognizes.

The first conventional visible light blocking method cannot control illuminance, so that errors caused by backlight, reflected light, and metering caused by visible light can be blocked, or errors due to backlight, reflected light, and metering due to infrared rays cannot be blocked.

Therefore, since you have to turn on the infrared light at both the day and night to take an image, not only the black and white images are captured regardless of the shooting time, and if the infrared light is turned off, you can analyze the image regardless of the day or night. There was no.

In addition, since the image analysis system installed in autonomous vehicles can analyze the shape of objects only by taking color images, it cannot be used except for the vehicle number recognition system, and is used as an image recognition equipment that must turn on infrared rays. This was impossible.

In addition, the most energy-consuming part is turning on infrared lights, which inevitably has a low energy efficiency rating.

To solve this problem, the conventional illuminance analysis method around the license plate allows basic illuminance analysis, but illuminance analysis is not accurate and color image shooting analysis is impossible. Equipment installation location and vehicle entry method, one illumination analysis location, slightly dark daytime illumination During each special case, such as the driving speed of the vehicle when entering the license plate of an electric vehicle, or the diffuse reflection of the car headlights due to snow or rain, images that do not recognize the vehicle number were often photographed.

The conventional technology that completely blocks visible light and uses only infrared light is a technology that forcibly blocks visible light and analyzes the vehicle number because it is impossible to adjust the artificial intelligence automatic illuminance required for instantaneous high-speed analysis. Therefore, in a way that color image images cannot be photographed, only the last 4 digits are entered for quick search when applied to the toll control system or when searching for a vehicle.

At this time, even when the same vehicle is searched, or when misrecognized or unrecognized, it takes a lot of time to search only with video, and when shooting with visible light even for black and white images, the contrast ratio of black and white is different. It is easy to find, but with the conventional technology that is photographed only with infrared rays, not only the phenomenon that the contrast ratio is almost the same and the recognition ability is lowered, but also the technology of recognizing the vehicle number by photographing a color image was almost impossible.

Although it is possible to find the infrared illuminance analysis value within the entire range of the captured image by applying the conventional vehicle number recognition system with visible light blocking, it was almost impossible to reach the threshold value because the annual illuminance measurement value is measured infinitely.

If the automatic illuminance control technology is applied to the existing camera, since it is impossible to analyze the illuminance value of the position to be added or subtracted after recording the image in real time, the register value (shutter, gain, aperture value, etc.) Since the auto iris lens has to be moved one step at a time, it has become impossible to use it due to a delay when using the vehicle number recognition equipment.

Therefore, it is impossible to use the existing automatic aperture control technology compared to the moving speed of the vehicle, and even if there is an artificial intelligent aperture that can be moved, a number of cases occur, so only the license plate location technology and the illumination around the license plate can be analyzed. There has been a need for partial image analysis technology.

When recognizing the front license plate, when various kinds of natural phenomena that cause unrecognized snow or other vehicle numbers occur, the conventional technology has no choice but to take an image of the rear license plate in the same way as the front license plate recognition method, so the recognition rate was not significantly improved. .

It was almost impossible for the vehicle number recognition technology using the conventional schedule illuminance control method to analyze the illuminance according to the natural environment change in the case of infinite in real time. Even if it can be analyzed, it is impossible to analyze the illuminance where the license plate of the vehicle is located in the image being photographed.Thus, a mechanical method is used in the entire image being photographed to completely block visible light, and the number is recognized only with infrared light. I had to shoot a video.

In the case of sunlight, since infrared wavelength light is emitted, it has an adverse effect upon occurrence of reflected light, backlight, and metering, resulting in increased errors. Another way is to create an annual illuminance schedule and create an annual illuminance to control the illuminance.However, since the sun revolves and rotates, it is different every hour, every season, or depending on the site installation location. There is a limit to preventing recognition errors.

The method of analyzing images by capturing images with a camera also reached a limit in autonomous vehicles, so it was forced to use various other sensors at the same time.

Therefore, even with annual climate change, the problem can be solved only by taking a threshold image that allows image analysis. In addition, in the vehicle number recognition system, it is an important point to ensure that the illuminance value around the license plate reaches the threshold value even in the entire range image being photographed. Became.

When the double-sided license plate recognizes the vehicle's front license plate, if it is unrecognized, the rear license plate can be recognized, but there is no way to know when the front license plate misrecognition occurs.

Therefore, it is possible to analyze the number in the number recognition image taken after adjusting the illuminance by the threshold value, but since there is a high possibility of false recognition, the rear license plate is read out of the threshold value by reading the illuminance value from the front license plate recognition image. There is a need for a smart comparison algorithm that compares which image has a high probability of causing an erroneous recognition rate among the images taken and analyzed with front and back numbers.

In the conventional technology, when backlighting occurs in the entire image being photographed, the entire screen can be photographed to reach a threshold value, but even in such a case, a backlight error occurs in which the area around the license plate becomes darker, and an image in which vehicle number analysis is impossible is photographed. It is necessary to separate and analyze only the partial image around the license plate of the vehicle, and if the image of the moving vehicle is captured, the shooting position is photographed differently every time according to the driver's driving habits, so the technology to accurately find the active license plate location and the partial illumination value around the license plate There is a need for a smart automatic illuminance control analysis technology that makes the illuminance around the license plate reach a critical value.

In addition, when a dark image with a limitation in automatic illuminance control in which the illuminance value around the license plate can no longer be adjusted due to fog, snow, or rain during the daytime is captured, the artificial intelligence that enables the image for number recognition to be taken by turning on the infrared light is turned on. There is a need for a technology that enables illuminance analysis.

In order to shoot an image capable of number analysis, when the time taken to move to the threshold illuminance information (shutter, gain, experence) that has been smartly analyzed for the camera is 20ms or less, and it is changing to a different illuminance during this change time. If the actual captured image does not reach the desired threshold, or if a subtracted image is captured, number analysis becomes impossible, or a rapidly changing weather situation is prevented from being recorded. There is a need for a smart automatic analysis technology for environmental changes that can be analyzed.

In addition, there is a need for an artificial intelligence illuminance control technology that slows down the rapidly changing illuminance by turning on the infrared light when falling illuminance occurs, taking an image, and taking an image with a register value below the appropriate mode value when increasing or decreasing.

Since the conventional vehicle number recognition technology is not correct in the illuminance analysis technology, it is not clear to distinguish between day and night boundaries, and even if it is an annual schedule illuminance software that can classify the night with the schedule illuminance, the equipment is installed even if the sun is divided into sunset. It depends on the location, so no matter where it is installed, it could not be accurately identified.

In addition, even with the conventional technology that allows the illumination analysis around the license plate, if it is raining at night and it is on the floor, diffuse reflection is generated by the headlights of the car and the illumination around the license plate is analyzed as a higher illumination, and a darker image can be taken. In order to prevent this, there is a need for a technology capable of shooting an image by fixing it to the night mode and turning on the infrared light.

Even this, when water is accumulated on the floor due to snow and rain at night, an error of analyzing the illuminance above the night has occurred because the actual illuminance and photographing illuminance are different due to the diffuse reflection of the vehicle's lighting. When analyzing illuminance that must change rapidly to another mode outside the range of adjacent groups (night, dawn, morning, appropriate, daytime) due to sudden change in the natural environment, the illuminance measurement according to the sudden change in the natural environment is a factor that causes a recognition error. As it has been, there is a need for an adjustment mode technology that can buffer the mode by one step because a sequential stage illumination control mode is required to have a dampening effect in case of sudden change in mode.

In addition, conventionally, in order to automatically adjust the illumination of the camera, it is necessary to control the illuminance-related function using the shutter, gain, and expander, but when the current illuminance value is analyzed and the values such as shutter, gain, and aperture are too asymmetric, The captured image has reached the threshold value, but an asymmetric threshold value has been generated that lowers the quality and sharpness of the actually captured image.

In addition, in the conventional vehicle number recognition system installed outdoors, an error phenomenon occurs when recognizing the number by photographing a vehicle with a camera built into the equipment, most of which have occurred for the following reasons.

When a shadow is generated by an object in the surrounding environment in a place where sunlight is shining, this shadow is caused by a phenomenon that covers part of the license plate of the vehicle.

In addition, an error occurred as an image in which the number cannot be recognized is photographed due to backlighting, reflected light, and metering depending on the angle and time of the sun.

In addition, when small clouds were moving continuously on a large scale, the shape and thickness of the clouds, the position of the sun, and the angle of illumination interacted, resulting in a complex phenomenon.

In addition, when there is a case in which the variation in the illumination intensity changes rapidly, or when snow or rain falls on the floor, the light of the headlight of the car is diffusely reflected by snow and rain, and the image is captured.

In addition, it also occurred when an object moving around the equipment blocked the window of the camera that photographed the front license plate image, and the image could not be captured.

In addition, it was caused by the reason that dirt was thrown on the front license plate of the vehicle due to snow and rain, making it impossible to read the vehicle number.

In addition, the vehicle number cannot be recognized when the daytime illumination is within 3m of the visible distance due to fog, or when the daytime illumination is still dark, etc., in the case of electric vehicle license plates, only the license plate is dark, so the illumination is daytime illumination. There was a case where the vehicle number could not be recognized because is not turned on.

As described above, most of the error phenomena that occur in natural phenomena are largely divided into two types.Infinity error illuminance is instantaneously generated in natural phenomena during video recording, making it impossible to adjust the illuminance or This is the reason why the front license plate cannot be read due to a physical phenomenon.

In this case, since one cannot adjust the illuminance itself, the visible light itself is forcibly blocked and the vehicle number is recognized with only infrared light without illuminance control.

In the other method, illumination control analysis is possible, but it is impossible to table the number of all cases where infinity illumination occurs, and even if it is made into a table, it takes too much time to analyze. A predictive analysis method has been used in which the limit values of the maximum and minimum moving ranges that can change abruptly compared to the illuminance are set, and three different images are created for number recognition.

Among the two conventional vehicle number recognizer systems, the technology method of recognizing vehicle number without adjustment of illuminance is inevitable to prevent errors due to infrared reflected light by sunlight, and because the image is photographed only with infrared light, it is distinguished between day and night. I had to shoot only black and white images, which are always nighttime.

In addition, even with other technical methods of recognizing vehicle numbers by adjusting the conventional illumination, the method of creating an annual illumination schedule takes a lot of analysis time, and in the case of various special natural phenomena, a number recognition error occurs or it is always necessary to shoot only black and white images. The recognition error rate was inevitably small.

In the case of shooting only black-and-white images, when applied to an autonomous driving system, it is impossible to apply because the type, direction of movement, and shape of the object must be analyzed, and the occurrence of misrecognition errors is suppressed because it is not possible to adjust illumination and analyze illumination. I couldn't have it.

When the infrared light is turned off, the vehicle number recognition itself becomes impossible, and the infrared light has the lowest energy efficiency rating because it consumes the most power among all parts in the vehicle number recognizer. In addition, irrespective of day or night, the infrared light must be turned on like a psychedelic light, and thus, the life span has been shortened.

Another way to solve this problem is that it is possible to adjust the illumination for each part in the entire area, but there is only a rack tangle area that allows one analysis area to be analyzed, depending on the equipment installation location, vehicle entry method, and vehicle speed. , As the position of the license plate to be photographed is photographed differently each time, a problem occurs that the analysis position is not accurate.

In the case of snow or rain falling on the floor, different values of illuminance are measured under the influence of the headlights of the vehicle entering the vehicle, and even the annually measured illuminance values are also infinity, and even if all the values can be plotted, the number of cases If there are too many, the speed of finding the threshold value becomes slow and cannot be used.

In addition, even if the threshold value is known, when the register value (shutter, gain, aperture value) is in a large range from † to the threshold value, the camera CPU speed does not keep up with the moving speed of the vehicle.

Therefore, even if any illuminance is measured, analysis of the position of the license plate and the partial area where the license plate is located in the image taken before the vehicle moves, and the artificial intelligence automatic illuminance control technology that quickly reaches the threshold value at a time must be invented to analyze various images in the photographed image. You can analyze the data accurately. When the vehicle enters the vehicle, the difference between the current measured illuminance value and the difference between the current measured illuminance value is small and the time required to move to the threshold value can be shortened.

In the case of shooting an image with three types of thresholds by making the maximum, minimum, and the same value compared to the previous illumination value, errors such as backlight, reflected light, and metering can be prevented to some extent. I can't prevent it.

That is, when small clouds are continuously moving, the illuminance is continuously rapidly rising, a shadow is formed only around the license plate of the vehicle, or an electric vehicle enters during the daytime and when it is slightly dark.

In addition, when the vehicle enters differently from the direction from which it is viewed due to the installation of the equipment, or when the side of the vehicle has to be photographed by various entry methods, or when the vehicle is taken outside the illumination analysis area where the license plate is located, etc. to be.

In addition, this is a case in which a large amount of change in threshold illuminance occurs due to a weather condition in which the illuminance change range is too rapidly changing out of the maximum and minimum width range of illuminance control, or when the equipment is operated for the first time, or by entering a vehicle within a few hours.

In such special natural environment changes, it has become difficult to adjust the automatic illumination of video shooting to prevent errors.Also, since false recognition errors are not known, an automatic illumination control algorithm of artificial intelligence capable of blocking false recognition errors in advance is included. There is a need for a vehicle number recognition system.

As described above, due to various factors of error occurrence, the conventional vehicle number recognizer could not know the occurrence of illuminance in the case of infinity, and could not automatically adjust the illuminance to reach a threshold value compared to the infinity generated illuminance value.

In the meantime, all cameras have only had to move one step by step to the range of the threshold illuminance area that should be moved and subtracted while shooting the video every time, since the time it takes to shoot the threshold image has not been a problem.

Therefore, it took too much time to capture an image of a moving vehicle and reach the threshold illuminance at one time, making it impossible to use it as a vehicle number recognition device.

As a solution to this, there was no choice but to install a sunglass-type structure that blocks all visible light, or to take an image with only infrared lights turned on at day and night without adjusting the illumination.

Therefore, when the infrared light is blocked, it is impossible to recognize the number at all, or the infrared light must be turned on even in the daytime, and there is a problem in that power consumption is the most caused by the recognition of the car number.

In addition, there is a reason why it is impossible to shoot color images because all of them are forced to shoot in the form of night shooting, and when applying to a manned or unmanned fare machine or autonomous vehicle used to find my vehicle, image analysis must be performed through color. There is a problem that can only be followed by restrictions.

In addition, if all existing cameras uniformly adjust the entire screen within the actual shooting range in order to increase the clarity when shooting video, the dark or bright parts are automatically adjusted to illuminance at the same time, so that the dark part of the image becomes darker, and the bright part of the image becomes brighter. There is a problem in that there is no choice but to shoot a losing video.

Therefore, even though automatic illumination can be adjusted, there is a need for a technology capable of changing the illumination around the license plate to a threshold value by analyzing the location and illumination of the vehicle license plate among the entire image capturing range.

In addition, since the number recognition system cannot know whether or not the number recognition system is misrecognized, there is a need for the ability to compare and analyze the illuminance value of artificial intelligence that can reduce false recognition errors.

In addition, in order to prevent the above-described error phenomenon, before photographing an image for image analysis, an illuminance analysis to adjust the threshold illuminance must be performed first. In addition, when analyzing the illuminance, a partial image of the license plate is located. Alternatively, there is a need for a method for changing the threshold illuminance based on a partial image that needs to be analyzed or the like.

In addition, through artificial intelligence that reads images in various cases and analyzes, the corresponding partial area is changed to a specific threshold value at a time at a high speed, and then the image for image analysis is taken, and the vehicle number and various image data A technology that analyzes the lights and an algorithm that automatically adjusts the illuminance through artificial intelligence based on the camera register values (shutter, gain, aperture level, etc.) is required so that an error-free image can be taken at any time in this case. Has become.

Korean Patent Application Publication No. 10-2019-0013080 (published date: 2019.02.11)

An object of the present invention in order to meet the above requirements while solving the above-described problems is installed in a parking lot entrance and an autonomous vehicle, etc., before photographing a vehicle entering the parking lot or before an autonomous vehicle photographs the surroundings, artificial By intelligently analyzing changes in surrounding illumination, adjusting the illumination level so that no errors occur, and then photographing an image, the vehicle number entering the parking lot is recognized, or by taking an image while an autonomous vehicle is moving, various data necessary from the video are recorded in real time. It is to provide an artificial intelligence image analysis system and method that allows the user to safely drive to the destination by analyzing it with an unmanned driving method.

In addition, the present invention for solving the problems of the prior art is a vehicle number recognition system and peripheral auxiliary devices such as an image analysis system of an autonomous vehicle, a solution method, and a high-intelligent automatic illumination control program capable of performing the above method. It provides hardware, computer-readable recording media, and media for transmitting compressed images and data.

In addition, the vehicle number recognition system according to the present invention can first analyze the image in advance and block various types of error occurrence before the image including all types of errors that occur when the image causing various errors as follows is captured. By configuring a smart hardware system and high-intelligence smart software, and by completely blocking the possibility of error occurrence through this configuration, number recognition errors are taken by taking images for number recognition or images that can analyze various image data of autonomous vehicles. It is to provide an artificial intelligence image analysis system and method that can use information necessary for various image analysis during autonomous driving and a method that does not occur at all.

The types of causes of the various errors are as follows.

In other words, depending on the angle that the light source is shining on, it moves differently for each season, time, and installation location, when it snows/rains, or when there is a fog less than enough to show a visible distance within 3m, a large amount of Such as when small clouds continue.

In addition, in other natural environments, when backlight, reflected light, or metering occurs, when the light of the light source shines differently in day and night, when snow falls while driving a vehicle, snow is accumulated on the front license plate, and the headlight installation of a car is defective. This is a case that affects video recording.

In addition, when water is accumulated on the floor due to rain or snow due to a sudden change in the natural environment, diffuse reflection occurs due to the car headlights of the vehicle being entered, which makes it impossible to distinguish between day and night.

In addition, during daytime, when the visible distance is shortened by fog within 3m, diffuse reflection occurs due to infrared lamp light, the image of the threshold value around the license plate cannot be photographed due to the natural environment that changes rapidly during illumination change to the threshold value. Such is the case when it is taken.

In addition, when a shadow is generated by surrounding or objects, an image is photographed differently from the location where the vehicle number recognition system is installed and the location where the vehicle license plate is photographed.

The present invention is an aspect of a method for solving the problems of the cause of the error. Since the initial operating time of the vehicle number recognition system is not separately set, an image capable of analyzing the number at one time is captured even if it is executed at any time. It is to provide an artificial intelligence image analysis system and method capable of finding a range of basic register values (shutter, gain, aperture value) related to camera illuminance that constitutes the basic illuminance.

In addition, the present invention can be accurately known only if it starts from an illuminance value without the sun or an ambient light source and increases it gradually, but there is no time to do so † An illuminance value such that vehicle number recognition is possible with only a minimum constant infrared illumination illuminance value in an illuminance state. It is to provide an artificial intelligence image analysis system and method capable of finding a basic register value (night illumination: shutter, gain, aperture value, etc.) in which an image is photographed.

In addition, the present invention first can quickly and accurately analyze the illuminance of the area around the license plate of the vehicle in order to block various types of error occurrences accompanying the location of the license plate of the vehicle within the range of the currently photographed image. A basic setting method that allows images to always be taken within a wide range of rack tangles for illuminance analysis where the license plate is located from the vehicle speed and entry method, and an artificial that can analyze each illuminance value within all rack tangle ranges where the license plate is located. It is to provide an intelligent video analysis system.

In addition, the present invention analyzes the vehicle size with an automatic vehicle type determination sensor first when a vehicle within the entry and exit shooting range enters, and a loop detection sensor or an infrared detection sensor detects the entered vehicle, and thus DIO (IN/OUT BOARD). When the detection signal is sent to the board and the DIO (IN/OUT BOARD) board sends the incoming vehicle signal to the PC, the PC is the front/rear signal, the left/right type signal, which camera is the priority of the illumination control, and the simultaneous illumination. It distinguishes whether or not the camera is in the order of control, and sends a signal to determine whether to shoot an image without turning on or on the infrared lamp to the DIO (IN/OUT BOARD) board, and the driver's speeding driving habits to the image shot within the entire shooting range. Regardless of the direction of vehicle entry, the vehicle license plate is cut off or the location of the vehicle license plate is always photographed within the illuminance analysis area, and an artificial intelligence video analysis system and method that can satisfy the number of various cases through various basic setting setting functions. Is to provide.

In addition, the present invention first analyzes all the illuminances of the area where the license plate is located before photographing the image for vehicle number recognition, and finds the illuminance adjustment range with high-intelligence smart software that can satisfy the number of various cases with the analyzed illuminance value. In any case, high-intelligence smart illumination control software that can shoot vehicle number images by setting up the illumination register values (shutter, gain, aperture value, etc.) that move to the threshold image range that can only be analyzed in any case. It is to provide an artificial intelligence image analysis system equipped with.

In addition, the present invention analyzes the illuminance analysis result as an infinite case, and in order to solve the case that the illuminance value at which the license plate of the vehicle is located reaches the threshold value as the analyzed result value, the threshold value is easily set at once. Initial illuminance with an illuminance schedule that can apply the types of modes that can classify the entire day's illumination range mode into several stages to find, and various types of detailed register values for illumination adjustment for each zone that can be adjusted to detailed illumination by division mode. When the result of the illuminance value of the image applied for number recognition does not reach the threshold value compared to the illuminance register value (shutter, gain, aperture value, etc.) used for confirmation (night use), the detailed illuminance zone to be applied for the next illuminance analysis It is to provide an artificial intelligence image analysis system and method including a high-intelligent smart illumination scheduling algorithm capable of automatically calculating a register value that is added or subtracted by a range.

In addition, the present invention includes a wide range of illuminance analysis area, which is a rack tangle that is expected to have a vehicle number within the entire image range, and is divided into a daylight analysis area and a night illumination analysis area, and the daytime illumination analysis area is vehicle It can be analyzed regardless of the speed of the vehicle and the entry method of the vehicle, and the entire area where the front and rear license plates are located can be analyzed.At night, the range of illumination analysis is available in areas and directions that are not affected by the vehicle's headlights. It is to provide an artificial intelligence image analysis system and method that are not included.

In addition, the present invention provides an artificial intelligence image analysis system and method that divides the daytime illuminance analysis area into an area for illuminance analysis when entering the left and an area for illuminance analysis when entering the right according to the equipment installation direction.

In addition, in the present invention, since the installation position and the photographing position of the vehicle number recognition system are different from each other, various types of backlighting, reflected light, and/or other types of backlighting, reflective light, and/or other types of backlights, It is to provide an artificial intelligence image analysis system and method capable of automatically analyzing and blocking errors in which vehicle number recognition is impossible due to metering and surrounding objects due to shadows or misrecognition.

In addition, the present invention is an artificial intelligence image that allows a perfect image to be captured after converting to a threshold image that enables accurate image analysis at once in any case by pre-recognizing changes in natural environment, climate, sunlight, etc. It is to provide an analysis system and method.

In addition, in the present invention, even if the change of illumination of the day is infinite, and different illuminations are generated for each season of the year, so that a pre-changed illumination analysis is possible, a large amount of small clouds are generated even while changing to a threshold value around the license plate. In the case of moving days, when the actual number recognition image is taken, it is often taken beyond the threshold value.Therefore, the artificial intelligence illuminance analysis and predictive cognitive intelligence capable of responding to such sudden increase and decrease illuminance are performed immediately before and before In order to grasp and analyze the flow of illuminance, and to comprehensively judge it to move to the expected illuminance, the vehicle has minimized the possibility of errors in unexpected analysis by mode group and mode group with a large frame that are taken all day. It is to provide an artificial intelligence video analysis system and method that enables smart illumination adjustment that can be moved to the buffer illumination range where number analysis is possible.

In addition, the present invention is the position of the sun in the position that generates backlight even in the case of daytime, snow, rain, or cloudy weather even when the expected illuminance of the entire wide range of rack tangles expected to be photographed including the license plate is accurately analyzed. In this case, the number of cases that can be analyzed even when a daytime video that only shows the vehicle number text in a form that is out of focus is not visible because the illumination around the license plate is lower than that of the entire vehicle taken with all cameras. It is to provide an artificial intelligence image analysis system and method to which intelligent illumination analysis technology is applied.

In addition, the present invention can read the vehicle license plate from the rear license plate or from the left and right when the front license plate is photographed when an image in which the front license plate cannot be recognized due to snow, rain or dirt throwing or poor access method, high speed, tail bite, etc. It is to provide an artificial intelligence image analysis system and method to which the technology is applied.

In addition, the present invention determines that it is impossible to adjust the illuminance due to the occurrence of illuminance in the case of infinity, and in order to block the visible light itself, a dark sunglasses component that blocks visible light is installed on the front of the camera, and always infrared rays regardless of day or night. Since it is only recorded in black and white images with the lamp turned on, the driver's awareness of finding their vehicle is lowered, and the infrared lamp is cut off. In order to prevent the phenomenon that the vehicle number recognition itself becomes impossible in the event of a loss, color images that enable analysis of various images such as a vehicle number recognition system or autonomous vehicle using an artificial intelligence illuminance control method, and infrared rays during the day It is to provide an artificial intelligence image analysis system and method that enables vehicle number recognition with only sunlight for energy saving purposes without lighting or light of a certain amount of light source.

In addition, the present invention is a selective high-intelligence algorithm that can lower the misrecognition rate by reading both front and rear license plates, a high-intelligence algorithm that turns on infrared light and shoots an image for illuminance recognition at the time of dawn, When the illuminance is recognized as daytime illuminance again, it is possible to automatically adjust the illuminance according to the level of water level adjustment of smart intelligence, which makes it possible to adjust the illuminance that is mitigated to the illuminance area that buffers the sudden illuminance conversion due to the high possibility that an error occurred image can be photographed It is to provide an artificial intelligence image analysis system and method to enable.

In order to achieve the above object, an artificial intelligence image analysis system according to an embodiment of the present invention includes: an image photographing apparatus for first obtaining an image for illuminance analysis by photographing a vehicle or an object, and then obtaining an image for object recognition; A basic illumination register unit in which basic illumination levels for adjusting the shutter, gain, and aperture level of the image photographing apparatus are stored according to start, night, and adjustment modes; An illuminance analysis unit for analyzing the image for illuminance analysis according to an artificial intelligence automatic illuminance control algorithm and selecting a daytime illuminance rac tangle region or a night illuminance rac tangle region according to an entry direction of the vehicle; For the selected rack tangle area, based on the illumination value of the start mode through an artificial intelligence automatic illumination control algorithm, the range of the illumination level that is adjusted within the zone range for each mode is selected and is provided as a critical illumination value. part; By applying the provided threshold illuminance value to the image capturing device, controlling the image capturing device to acquire the object recognition image by photographing a vehicle or object based on the threshold illuminance value, and based on the object recognition image Accordingly, it may include a central processing unit (CPU) that recognizes the vehicle number or controls the vehicle to be autonomously driven.

In addition, the central processing unit may store the threshold illuminance value as a basic register value and use it to analyze an illuminance analysis image obtained by photographing a vehicle or an object next time.

In addition, a mode selection unit for selecting one of a night mode, a dawn mode, a morning mode, an appropriate mode, a day mode, a fixed night mode, and an adjustment mode based on the image for illuminance analysis may be further included.

In addition, it may further include an equipment environment setting unit for setting the vehicle entry method, equipment installation for each type, vehicle number recognition method, and illumination control priority.

In addition, the image capturing apparatus may include a front roof sensor that recognizes the front of the vehicle; A rear roof sensor that recognizes the rear of the vehicle; A front camera unit for photographing the front of the vehicle; A rear camera unit for photographing the rear of the vehicle; A front infrared illumination unit that transmits infrared illumination to the front of the vehicle; A rear infrared illuminator for transmitting infrared illumination to the rear of the vehicle; A vehicle type determination sensor for determining a vehicle type based on the object recognition image; And transmitting a vehicle entry signal to the central processing unit based on the determined vehicle type, the front roof sensor, and the signal detected through the rear roof sensor, and taking a front view based on a control command received from the central processing unit. Input/output for transmitting a command signal to the front camera unit or transmitting a photographing command signal to the rear camera unit, and controlling an infrared light to be transmitted to the front or rear of the vehicle through the front infrared illumination unit or the rear infrared illumination unit ( DIO) controller.

In addition, a circuit breaker operation unit that performs an operation of raising or lowering the breaker so that the vehicle can pass through; An electric sign module that displays the analysis data of the central processing unit, the analysis data of the illuminance analysis unit, and the analysis data of the zone range analysis unit for each mode in text or screen; And an electric sign controller for controlling a display state of the electric sign module.

In addition, a communication unit for communicating with an external device is further included, and the central processing unit may control to transmit various kinds of analysis data to the external device through the communication unit.

In addition, the central processing unit blocks all errors by first analyzing the illuminance at the location where the vehicle license plate is located in the photographed illumination analysis image before the object recognition image is captured through the image capturing device. The object after changing the camera register value including the shutter, gain, and aperture values for the image capture device through an artificial intelligence automatic illumination control algorithm so that the image is captured based on the illumination threshold value based on the image at the location Recognition images can be taken and used for vehicle number recognition, various image data analysis, and object recognition for autonomous driving.

In addition, the central processing unit controls the equipment environment setting unit to automatically set various environments so that the license plate of the moving vehicle is always included in the wide rack tangle to be photographed, and analyzes the range of the illuminance analysis area of the rack tangle. It is divided into area and night illumination analysis area, and the rack tangle illumination analysis range can be automatically set according to the vehicle entry method and equipment installation location.

In addition, the central processing unit, for the illuminance analysis area, through the illuminance analysis unit, set the rack tangle range of the entire illuminance analysis area to zone 0 to 17, and the rack tangle range of the weekly illuminance analysis area to zone 1 to Set to 16 zones, set the range of rack tangles in the night illumination analysis area to zones 0 and 17, and set the range of rack tangles on the left side, which has less influence on the vehicle's headlights, from zones 6 to 16 depending on the equipment installation location. And, the range of the right-hand mounting rack tangle can be set to 1 to 8 and 10 to 12 areas.

In addition, the central processing unit, according to the artificial intelligence automatic illumination control algorithm, for the illumination of a detailed zone in the illumination mode, a shutter value per level is an illumination value of 10 stages, a level illumination and a gain value of 4 stages. When determining the falling illuminance based on the appropriate mode, it is divided into a maximum of 14 zone steps, and if it is determined that the illuminance is higher than a certain standard, the register values for the shutter, gain, and aperture are rapidly lowered, and it is determined as the rising illuminance above a certain standard. It is possible to move the register value in which the illuminance value increases for each zone of up to 6 stages from the time of daylight mode to the appropriate mode.

In addition, the central processing unit, when all of the illuminance values within the rack tangle range are analyzed at night, and the illuminance value is greater than or equal to the dawn mode and 6 or more rack tangle values than the total 11 rack tangle values belong to the night mode, The next lower rack tangle illumination value than the maximum illumination value differs by 30 or more, or 20:00 to 04:00, which is a fixed night time per year, or 2 or more rack tangle values with low illumination belong to night time. In some cases, it is possible to control to capture an object recognition image while the infrared light is turned on in synchronization with the camera for a short moment by recognizing it as a night mode.

In addition, the central processing unit, when dirt due to snow or rain is thrown on the front license plate of the vehicle, or a moving object obscures the front camera unit, causing unrecognized, or when the front loop fails, the rear loop sensor, the It is possible to control to recognize the rear license plate of the vehicle through the rear camera unit and the rear infrared illumination unit.

In addition, the central processing unit automatically determines which image is closer to the threshold value in the appropriate mode in which the illuminance value of the dual image recognizing the front vehicle number and the rear vehicle number of the vehicle is By comparing and analyzing, the number recognition data of the image close to the appropriate mode can be used as a basic value.

In addition, in the case of recognizing the front vehicle number and the rear vehicle number of the vehicle, the central processing unit performs counting after determining the vehicle type or counting before determining the vehicle type in order to prevent the occurrence of an exception error through parking counting. It is possible to execute intelligent counting logic considering the maximum signal delay time so that the logic does not fall into an infinite loop when it is determined whether or not a failure of the front or rear loop sensor has occurred.

In addition, the central processing unit may control the image capturing device to obtain the image for illuminance analysis and the image for object recognition as a color image when acquiring the image for image recognition.

On the other hand, in the artificial intelligence image analysis method according to an embodiment of the present invention for achieving the above object, the basic illuminance for adjusting the shutter, gain, and aperture level of the image photographing apparatus is determined according to the start, night, and adjustment modes. An artificial intelligence image analysis method of a system including a basic illuminance register unit stored respectively, and a central processing unit that controls an image captured by the image capturing device to be analyzed according to an artificial intelligence automatic illuminance control algorithm, comprising: (a) the image capturing Before the device acquires an image for object recognition, first obtaining an image for illuminance analysis by photographing a vehicle or an object; (b) selecting, by an illuminance analysis unit, the illuminance analysis image according to the artificial intelligence automatic illuminance control algorithm and selecting a daylight illuminance rac tangle region or a night illuminance rac tangle region according to an entry direction of the vehicle; (c) The zone range analysis unit for each mode selects a range of illumination levels to be added or decreased within the zone range for each mode based on the illumination value of the start mode through the artificial intelligence automatic illumination control algorithm for the selected rack tangle area. Providing an illuminance value; (d) applying, by the central processing unit, the provided threshold illuminance value to the image capturing apparatus; (e) obtaining, by the image photographing apparatus, an image for object recognition by photographing a vehicle or an object based on the threshold illuminance value; And (f) controlling, by the central processing unit, to recognize the vehicle number based on the object recognition image or to perform autonomous driving of the vehicle.

On the other hand, the artificial intelligence image analysis system according to an embodiment of the present invention for achieving the above object is applied to the vehicle number recognition system, so that an environment of any error occurrence factor occurs before photographing a number recognition image, By analyzing with smart high-intelligence software, it is possible to analyze various images required for license plate or autonomous driving, so that the shooting state based on the partial area is changed to reach a threshold value, and then various images for image analysis can be taken.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention analyzes the vehicle size with a vehicle type determination sensor within the entry/exit vehicle shooting range, and at least one so that a roof detection sensor or an infrared detection sensor captures an image of the entering vehicle. When a signal is sent to the photographing unit, before taking the image for number recognition for the captured image, first, calculate the illuminance value from each rectangle (0 to 17) in the wide illuminance analysis area and select the rac tangle to be applied. , In various cases, the mode (night, dawn, morning, appropriate, daytime, adjustment) value was first obtained, and the zone range to be adjusted for each mode was found, and the shooting unit was moved to the range of threshold values (shutter, gain, aperture). Take an image for number recognition, analyze the number and obtain the illuminance value from the image, use the register value (shutter, gain, aperture) to create the applied threshold illuminance, and use the basic register value when analyzing the illuminance value of the next vehicle. Can be used.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention may provide a high-intelligent software applied to the vehicle number recognition system by dividing it into a smart illumination control algorithm of artificial intelligence, a number recognition algorithm, and an operating system algorithm.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention includes a group rack tangle illumination analysis unit (zones 0 to 17) that separately analyzes only the illumination around the license plate within the artificial intelligence smart illumination control algorithm, and each rack tangle. Mode selection unit (Night, Dawn, Morning, Appropriate, Daytime, Adjustment) that selects intelligent illuminance according to the number of cases as a result value of illuminance analysis by area and determines which mode the current illuminance belongs to. Mode, etc.), and the shutter, gain, and aperture values that need to be adjusted by comparing them with the shutter, gain, and aperture values used to obtain the illuminance values around the current license plate. The adjustment unit for each mode determines how much to add or decrease, and automatically calculates the range of the adjustment level by comparing and determining the range of the adjustment level. After applying it to the camera, an image for number recognition can be taken.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention includes whether to turn on the light in the illuminance analysis unit and take an image for number analysis, depending on the status of receiving the reservation data and whether to turn on/off the infrared light. As it is, the input/output controller (DIO) board is sent to the input/output controller (DIO) board, and the input/output controller (DIO) board can be operated according to the synchronization signal of infrared lighting and image capture for number recognition.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention captures an image without turning on the infrared light when analyzing the illuminance around the license plate, calculates the illuminance value for each rack tangle, and calculates the smart automatic adjustment value. After setting up the camera according to the result value, you can shoot the image for number recognition.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention first determines the case where the infrared light is always turned on, the case where the infrared light is turned on, and the case where the infrared light does not need to be turned on is first determined, and then according to the determination result. After the video is captured by operating properly, the vehicle number can be recognized from the video.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention calculates the vehicle number recognition data and the illuminance value, but in other modes except for the night mode, the illuminance value obtained for the vehicle number recognition image is far out of the threshold illuminance. In this case, the rear license plate to reduce the false recognition rate also uses the same method as the front license plate number recognition logic method to analyze the number and calculate the illuminance value, and check which of the recognized illuminance values for the front and rear numbers is close to the threshold. Smart high-intelligence can be executed to compare whether to use it as recognition data.

In addition, in the artificial intelligence image analysis system according to an embodiment of the present invention, since the area where the vehicle number is located is photographed differently according to the driving habit speed at the moment the vehicle in the daytime movement arrives at the shooting range, the vehicle number recognition system is installed in advance. By selecting, it is possible to set the illuminance recognition range for each entrance direction to analyze the illuminance of the exact license plate position.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention is the opposite of the entry direction because the vehicle headlights due to snow, rain, or abnormal weather may be reflected in the entry direction of the vehicle moving at night. The rack tangle area can be automatically set so that it is not affected, or intelligent control can be executed to automatically select the lowest illuminance value to prevent errors due to diffuse reflection in both directions.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention does not matter to drive at an average or lower speed due to the driver's driving habit, but is photographed in order to recognize the vehicle number of a driver with a speeding driving habit. Divide the video range by a quarter, and shoot the top (top left, right: 2 locations) so that the vehicle license plate is not cut off by high-speed driving of the vehicle to secure the maximum recognition effective distance and enter the left The required direction of the vehicle is to be taken from the top left, and the direction of the vehicle that is required to enter the right is taken from the top right, and the priority of illumination control can be customized according to the entry direction with the most entry method. .

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention allows the vehicle within the entire shooting range to first create a threshold image capable of image analysis even if only the license plate occurs before shooting the image for number recognition. Since the position where the license plate is photographed is variable according to the speed of the vehicle, the entire virtual area where the license plate is photographed can be analyzed in any case.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention divides the range of the daytime illumination analysis area into a maximum of 16 areas or more rack tangle areas, and is not affected by sunlight at night, but is not affected by the vehicle headlight. It is affected by snow or rain accumulated on the floor due to snow or rain due to natural climate change, but it is most affected by the way the vehicle enters, so the entry direction data that can distinguish the vehicle entry direction Can provide.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention adjusts the illuminance so that the rear license plate can be recognized through the same smart illumination control method when the front license plate is not visible due to snow or dirt throwing, and then the rear license plate is opened. I can recognize it.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention can accurately classify the size of the vehicle into small and other vehicles (large) automatically through the vehicle type determination sensor, and is loaded on vehicle number recognition data and transmitted together. In addition, it is possible to provide a function that automatically applies to a vehicle subject to a discount when paying a fee to provide a discount.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention is implemented as a vehicle number recognition system, and when equipment is installed on the left and right sides, the most entry direction is given priority to increase the vehicle number recognition speed. Provides a setting mode that allows you to select a method that can be selected or recognized simultaneously regardless of the direction of entry.When the front and rear vehicle number recognition function is set, the front license plate of the vehicle currently entering is recognized and then the rear license plate is recognized. , If the next vehicle enters before recognizing the rear license plate, the front-priority counting function recognizes the front part of the next vehicle first, and an error occurs if the front and rear license plate photographing signals do not come in for a certain period of time (arbitrary vehicle passing time). It is possible to provide a function of setting a delay time of vehicle passing to be reset.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention reads and analyzes the illuminance values in various cases with the smart automatic illuminance control technology of the artificial intelligence, changes them to the read threshold, and then shoots the image. If the illuminance is rapidly changing without a break, it may occur outside the threshold (appropriate mode illuminance value).If the image is taken above the threshold (appropriate illuminance range value), the vehicle number recognition rate error is also increased. Intelligent illuminance control to lower can be performed.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention has to decrease the illuminance value exponentially as a dimension higher than the threshold illuminance is read, and the range is in the dawn mode or morning, which is the illuminance when it is lower than the threshold value. The illuminance control and the predictive cognitive high-intelligence can be adjusted in case of a sudden rise again only after moving within the mode range. When the illuminance is analyzed to a level near the threshold value, high-intelligence illuminance adjustment that descends step by step can be performed.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention is too bright to be viewed in the night or dawn mode, etc., when the light of the sun or the light source is high, but the intensity of the light is over the sunset as the backlight is generated. When the camera register value (shutter, gain, aperture, etc.) cannot be adjusted high because it is bright and there is no ambient light like at night, it provides a high-intelligence dimming function that turns on both low illumination and infrared illumination at the same time. can do.

In addition, in the artificial intelligence video analysis system according to an embodiment of the present invention, the illuminance value for each mode is determined in advance in a range such as a maximum or a minimum, and the daily illuminance value varies depending on the camera type based on the brightest day. Based on the camera whose illumination level is 0 ~ 255 depending on the type, night mode (0 ~ 24 levels) illumination, dawn mode (25 ~ 32 levels) illumination, morning mode (33 ~ 90 levels) illumination, appropriate mode (91) ~120 level) illuminance, day mode (121-255 level) illumination, and the adjustment mode is automatically selected as a special function control mode that prevents sudden drop to the night mode level illuminance above the dawn mode. In order to prevent the illuminance from occurring, it is possible to adjust the illuminance by step increase or decrease for each mode, and to adjust the illuminance that decreases exponentially according to the maximum measured illuminance value in the mode of daytime or higher that requires a rapid fall.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention, when calculating the basic illuminance value (mode, zone classification) of illuminance adjustment, always recognizes the night register value or the previous vehicle number in a state where the infrared light is not turned on. The register value used at the time, or the register value to be transferred to the threshold value calculated through automatic illumination adjustment of artificial intelligence to reach the appropriate mode when the illumination value used to recognize the previous vehicle number is out of the appropriate mode is applied. In the late night mode, it is possible to analyze the illuminance in the 0 and 17 rack tangle areas, and in the case of the dawn mode or higher, all the illuminance in the 1 to 16 rack tangle areas can be analyzed.

In addition, the artificial intelligence video analysis system according to an embodiment of the present invention, when it is determined as the night mode in the illumination mode or more than the dawn mode, when the vehicle enters the video recording range of the rack tangle areas 1 to 16 around the license plate. , In order to calculate the naturally occurring illumination value, in any case, after blocking the light from the infrared lamp, first take an image for illumination analysis, and then enter the night illumination mode at each illumination value in the entire rack tangle area (1~16). You can switch to night mode only when the corresponding illuminance value is read.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention, when photographing an image for object analysis while the infrared light is on, the illumination analysis result value is included in the night or dawn illumination mode value, or the dawn mode illumination The result of analyzing the rack tangle area (zones 1~16) within the mode above the value. If the size of the rack tangle value different from the maximum rack tangle value is 50% or more, two or more rack tangle values are analyzed within the range of night or dawn mode, or when the infrared light is always turned on, etc. Regardless of the mode range, images for object analysis can be taken with the infrared light turned on.

In addition, the artificial intelligence video analysis system according to an embodiment of the present invention, when it rains in a state of more than the dawn mode, when backlighting occurs, or when the illumination is higher than the morning illumination mode, but the license plate color of the electric vehicle is dark, so the license plate illumination value is not low, but the vehicle number analysis An image for object analysis can be taken while the infrared light is turned on through the artificial intelligence smart illuminance determination that automatically reads when a special phenomenon occurs that becomes impossible.

In addition, the artificial intelligence video analysis system according to an embodiment of the present invention, in the case of a night illumination mode, and snow or water is accumulated on the floor due to snow or rain, diffuse reflection is generated from the headlights of the entering vehicle, so when switching to another mode, too much. The high-intelligence automatic adjustment software part in charge of the number of cases in the smart illumination control that prevents number recognition from becoming impossible because it is photographed in darkness is the rack tangle opposite to the vehicle entry direction among the two rack tangle areas (0, 17) for night use. It is based on the standard of the illuminance analysis area and the lowest illuminance analysis value, and can be switched to another mode via the adjustment mode only when all of the rack tangle areas for night analysis (Zone 0, 17) are out of the night illuminance range. .

In addition, the artificial intelligence video analysis system according to an embodiment of the present invention, when the illumination analysis result in the morning mode does not reach the appropriate mode illumination value, increases the gain by one step (4 steps) in the register, and increases by more than 98 steps. When it reaches the shutter value (10 steps), the shutter value (10 steps) is increased, but if the ratio between the gain and the shutter value differs by more than 10%, artificial intelligence prevents the difference from more than 10% through the high-intelligent raster value uniform increase or decrease processing. It is possible to provide an artificial intelligence automatic illumination control technology that automatically increases or decreases after going through the arithmetic process of, and automatically increases the maximum value of the shutter to the limit range so that it does not exceed 140.

In addition, the artificial intelligence video analysis system according to an embodiment of the present invention, the entire mode change range (night to day mode) is an illuminance value of 0 to 255 level, even if the illuminance value measured in any mode is always a threshold illuminance mode. Through artificial intelligence automatic adjustment software so that it can be moved to the appropriate mode illuminance value (91~120 levels), the result of high-intelligence analysis based on the number of various cases can be comprehensively determined and then moved.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention recognizes the front license plate due to the occurrence of dirt thrown on the front license plate due to snow or rain, only part of the license plate being covered by the shadow, diffuse reflection of vehicle headlights, or interference from surrounding objects. Even when the camera is covered, or when an error occurs due to speeding or biting the tail, the image of the rear of the vehicle can be analyzed using the same high-intelligent illumination control technology.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention can be replaced because it is possible to determine unrecognized among various errors caused by backlight, reflected light, and metering due to sunlight. It is impossible to determine the false recognition itself, but if various errors occur, the image analysis value reaching the threshold illuminance can be applied to the threshold value (appropriate mode value) because the probability of occurrence of false recognition of the license plate image is low. have.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention does not block light rays during image capture, so that color images are captured during the day, and images close to black and white images are captured only at night. , You can increase the awareness of finding videos in an unmanned fare calculator or a car finding kiosk.

In addition, the artificial intelligence image analysis system according to an embodiment of the present invention is applied to the vehicle number recognition system, and the entry direction of the vehicle is changed according to the installation location, and the effective shooting distance of the vehicle being entered can be shortened. Since the license plate of a vehicle entering rapidly can be cut and photographed, it is possible to provide a function to set the left or right entry method for each equipment installation location in order to photograph the longest effective distance.

In addition, in the artificial intelligence video analysis system according to an embodiment of the present invention, since the vehicle entry method such as the left entry, the right entry, and the straight entry is different from the road on which the vehicle is going to the parking lot, each entry method is As a result, an image that makes it impossible to photograph the license plate is taken, so even if a dual device on the left and right sides is installed in a single passage to recognize the number, regardless of the number of entry methods, priority is given to the place with the largest number of vehicles entering the left or right side per day. Since the recognition speed can be recognized more quickly by placing a priority, the function setting to select it and the recognized left and right image analysis data are selected to adjust the illumination by selecting only one camera set for fast and accurate recognition even when adjusting the illumination. The recognition speed can be faster and more accurate with the ability to apply the same illuminance value to the camera or adjust illuminance at the same time on both sides.

In addition, the artificial intelligence video analysis system according to an embodiment of the present invention, in order to find the start basic register value (shutter, gain, aperture value, etc.) when the first vehicle number recognition system is executed, at night when the lowest illumination is detected during the day. The case corresponds to †illuminance, but “because the vehicle number cannot be recognized at 0 illuminance, the register threshold minimum illuminance value that allows number recognition while the infrared light is turned on to find the register value that transmits light of a certain illuminance. First finds and sets it as a basic register value that can be applied regardless of day or night.When this basic register value is applied, the register value is a register value that enables vehicle number recognition within a certain speed. It finds and applies the range to which the basic register value can be applied within 10ms, which is a possible speed. However, in order to apply this value only in night mode, the minimum shutter value can be applied so as not to exceed 10 steps and 16 gain steps.

In addition, since the artificial intelligence image analysis system according to an embodiment of the present invention does not know in which time period the basic register value is photographed, image analysis is possible in any case when the vehicle number recognition system or autonomous vehicle equipment is initially started. In order to shoot an image, the basic register value is applied by default, and the current ambient illuminance mode is analyzed through this light. If the value is lower than the night illuminance level, turn on the infrared light to the night register value and search for various images. You can shoot the required illuminance value video.

According to the present invention, before photographing a vehicle entering a parking lot or before a self-driving vehicle photographs the surroundings, an image is taken after automatically adjusting the illuminance so as not to cause an error by analyzing the ambient illuminance change with artificial intelligence, It recognizes the vehicle number entering the parking lot, or captures an image while an autonomous vehicle is moving, analyzes various data necessary from the image in real time, and can safely drive to the destination in an unmanned driving method.

In addition, the present invention is a type in which an error occurs at the moment the vehicle enters the image capturing range due to the sun and ambient light sources, such as backlight, reflected light, and metering, and only a part of the license plate is shadowed, or the front license plate filth due to snow or rain is thrown. In any case, prior to shooting the video, including the occurrence of the phenomenon, the visibility is less than 3m due to fog, the diffuse reflection due to the headlight of the vehicle, the moving object temporarily covers the front license plate camera, etc. However, if it is impossible to block errors artificially, the image is captured by adjusting the illuminance of the rear license plate so that the vehicle number recognition is always possible.

In addition, when the present invention is applied to an autonomous driving system, an image can be captured with a camera to analyze an image in real time. However, when a natural phenomena in which image analysis itself is impossible by photographing a natural phenomenon in the prior art occurs, image analysis becomes impossible. The video is not only captured, but also leads to a vehicle accident, and the conventional method cannot follow the illumination adjustment time compared to the speed of the moving vehicle, so there is a limitation when applied to an autonomous vehicle with only a camera, but through the present invention With the existing camera alone, it is possible to shoot a threshold image that is an appropriate illuminance only once in a moment in any case, as well as a single threshold image based on the image for each area that needs to be analyzed. Only one can bring the effect that can be applied to all the equipment of autonomous driving.

In addition, the present invention, through high-intelligent smart automatic illumination control as described above, the vehicle speed, entry method, equipment installation direction, the front license plate in the case where the front license plate cannot be recognized, the photographing camera is covered, or the front license plate It is possible to provide a vehicle number recognition system in which vehicle number analysis is possible even in all cases due to reasons such as dirt throwing.

In addition, the present invention is a result obtained by analyzing an image that has been automatically illuminated by artificial intelligence, but in the case of a sudden change in natural environment climate change, the actual photographed image may be photographed out of the threshold illuminance. It is impossible to determine the false recognition state due to increased occurrence, but it is a method of preliminary analysis of the predicted state of false recognition, and which license plate is closest to the threshold value (appropriate illuminance value) through recognition of the front and rear license plates of the same vehicle. It is possible to provide an intelligent vehicle number recognition system capable of analyzing even misrecognition errors by automatically selecting and applying to determine whether or not a photo has been taken.

In addition, since all conventional vehicle number recognition systems cannot perform illuminance analysis, the only way to completely block visible light itself is to take images by installing a dark sunglasses-type structure device on the front of the camera. The color image itself could not be photographed, and in order to find a vehicle, it was only necessary to search with the vehicle number, but the present invention does not block visible light, so it is possible to shoot a color image, which is necessary for image analysis when applied to an autonomous driving system. All analyzes such as color analysis and shape analysis of objects to be performed have been enabled, and there is an effect of analyzing the image with color images so that vehicle damage can be easily recognized by the fare machine of the vehicle number recognition system.

In addition, the conventional vehicle number recognizer is not only unable to automatically adjust the illuminance and must completely block visible light and take an image. Therefore, since it is impossible to take an image without turning on the infrared light every time the image is taken, the infrared light is turned on both day and night. Although the image was photographed in the state of being turned on, the present invention turns on the infrared light by accurately distinguishing whether the illumination around the license plate is night or day in the entire range of images being photographed, so that the infrared light that consumes the most energy is turned on at night. Only when the determined illuminance is analyzed, the energy saving effect can be obtained by turning on the infrared light and taking an image.

In addition, conventionally, the installation of the vehicle number recognition equipment is limited because the access road of the building is short, the vehicle enters in various directions, or changes in the environment in which the front license plate of the vehicle cannot be photographed occur. Although many cases have occurred, the present invention can obtain an innovative vehicle recognition technology that does not apply to all such limiting factors.

1 is a block diagram schematically showing the overall configuration of an artificial intelligence image analysis system according to an embodiment of the present invention.
2 is a diagram illustrating an example of classifying all modes of a day in an artificial intelligence image analysis system according to an embodiment of the present invention.
FIG. 3 is a diagram showing a high-intelligent smart automatic illumination control process of an artificial intelligence image analysis system according to an embodiment of the present invention, and an adjustment range of the zone illumination intensity.
FIG. 4 is an environment setting diagram illustrating a basic setting method for taking pictures regardless of vehicle speed compared to various vehicle entry methods and equipment installation locations according to an embodiment of the present invention.
5A and 5B are screen area display diagrams showing a wide range of rack tangle areas in which the license plate of a vehicle is photographed and intelligent day and night analysis areas compared to analysis illuminance.
6 is a diagram showing an overall illumination control system diagram of an artificial intelligence image analysis system according to an embodiment of the present invention.
7 is a diagram illustrating that a central processing unit executes a recognition algorithm and a system control algorithm in an artificial intelligence image analysis system according to an embodiment of the present invention.
8 is a diagram illustrating an operation flowchart for explaining an artificial intelligence image analysis method according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those of ordinary skill in the art may easily implement the present invention. The present invention may be implemented in various forms and is not limited to the embodiments described herein.

In order to clearly describe the present invention, parts irrelevant to the description have been omitted, and the same reference numerals are assigned to the same or similar components throughout the specification.

Throughout the specification, when a part is said to be "connected" with another part, this includes not only "directly connected" but also "electrically connected" with another element interposed therebetween. . In addition, when a part "includes" a certain component, it means that other components may be further included rather than excluding other components unless specifically stated to the contrary.

When a part is referred to as being "on" another part, it may be directly on top of another part, or other parts may be involved in between. In contrast, when a part is referred to as being "directly above" another part, no other part is involved in between.

Terms such as first, second and third are used to describe various parts, components, regions, layers, and/or sections, but are not limited thereto. These terms are only used to distinguish one part, component, region, layer or section from another part, component, region, layer or section. Accordingly, a first part, component, region, layer or section described below may be referred to as a second part, component, region, layer or section without departing from the scope of the present invention.

The terminology used herein is for referring only to specific embodiments and is not intended to limit the present invention. Singular forms as used herein also include plural forms unless the phrases clearly indicate the opposite. The meaning of “comprising” as used in the specification specifies a specific characteristic, region, integer, step, action, element and/or component, and the presence of another characteristic, region, integer, step, action, element and/or component, or It does not exclude additions.

Terms indicating a relative space such as "below" and "above" may be used to more easily describe the relationship of one part to another part shown in the drawings. These terms are intended to include other meanings or operations of the device in use together with their intended meaning in the drawings. For example, if the device in the drawings is turned over, certain parts described as being "below" other parts are described as being "above" other parts. Thus, the exemplary term “down” includes both up and down directions. The device can be rotated by 90 degrees or by other angles, and terms that refer to relative space are interpreted accordingly.

Although not defined differently, all terms including technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs. Terms defined in a commonly used dictionary are additionally interpreted as having a meaning consistent with the related technical literature and the presently disclosed content, and are not interpreted in an ideal or very formal meaning unless defined.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those of ordinary skill in the art may easily implement the present invention. However, the present invention may be implemented in various forms and is not limited to the embodiments described herein.

First, in order to help understand the present invention, the general concept of the present invention will be described.

In the present invention, when a camera used in a vehicle number recognition system or an autonomous vehicle, etc., captures an image within an area that is immediately needed in real time from a vehicle entering or moving within the shooting range, the sun, shadow, snow, rain, fog, car head Various errors occur due to light, various entry methods of the vehicle, speeding, biting the tail, and the reason for covering the camera window due to objects moving around the equipment, etc. In the case of city, the rear license plate of the vehicle is photographed, in any case, when the vehicle number is recognized, unrecognized or misrecognized, etc. are blocked, the image is captured, and then the vehicle number is recognized in the image, or in the case of autonomous vehicles, various images It is possible to accurately analyze vehicle number recognition and various image analysis at any time by taking images through smart high-intelligence automatic illuminance control, which blocks the natural environment occurrence conditions that may cause errors in consideration of all the above. To be able to.

In addition, in the present invention, it is most important to take an image for obtaining various image analysis data after adjusting the illuminance with artificial intelligence in advance, such as when capturing a vehicle number recognition image or an image in an autonomous driving state. .

In addition, in the case of the vehicle number recognition system, priority is given to changing the image within the entire shooting range to a threshold value (appropriate illuminance) image for which only a specific area where the license plate is determined is used.

In addition, according to the present invention, the entire screen is always captured within a threshold value range so that an image that makes various image analysis impossible due to a sudden change in natural environment is not captured in image capturing of an autonomous vehicle.

In addition, according to the present invention, even if an error occurs in any case by analyzing the real-time illumination, the automatic illumination control of the artificial intelligence is performed so that an image capable of various image analysis can be captured after blocking in advance.

1 is a block diagram schematically showing the overall configuration of an artificial intelligence image analysis system according to an embodiment of the present invention.

Referring to FIG. 1, an artificial intelligence image analysis system 100 according to an embodiment of the present invention includes an image capturing apparatus 101 to 107, an equipment environment setting unit 112, an illuminance analysis unit 113, and a mode selection unit. 114, a zone range for each mode (zone range analysis for each mode, 115), a basic illuminance register 116, and a central processing unit (CPU) 117.

In addition, it may further include a vehicle type discrimination sensor 10, a circuit breaker operation unit 108 including a circuit breaker, an electronic display module 109, an electronic display controller 110, a power supply unit 111 and a communication unit 118.

The image capturing apparatuses 101 to 107 are devices that first acquire an image for illuminance analysis by photographing a vehicle or an object and then acquire an image for object recognition later, and include a front loop detection sensor 101, a rear roof detection sensor 102, and Input/output (DIO) controller 103, front infrared illumination unit 104, rear infrared illumination unit 105, front camera (front CAM, 106), rear camera (rear CAM, 107), and the like.

In an embodiment of the present invention, the terms may be used in parallel as the front loop detection sensor 101 and the rear loop detection sensor 102 as shown in FIG. 1 as the front loop detection 101 and the rear loop detection 102. have. Further, the input/output controller 103 can be used in parallel as the DIO CONTROLLER 103 or the DIO controller 103 or the like. In addition, the front infrared illumination unit 104 and the rear infrared illumination unit 105 may be used in parallel as the front infrared illumination 104 and the rear infrared illumination 105. In addition, the front camera unit 106 is a front CAM 106, the rear camera unit 107 is a rear CAM 106, a front camera and a rear camera 106, 107, or front and rear cameras 106, 107) can be used in parallel. In addition, the electric sign controller 110 may be used in parallel with the electric sign controller 110 or the electric sign controller 110. In addition, the zone range analysis unit 115 for each mode may be used in parallel as the zone range adjustment unit 115 for each mode.

The vehicle type determination sensor 10 may determine a vehicle type such as a small, medium, large, truck, bus, or truck according to the size of the vehicle based on the object recognition image.

The front roof sensor 101 recognizes the front of the entering vehicle, and the rear roof sensor 102 recognizes the rear of the vehicle.

The input/output (DIO) controller 103 transmits the vehicle entry signal to the central processing unit based on the determined vehicle type and signals detected through the front loop sensor and the rear roof sensor, and based on the control command received from the central processing unit. The front photographing command signal is transmitted to the front camera unit, or the photographing command signal is transmitted to the rear camera unit, and infrared lights are transmitted to the front or rear of the vehicle through the front infrared illumination unit or the rear infrared illumination unit.

The front infrared illumination unit 104 transmits infrared illumination to the front of the vehicle.

The rear infrared illumination unit 105 transmits infrared illumination to the rear of the vehicle.

The front camera unit 106 photographs the front of the vehicle, and the rear camera unit 107 photographs the rear of the vehicle.

The circuit breaker operation unit 108 raises or lowers the circuit breaker so that the vehicle can pass.

The electric sign module 109 displays analysis data of the central processing unit, analysis data of an illuminance analysis unit, and analysis data of a zone range analysis unit for each mode in text or a screen.

The billboard controller 110 controls the display state of the billboard module.

The power supply 111 supplies power to the entire device.

The equipment environment setting unit 112 may set a vehicle entry method, a type of equipment installation, a vehicle number recognition method, and an illumination control priority.

The illuminance analysis unit 113 may analyze an image for illuminance analysis according to an artificial intelligence automatic illuminance control algorithm, and select a daytime illuminance rectangle region or a night illuminance rac tangle region according to the entry direction of the vehicle.

The mode selection unit 114 may select one of a night mode, a dawn mode, a morning mode, an appropriate mode, a day mode, a fixed night mode, and an adjustment mode based on the image for illuminance analysis.

The zone range adjustment unit 115 for each mode selects a range of illumination levels to be added or decreased within the zone range for each mode based on the illumination value of the start mode through an artificial intelligence automatic illumination control algorithm for the selected rack tangle area. Provided by value.

The basic illumination register unit 116 stores basic illumination levels for adjusting the shutter, gain, and aperture levels of the image capturing apparatus according to start, night, and adjustment modes, respectively.

The central processing unit 117 applies the provided threshold illuminance value to the image capturing device, and controls the image capturing device to acquire an object recognition image by photographing a vehicle or an object based on the critical illuminance value. Based on the vehicle number, it is possible to recognize the vehicle number or control the vehicle to perform autonomous driving.

In addition, the central processing unit 117 may store the threshold illuminance value as a basic register value and use it to analyze an illuminance analysis image obtained by photographing a vehicle or an object next time.

The communication unit 118 communicates with an external device.

The central processing unit 117 may control to transmit various kinds of analysis data to an external device through a communication unit.

In addition, the central processing unit 117 first analyzes the illuminance at the place where the vehicle license plate is located in the photographed illumination analysis image before capturing the object recognition image through the image capturing device to block all errors. Object recognition after changing the camera register values including shutter, gain, and aperture values for the image capture device through an artificial intelligence automatic illuminance control algorithm so that the image is captured based on the illuminance threshold based on the image at the location where the image is located. A dragon image can be taken and used for vehicle number recognition, various image data analysis, and object recognition for autonomous driving.

In addition, the central processing unit 117 controls the equipment environment setting unit to automatically set various environments so that the license plate of the moving vehicle is always included in the wide rack tangle and photographed, and the range of the illuminance analysis area of the rack tangle is adjusted to It is divided into an analysis area and a night illumination analysis area, and the rack tangle illumination analysis range can be automatically set according to the vehicle entry method and equipment installation location.

In addition, the central processing unit 117, through the illuminance analysis unit, for the illuminance analysis area, sets the rack tangle range of the entire illuminance analysis area to zone 0 to 17, and sets the rack tangle range of the weekly illuminance analysis area to 1 zone. The range of the rack tangle of the night illumination analysis area is set to 0 and 17 areas, and the range of the left-hand installation rack tangle that has little effect on the headlights of the vehicle is set to 6 to 16 areas according to the equipment installation location. It can be set, and the range of the right mounting rack tangle can be set to 1 to 8 and 10 to 12 areas.

In addition, the central processing unit 117, according to the artificial intelligence automatic illumination control algorithm, for the illumination of the detailed zone in the illumination mode, the shutter value per level is the illumination value, the level of illumination and the gain value are 4 stages. When determining the descending illuminance based on the appropriate mode, it is divided into a maximum of 14 zone steps, and if it is determined that the illuminance is higher than a certain standard, the register values for the shutter, gain, and aperture are rapidly lowered, and the rising illuminance is higher than a certain standard. When determining, it is possible to move the register value in which the illuminance value increases for each zone of up to 6 stages from the dawn mode to the appropriate mode.

In addition, the central processing unit 117, when all of the illuminance values within the rack tangle range are analyzed at night, and when the illuminance value is greater than or equal to the dawn mode, 6 or more rack tangle values are included in the night mode. Or, there is a difference of 30 or more in the next lower rack tangle illuminance value than the rack tangle of the maximum illuminance value, or 2 or more rack tangle values with low illuminance at night from 20:00 to 04:00, which is a fixed night time. If it belongs, it is recognized as a night mode, and the image for object recognition can be captured while the infrared light is turned on.

In addition, the central processing unit 117, when dirt due to snow or rain is thrown on the front license plate of the vehicle, or a moving object obscures the front camera, causing unrecognized, or in case of a front loop failure, the rear roof sensor and the rear camera unit And it is possible to control to recognize the rear license plate of the vehicle through the rear infrared lighting unit.

In addition, in order to reduce the error rate of misrecognition, the central processing unit 117 determines which image is closer to the threshold value in the appropriate mode in which the illuminance value of the dual image recognizing the front vehicle number and the rear vehicle number of the vehicle is By automatically comparing and analyzing, number recognition data of an image close to the appropriate mode can be used as a base value.

In addition, the central processing unit 117, in the case of recognizing the front vehicle number and the rear vehicle number of the vehicle, to prevent the occurrence of an exception error through parking counting, whether to count after vehicle type determination or counting before vehicle type determination In the case of determining whether to perform the operation or whether a failure of the front or rear loop sensor occurs, the intelligent counting logic that reflects the maximum signal delay time can be executed so that the logic does not fall into an infinite loop.

In addition, the central processing unit 117 may control the image capturing device to acquire an image for illuminance analysis and an image for object recognition as a color image.

On the other hand, the artificial intelligence image analysis system 100 according to an embodiment of the present invention, although not shown in the drawing, is an image capture storage unit for storing the front image and the rear image captured by the image capture devices 101 to 107 ; The central processing unit 117 may further include an algorithm storage unit for storing system control algorithms, vehicle number recognition, and various image analysis algorithms in software form.

Meanwhile, in FIG. 1, the artificial intelligence image analysis system 100 of the present invention detects the vehicle by front and rear loop detection sensors 101 and 102 in order to photograph the front license plate of the vehicle when the vehicle enters the image capturing range. Then, first, an image for illuminance analysis is taken to obtain an illuminance value in a current state in which a natural phenomenon occurs.

Then, the DIO CONTROLLER 103 generates a front image shooting trigger signal and transmits it to the front camera 106, and the front CAM 106 shoots the front image, and various errors occur in the captured image. In order to block the number, the vehicle entry method and the single or dual vehicle number recognition value are read as previously set in the equipment environment setting unit 112, and the illuminance analysis unit 113 based on the basic setting function based on the illuminance control priority, etc. ).

Subsequently, the artificial intelligence image analysis system 100 of the present invention has an error due to the mode selection unit 114 and the zone range adjustment unit 115 for each mode through the result of illuminance analysis based on the rack tangle area around the license plate. After changing the register value of the front camera 103 to the range of the threshold value (appropriate mode) that does not occur, the captured images 121 and 122 are used to recognize the vehicle number and analyze the captured illumination and various images during number analysis (123 ), and after recognizing visited vehicles, regular vehicles, and various time commuter passes, etc. through the result value, opening and closing of the circuit breaker and image data are stored, or various analysis data and image data are transmitted to an external server through the communication unit 119, and When analyzing data, the circuit breaker 108 is opened and the recognized data is sent to the electronic display controller 110 and the result is displayed through the electronic display module 109.

In FIG. 1, the artificial intelligence image analysis system 100 is a basic step of preparing an image for autonomous driving or an illuminance value around a license plate when the vehicle number recognition operation is initially operated, which is a system control algorithm of a PC. The CPU 117 transmits front and rear image photographing signals to the front and rear cameras 106 and 107 through the DIO CONTROLLER 103 to take an image.

At this time, the artificial intelligence image analysis system 100 obtains the start mode value of the basic register unit 116 and analyzes the current illuminance value, through a high-intelligent smart automatic illuminance adjustment calculation logic software (algorithm), the start mode register value If it is the lowest illuminance of the day and the illuminance value at which basic video recording is possible, the range of night (MIN) to day (MAX) illuminance mode is classified based on this illuminance value, and the register value of the currently used start mode is compared. It analyzes how much of each star must be moved to reach the zone zone register value to reach the threshold.

Subsequently, the artificial intelligence image analysis system 100 changes the register values of the front and rear cameras 106 and 107, and then waits in a ready state for operating the equipment.

At this time, in the case of the night mode, the artificial intelligence video analysis system 100 retrieves the night mode register value (shutter: 80, gain: 70, aperture: AUTO) and uses the value for illuminance analysis when entering the vehicle to the front camera 106. After applying, the captured image is applied to the vehicle number analysis value and the illuminance analysis value.

At this time, if the analyzed illuminance value during vehicle analysis is out of the threshold value (appropriate mode), the artificial intelligence video analysis system 100 changes the addition or decrease to become the threshold value for illuminance analysis in preparation for the next vehicle entry, and then the front camera 106 By changing the register value in advance, always prepare it in advance to reach the illuminance analysis threshold when entering the next vehicle.

In Figure 1, the artificial intelligence image analysis system 100, when the vehicle arrives at a position where the front and rear numbers are photographed, the front and rear roof sensors 101 and 102 detect and send it to the DIO CONTROLLER 103, and , A trigger signal is sent to the rear image capturing cameras 106 and 107 to take an image, and the front and rear CAMs 106 and 107 that receive this signal shoot the image.

The artificial intelligence image analysis system 100 according to an embodiment of the present invention includes an equipment environment setting unit 112 and an illuminance analysis unit 113, a mode selection unit 114, a zone range adjustment unit 115 for each mode, and a basic Although the illuminance register unit 116 or the like can be configured with hardware, it can also be configured with each logic of the artificial intelligence automatic illuminance control algorithm 117. That is, the artificial intelligence automatic illumination control algorithm 117 includes an equipment environment setting unit 112 and an illumination analysis unit 113, a mode selection unit 114, a zone range adjustment unit 115 for each mode, and a basic illumination register unit 116. ), etc.

The artificial intelligence image analysis system 100 according to the present invention adjusts the illuminance with a threshold value based on the license plate or a threshold based on the entire screen (autonomous driving system) so as to take an image so that no error occurs. After making the change, you can take an image to obtain data that requires various analysis.

At this time, when an image is taken in which license plate recognition is impossible when applied to a vehicle number recognizer that recognizes the front license plate, the vehicle number can be recognized as a dual with the same logic to block the unrecognized itself.

In the event of misrecognition, the artificial intelligence is applied as a result of which camera 106, 107 of the illuminance values analyzed after the front and rear cameras 106 and 107 recognized the vehicle number, is close to the threshold value. By storing all of the dual image data, it is possible to find and correct which one of the currently analyzed image data among the images when leaving the car.

In FIG. 1, when moving the register values of the front and rear cameras 106 and 107 to the threshold illuminance analysis range, only the illuminance of the license plate portion while in the night mode, the dawn mode, and the morning mode is analyzed as a night mode, or always at night. In the in-time, the front and rear infrared lights 104 and 105 are turned on, and an image is captured with the front and rear cameras 106 and 107 to match the synchronization time.

2 is a diagram illustrating an example of classifying all modes of a day in an artificial intelligence image analysis system according to an embodiment of the present invention.

That is, FIG. 2 shows the illuminance level for each day mode applied in any case compared to the total illuminance measured annually by taking an image for illuminance analysis before photographing the image for vehicle number recognition in the artificial intelligence image analysis system according to an embodiment of the present invention. It shows the range.

In FIG. 2, the case of the yearly illumination analysis result or the analysis result of the whole day mode classification 119 is infinite, and it takes a lot of time to analyze a large amount of basic data even if it is possible to schedule illumination control in the case of infinity. Can take.

Since it is impossible to schedule all illuminances in the case of infinity, when applied to a vehicle number recognition system or an autonomous vehicle, the total time required for illuminance analysis time and change to the analyzed smart illuminance value must not exceed 20ms to shoot number analysis video. Or, it can be within a time range in which an image required for autonomous driving or the like can be captured.

Therefore, in order to analyze at a high speed, the range for each mode (119) and the illumination value for each mode are identified by finding a sample of the illumination mode for a day (00:00~23:59) satisfying all cases per year and classifying the priority mode. Fig. 2 shows the result of dividing the included zone-by-zone 120 and the moving range when the illuminance values per all individual zones are to be added or subtracted by the calculation logic of artificial intelligence.

As shown in FIG. 2, the types according to the full-day mode classification 119 may be classified into a night mode, a dawn mode, a morning mode, an appropriate mode, a day mode, and a special mode (fixed night mode and adjustment mode). .

Therefore, the artificial intelligence image analysis system 100 of the present invention obtains an analysis result of a wide range of rack tangle values around a license plate through a high-intelligent smart illumination control analysis algorithm, compares it with the number of various cases, and uses the read illumination value. Find out if it belongs to.

Table 1 below shows the range of illumination values for each mode and the special mode classification (fixed night mode, adjustment mode).

That is, Table 1 shows the range of illumination levels for each mode in relation to the classification of all modes of the day in FIG. 2, and classifies the range of illumination levels for each mode for the entire day, and the values vary according to camera types.

The full day mode is based on the measured illumination, night mode (0 to 24 levels), dawn mode (21 to 32 levels), morning mode (33 to 90 levels), appropriate mode (91 to 120 levels), and day mode (121 to 255). Level), etc.

Night mode can be applied when there is no sunlight at all or when the light source is as low as 24 or less.

Dawning mode can be applied when there is sunlight, but the illuminance is as low as 33 or less, and when the backlight is generated at sunset or when the sun rises, the illuminance is less than the illuminance level of 32.

The morning mode can be applied when the brightness of the sunlight is insufficient, but the illuminance is sufficient for number recognition.

Appropriate mode can be applied when an image with an almost perfect level of illumination is taken for number recognition.

Daytime mode can be applied when the sunlight is too bright, when an error occurs at a high level of illumination due to reflected light, or when a license plate cannot be found due to metering.

The fixed night mode can be applied to a time zone (20:00 to 04:00) in which night illumination is always measured regardless of the season or time zone even though it is at night.

The artificial intelligence video analysis system of the present invention is a daytime time zone in which an abnormality in the morning mode is measured, and when the light that changes to the night mode instantaneously due to a sudden change in the natural environment is analyzed, it is immediately converted to the original state when entering the next vehicle. Because it can be changed, if the surrounding mode is skipped when the ambient mode is skipped, an image that causes an error can be photographed, so it is moved to the ambient mode that buffers it once, and when the skipping intensity mode is measured repeatedly Only move them at once.

In FIG. 2, the artificial intelligence image analysis system of the present invention analyzes an illuminance value that cannot recognize a number at night, and if it is changed to a register value capable of analyzing illuminance without infrared lighting, the image capture speed is slowed, so the vehicle number is recognized. The speed may be affected, and when the illuminance is low, the driver sets the illuminance value of the minimum amount of light that does not interfere with driving.

Table 2 below shows the basic illuminance register values.

The artificial intelligence image analysis system of the present invention registers in advance a basic illuminance register value 121 that allows sufficient number recognition to be photographed even with an infrared lamp having almost no visible light (average value = shutter: 60, gain: 70), it can be applied differently depending on the installation location of the equipment.

For example, it is to set the night register value (shutter: 40 to 140, gain: 40 to 98) and the illuminance register value (average value = 80 to 140, gain: 84 to 98) of the adjustment mode.

Since the shutter boundary value changes the illuminance rapidly, it is easy to use the shutter value, so the average one-step value has a value based on 10 levels.

FIG. 3 is a diagram showing a high-intelligent smart automatic illumination control process of an artificial intelligence image analysis system according to an embodiment of the present invention, and an adjustment range of the zone illumination intensity.

Referring to Figure 3, the artificial intelligence image analysis system 100 of the present invention, when the vehicle enters the shooting range, the front loop detection sensor 101 or the rear loop detection sensor 102, the front and rear loop signals 121 ) Is generated, and before the front camera unit 106 and the rear camera unit 107 capture the vehicle number analysis image, the image for illumination analysis is first captured (122).

At this time, since the artificial intelligence image analysis system 100 of the present invention is photographed through the front camera unit 106 and the rear camera unit 107 in which the register values used when capturing the previous number analysis image are set, , The cam register value is changed to be applied differently to the rear cameras 106 and 107 (123).

Next, the artificial intelligence image analysis system 100 is applied to the illuminance analysis in day mode (124') and night mode (125') for a wide range of rack tangles where the license plate is located through the image for illuminance analysis. The range of rack tangle is applied differently.

In other words, the artificial intelligence video analysis system 100 is in the day mode 124', depending on the vehicle entry method, the rack tangle recognition range (1~8 zone, 10~16 zone) when entering the left and the rack tangle range when entering the right side ( 6~16) are classified and applied.

In addition, in the night mode (125'), the artificial intelligence video analysis system 100 can divide the left entry into zone 0, the right entry into zone 7, etc., or apply a small value after analyzing the left and right rack tangle illumination. have.

In addition, the artificial intelligence image analysis system 100 may select and apply a mode division compared to an applied illumination value and a zone range for division of illumination in a mode (126').

In addition, the artificial intelligence image analysis system 100 may analyze the total illuminance including the rack tangle analysis area suitable for the equipment installation location. At this time, the reason why the rack tangle area for illuminance analysis is different for each equipment installation location is that when a change in illuminance around the license plate of the vehicle is expected, it affects the entire daytime mode in the generic Dawning mode illuminated by sunlight.

In day mode, you can analyze the illuminance around the license plate, and an image in which the license plate is not visible may be photographed depending on how the vehicle enters into the illuminance around the license plate during the day. Can give Therefore, the image capturing priority may be set differently for each entry direction.

In addition, the artificial intelligence image analysis system 100 can properly shoot regardless of the speed even if the vehicle speed varies according to the driver's driving habit according to the installation of the equipment, and when driving at night 125', the headlights of the speeding vehicle are affected. It is located in the lower left and right of the rack tangle range that is not affected as much as possible because it is affected by car headlights when it is exposed to rain, snow or rain, etc., and the minimum value of the two can be set as the reference illuminance value. It is possible to fundamentally prevent the occurrence of each error due to the headlight.

In FIG. 3, the artificial intelligence image analysis system 100 calculates and applies the illuminance value of the rack tangle range divided into day and night, and automatically analyzes the number of various cases through a high-intelligent smart illuminance adjustment algorithm, and the applied illuminance value Contrast mode division and illumination within the mode can be selected as a zone range for addition/decrement division (126').

In addition, the artificial intelligence image analysis system 100 can select and control infrared lighting compared to day and night logic (127), whether to shoot an image with the infrared lighting turned on (128) or not turned on before image shooting. It is possible to separately select and control whether or not an image is to be captured 129.

In addition, the artificial intelligence video analysis system 100 automatically analyzes the application of the luminance range 130 within the zone range for each mode to reach the threshold value compared to the immediately preceding luminance value for each mode, and increases/decreases and decreases the register belonging to the luminance level range. You can change the camera register value accordingly. At this time, the aperture value is automatically shifted, and in the case of the proper mode, the image is controlled to be taken without changing the register value, and if the ratio of the shutter and the gain differs by more than 10%, the image quality can be taken with high clarity. It can be leveled by forcibly changing the value using the ratio calculation formula of artificial intelligence.

In addition, the artificial intelligence image analysis system 100, after applying (131) the register value of the result of smart illumination analysis for number analysis image capture, and then photographing 124 the number analysis image through the front and rear cameras, Through the number analysis and illumination analysis 125, it is applied to the basic register value 126 to be used when the next vehicle enters.

In addition, the artificial intelligence image analysis system 100 analyzes zone zones for each mode with a high-intelligent smart automatic illumination control algorithm and calculates a register value for adding or subtracting a zone range to reach a threshold value.

FIG. 4 is an environment setting diagram illustrating a basic setting method for taking pictures regardless of vehicle speed compared to various vehicle entry methods and equipment installation locations according to an embodiment of the present invention.

That is, FIG. 4 shows the analysis range when the area to which the rack tangle belongs is day or night on the screen to be photographed and the range of the rack tangle area used for analysis when entering the vehicle by equipment installation location. This is an environment setting technology in which the area where the license plate is located can be included even when entering the vehicle number in any case in the artificial intelligence image analysis system 100 of the present invention.

Referring to FIG. 4, the artificial intelligence image analysis system 100 according to an embodiment of the present invention determines the recognition range area so that the headlight light source of the vehicle is not affected in the rack tangle area required for analysis for each possible entry direction. Can be set automatically.

In addition, the artificial intelligence video analysis system 100, the rack tangle position so as not to be affected by the headlights of the vehicle even at night, and even if rain, snow, etc. It provides intelligent rack tangle area automatic selection technology with minimum value applied to avoid being affected.

In addition, the artificial intelligence image analysis system 100 analyzes the daylight illumination by equipment installation direction according to the range to which the two groups belong to the rack tangle 122 group for daytime illuminance analysis and the rack tangle 123 group for night illumination analysis. Set the range of rack tangle for 1~16 zones.

Here, the artificial intelligence video analysis system 100 sets the rack tangle range for daytime illumination analysis of the left entering vehicle for each vehicle entry direction into zones 1 to 8 and 10 to 12, and the rack tangle range for daytime illumination analysis of the right entering vehicle Is set to Zone 6~16.

In addition, the artificial intelligence video analysis system 100 sets the rack tangle range for night illumination analysis to zone 0 and zone 17, and the lower value of the two rack tangle areas to accurately classify without error regardless of the division of the entry direction. Can be reflected.

5A and 5B are screen area display diagrams showing a wide range of rack tangle areas in which the license plate of a vehicle is photographed and intelligent day and night analysis areas compared to analysis illuminance.

In other words, FIG. 5A shows the video recording location setting range for each equipment installation location when photographing the front license plate in the image recording range expansion setting compared to the vehicle speed, and FIG. 5B shows the image recording location setting range for each equipment installation location when photographing the rear license plate. Is shown.

5, the environment setting of the artificial intelligence image analysis system 100 of the present invention is a basic setting function for minimizing the occurrence of false recognition and non-recognition.

The artificial intelligence image analysis system 100 is based on the installation of equipment that allows all images to be captured regardless of the vehicle speed associated with the driver's driving habit in order to allow the image to be captured for as long as possible to capture an effective distance that can be recognized for a long time. The shooting location is taken at the top and bottom of the screen.

First, when photographing the front license plate, an image photographing position for each equipment installation position is set as follows (126).

When installing the equipment on the left (126-a), adjust the shooting angle of the front camera 106 so that the image can be photographed on the screen 3, and when installing the equipment on the right (126-b), the shooting can be taken on the first screen. The shooting angle of the rear camera 107 is adjusted.

In addition, if there are multiple vehicle entry directions (126-c), the most entry point among the entry directions is set as priority, the vehicle entering the left is set to shoot from screen 3 to the second screen, and the vehicle entering the right is Set to shoot from screen 1 to screen 4.

Next, when photographing the rear license plate, the image capturing position for each equipment installation position is set as follows (127).

When installing the equipment on the left side (127-a), adjust the shooting angle of the front camera 106 so that it can be photographed on screen 4, and when installing the equipment on the right side (127-b), so that it can be shot on screen 2 The shooting angle of the rear camera 107 is adjusted.

In addition, in the case of multiple vehicle entry directions (127-c), the most entry point among the entry directions is set as priority, the vehicle entering the left is set to shoot from screen 4 to the screen 1, and the vehicle entering the right is Set to shoot from screen 2 to screen 3.

6 is a diagram showing an overall illumination control system diagram of an artificial intelligence image analysis system according to an embodiment of the present invention.

6, the artificial intelligence image analysis system 100 according to the present invention, when a vehicle enters the front and rear, the front and rear cameras for photographing (106, 107) capture an image and send it to a high-intelligent smart illumination algorithm. .

That is, when the front loop detection sensor 101 detects a vehicle entering the front, it transmits a front loop signal to the front camera 106, and the front camera 106 captures an image for front illumination (128).

In addition, when the rear roof detection sensor 102 detects a vehicle entry to the rear, it transmits a rear loop signal to the rear camera 107, and the rear camera 107 captures an image for rear illumination (129).

The artificial intelligence calculation logic algorithm calculates all cases so that the threshold illuminance appropriate for each case always reaches the appropriate board.

The high-intelligence smart illuminance algorithm transmits the result value to the front and rear cameras 106 and 107 through the high-intelligent smart illuminance control analysis 130 of the front and rear license plate images to change the front and rear camera register values and high-intelligent smart illuminance results. Through the (131) process, the front and rear camera appropriate illuminance register values are changed (132). After that, when the vehicle number analysis and the illuminance analysis image are taken from the front and rear cameras 106 and 107 (133) and the vehicle number analysis and the illuminance analysis result 134 for the vehicle number are obtained, the illuminance value of the image used as the number analyzer. It is compared and analyzed once more whether or not the threshold value (appropriate illuminance) has been accurately reached (135). That is, in order to reflect on the next illuminance analysis image, the critical illuminance and the illuminance value for number analysis are compared and analyzed (135).

Next, the high-intelligence smart illuminance algorithm passes the same register value when it belongs to the appropriate illuminance 136 and uses it as the basic value for the next illuminance, and if it falls within the range of the night illuminance or the fixed-time illuminance region 137, the register of the night mode unconditionally Calculate the illuminance value by transforming the value into the next illuminance value.

In addition, the high-intelligence smart illuminance algorithm finds out the zone sector through the technology of applying a range of additives and subtractions within the zone range for each mode (130) when it is analyzed within the range of light from dawn to morning or daylight (138), and enters the next vehicle. Analyze the city illuminance.

In addition, the high-intelligence smart illumination algorithm prepares to modify the register values of the front and rear image capturing cameras 106 and 107 so that they can be applied to the basic register value.

7 is a diagram illustrating that a central processing unit executes a recognition algorithm and a system control algorithm in an artificial intelligence image analysis system according to an embodiment of the present invention.

Referring to FIG. 7, in the artificial intelligence image analysis system 100 according to the present invention, the central processing unit 117 executes the system control algorithm to process the front image 130 or the rear image 140 in an image capture storage unit. Can be saved on.

Further, the central processing unit 117 may retrieve the front image 130 or the rear image 140 stored in the image capture storage unit, and perform vehicle number recognition and various image analysis through a recognition algorithm.

Further, the central processing unit 117 may control the overall operation of the artificial intelligence image analysis system 100 and data transmission/reception through a system control algorithm.

8 is a diagram illustrating an operation flowchart for explaining an artificial intelligence image analysis method according to an embodiment of the present invention.

8, the artificial intelligence image analysis system 100 according to an embodiment of the present invention, before the image capturing apparatuses 101 to 107 acquire an image for object recognition, photographs a vehicle or an object and analyzes illumination first. A dragon image is acquired (S810).

Subsequently, the illuminance analysis unit 113 analyzes the illuminance analysis image according to an artificial intelligence automatic illuminance control algorithm, and selects a daytime illuminance rac tangle region or a night illuminance rac tangle region according to the entry direction of the vehicle (S820). .

Subsequently, the zone range adjusting unit 115 for each mode selects a range of illumination levels to be adjusted within the zone range for each mode based on the illumination value of the start mode through an artificial intelligence automatic illumination control algorithm for the selected rack tangle area. It is provided as a threshold illuminance value (S830).

Subsequently, the central processing unit 117 applies the provided threshold illuminance value to the image capturing apparatuses 101 to 107 (S840).

Subsequently, the image capturing apparatuses 101 to 107 acquire an object recognition image by capturing a vehicle or an object based on the threshold illuminance value (S850).

Subsequently, the central processing unit 117 recognizes the vehicle number based on the image for object recognition or controls the vehicle to run autonomously (S860).

As described above, according to the present invention, it is installed in a parking lot entrance and an autonomous vehicle, and an error occurs by analyzing changes in ambient illumination with artificial intelligence before photographing a vehicle entering the parking lot or before an autonomous vehicle photographs the surroundings. By adjusting the illuminance so that it does not occur, and then photographing an image, the vehicle number entering the parking lot is recognized, or the image is captured while an autonomous vehicle is moving, and various data necessary from the image are analyzed in real time to safely drive to the destination by unmanned driving method. The artificial intelligence video analysis system and method can be realized.

Those skilled in the art to which the present invention pertains, since the present invention may be implemented in other specific forms without changing the technical spirit or essential features thereof, the embodiments described above are illustrative in all respects and should be understood as non-limiting. Only. The scope of the present invention is indicated by the claims to be described later rather than the detailed description, and all changes or modified forms derived from the meaning and scope of the claims and their equivalent concepts should be interpreted as being included in the scope of the present invention. .

10: vehicle type identification sensor
101: front loop detection sensor
102: rear loop detection sensor
103: DIO(IN/OUT) CONTROLLER
104: front infrared illuminator
105: rear infrared illuminator
106: front camera
107: rear camera
108: circuit breaker
109: electronic board
110: electronic display controller
111: power supply
112: Equipment environment setting unit
113: illuminance analysis unit
114: mode selection unit
115: Zone range for each mode
116: basic illuminance register setting unit
117: CPU
118: Ministry of Communications
119: All modes per day classification
120: Classification of illuminance value level range by mode and special mode
121: basic illuminance register value
122: Rack tangle for daytime illumination analysis
123: Rack tangle for night analysis
124': Range of rack tangle: 1 to 16
125': Range of rack tangle: 0, 7
126: When shooting the front license plate, set the video recording location by equipment location
127: When shooting the rear license plate, set the video recording location by equipment location
128: Image shooting for front illumination
129: Image shooting for rear illumination
130: High-intelligent smart illumination control analysis of front and rear license plate images
131: Reflecting the result of high-intelligent smart illumination adjustment of the front and rear camera register values
132: Change the value of the proper illumination register for the front and rear cameras
133: Vehicle number analysis and illumination analysis image shooting
134: Vehicle number analysis and illumination analysis result for vehicle analysis
135: Comparison of critical illuminance and illuminance value for number analysis
136: appropriate illuminance (pass)
137: night illumination, fixed time illumination
138: Dawn~morning illumination, daylight illumination
139: front image
140: rear image

Claims (17)

A front loop detection sensor 101 for recognizing a front surface of the vehicle, which first acquires an image for illuminance analysis by photographing a vehicle or an object and later acquires an image for object recognition; A rear loop detection sensor 102 for recognizing the rear of the vehicle; A front camera unit 106 for photographing the front of the vehicle; A rear camera unit 107 for photographing the rear of the vehicle; A front infrared illumination unit 104 for transmitting infrared illumination to the front of the vehicle; A rear infrared illumination unit 105 for transmitting infrared illumination to the rear of the vehicle; Includes a vehicle type discrimination sensor 10 that determines a vehicle type based on an image for object recognition, and enters a vehicle based on the determined vehicle type and signals detected through the front roof detection sensor 101 and the rear roof detection sensor 102 The signal is transmitted to the central processing unit (CPU), and a front photographing command signal is transmitted to the front camera unit 106 based on a control command received from the central processing unit (CPU), or a photographing command signal is transmitted to the rear camera unit 107 ), and comprising an input/output (DIO) controller for controlling the infrared light to be transmitted to the front or rear of the vehicle through the front infrared light unit 104 or the rear infrared light unit 105;
An equipment environment setting unit 112 capable of setting vehicle entry method and equipment installation for each type, vehicle number recognition method, and illumination control priority;
An illuminance analysis unit 113 that analyzes an image for illuminance analysis according to an artificial intelligence automatic illuminance control and selects a daylight rac tangle region or a night illuminance rac tangle region according to an entry direction of the vehicle;
A mode selection unit 114 for selecting one of a night mode, a dawn mode, a morning mode, an appropriate mode, a day mode, a fixed night mode, and an adjustment mode based on the image for illuminance analysis;
Zone range analysis unit for each mode that selects the range of illumination levels to be adjusted within the zone range for each mode based on the illumination value of the start mode through artificial intelligence automatic illumination control for the selected rack tangle area and provides it as a critical illumination value ( 115);
A basic illuminance register unit 116 in which basic illuminance for adjusting a level of a shutter, a gain, and an aperture of the image photographing apparatus is stored according to a start, night, and adjustment mode, respectively;
Applying the provided threshold illuminance value to the image capturing device to control the image capturing device to acquire an image for object recognition by photographing a vehicle or object based on the critical illuminance value, and to recognize the vehicle number based on the object recognition image, Artificial intelligence image analysis system comprising; a central processing unit (CPU) for controlling the autonomous driving of the vehicle to be executed.
The method of claim 1,
The central processing unit (CPU) is an artificial intelligence image analysis system that stores a threshold illuminance value as a basic illuminance register value and uses it to analyze an illuminance analysis image obtained by photographing a vehicle or object next time.

delete delete delete delete delete The method of claim 1,
The central processing unit (CPU) first analyzes the illuminance at the place where the vehicle license plate is located in the image for image recognition of the object captured through the image capturing device before capturing the image for object recognition. Object recognition image after changing camera register values including shutter, gain, and aperture values for the image capturing device through artificial intelligence automatic illuminance control so that the image is captured based on the illuminance threshold based on the image of the location. An artificial intelligence image analysis system that shoots and uses the image for vehicle number recognition, various image data analysis, and object recognition for autonomous driving.
The method of claim 1,
The central processing unit (CPU) controls the equipment environment setting unit 112 to automatically set various environments so that the license plate of the moving vehicle is always included in the wide rack tangle to be photographed, and the range of the illuminance analysis area of the rack tangle is set to daylight illumination. An artificial intelligence image analysis system that automatically sets the rack tangle illumination analysis range according to the vehicle entry method and equipment installation location by dividing the analysis area and the night illumination analysis area.
The method of claim 1,
The central processing unit (CPU) sets the rack tangle range of the entire illumination analysis area to zone 0 to 17 through the illumination analysis unit for the illumination analysis area, and sets the rack tangle range of the weekly illumination analysis area to 1 to 16 zones. And, the rack tangle range of the night illumination analysis area is set to Zone 0 and Zone 17, and the rack tangle range on the left side, which has little effect on the headlights of the vehicle, is set to zones 6 to 16, depending on the equipment installation location. An artificial intelligence image analysis system that allows the rack tangle range to be set to zones 1 to 8 and zones 10 to 12.
The method of claim 1,
The central processing unit (CPU) has a shutter value of 10 levels per level for a detailed zone illumination in an illumination mode according to artificial intelligence automatic illumination control, and a level illumination and gain value of 4 levels. When determining the falling illuminance based on, it is divided into a maximum of 14 zone steps, and if it is determined that the illuminance is higher than a certain standard, the register values for the shutter, gain, and aperture are rapidly lowered. An artificial intelligence video analysis system that moves the register value that increases the illuminance value for each zone in up to 6 stages before the appropriate mode.
The method of claim 1,
The central processing unit (CPU) is the case when the entire luminance value within the rack tangle range is analyzed at night, and the luminance value is greater than or equal to the dawn mode, and 6 or more rack tangle values than the total 11 rack tangle values belong to the night mode, or If there is a difference of 30 or more in the next lower rack tangle illuminance value than the rack tangle of the value, or the time zone from 20:00 to 04:00, which is a fixed night time, or two or more rack tangle values with low illuminance are in the night, night An artificial intelligence image analysis system that recognizes as a mode and controls to capture an image for object recognition while the infrared light is turned on.

delete The method of claim 1,
The central processing unit (CPU) automatically compares which image is closer to the threshold value in the appropriate mode in which the illuminance value of the dual image recognizing the front vehicle number and the rear vehicle number of the vehicle is recognized to reduce the error rate of false recognition occurrence. An artificial intelligence image analysis system that analyzes and uses the number recognition data of the image close to the appropriate mode as a basic value.
The method of claim 1,
When recognizing the front vehicle number and the rear vehicle number of the vehicle, the central processing unit (CPU) counts after vehicle type determination to prevent errors from occurring through parking counting, whether to count before vehicle type determination, front or rear. An artificial intelligence video analysis system that executes intelligent counting logic considering the maximum signal delay time so that the logic does not fall into an infinite loop when it is determined whether a loop detection sensor has failed.
The method of claim 1,
The central processing unit (CPU) is an artificial intelligence image analysis system that controls an image capturing device to acquire a color image when acquiring an image for illuminance analysis and an image for object recognition.
A basic illumination register unit in which the basic illumination for adjusting the shutter, gain, and aperture level of the video recording device is stored according to the start, night, and adjustment mode, and the image captured by the video recording device is automatically controlled by artificial intelligence. In the artificial intelligence image analysis method of a system including a central processing unit (CPU) that is controlled to analyze according to,
(a) before acquiring an image for object recognition, obtaining an image for illuminance analysis of first obtaining an image for illuminance analysis by photographing a vehicle or an object;
(b) analyzing the image for illuminance analysis according to artificial intelligence automatic illuminance control, and selecting a daytime illuminance rac tangle region or a night illuminance rac tangle region according to the entry direction of the vehicle;
(c) accurately finding the position of the license plate of the vehicle through artificial intelligence automatic illumination control that obtains the illumination value for each area in the day and night rack tangle areas;
(d) Through the rack tangle area that divides day and night, artificial intelligence automatic illumination control applies the start register values (shutter, gain, experser) to measure various modes (night, dawn, morning, proper, daytime, etc.) Determining which mode of the day belongs to the adjustment, constant illumination time mode, etc.);
(e) In various modes (night, dawn, morning, proper, daytime, adjustment, constant illumination time mode, etc.), the zone range analysis unit for each mode finds the zone illumination range precisely subdivided for each mode through artificial intelligence automatic illumination control. ;
(f) With the illuminance found in various rack tangles, it is possible to meet all the directions in which the illuminance is moving through artificial intelligence automatic illuminance adjustment to reach the threshold by finding the register value (shutter, gain, exper) for each zone. Intelligent control, light sensitivity adjustment step;
(g) forcibly changing the register value of the threshold illuminance value and then applying the image for number analysis to the image capturing device by the central processing unit (CPU);
(h) obtaining, by an imaging device, an image for object recognition by photographing a vehicle or an object based on a threshold illuminance value;
(i) Step of allowing the central processing unit (CPU) to recognize the vehicle number based on the object recognition image or to analyze the resulting illuminance value, and automatically distinguish whether or not to shoot the next illuminance analysis image with the result value. ;
(j) Automatic illumination adjustment for each subdivided area of the entire screen so that an image capable of realizing autonomous driving of the vehicle can be captured.

Images