JPH07210799A - Measuring instrument for parking condition - Google Patents

Abstract

PURPOSE:To obtain a high hit rate of parking condition measurement even on holidays or in specific time zones wherein the hit rate decreases. CONSTITUTION:The parking condition measuring instrument which measures the parking condition of the whole parking lot from the parking condition of a representative area in a specific block of the parking lot by using a technique of a neural network is equipped with a passage rate measuring instrument 21 which measures passing vehicles in an entry or exiting path, a parking rate measuring instrument 22 which measures parked vehicles in the representative area, an information processing part 24 consisting of plural neural networks which input the signals from the passage rate measuring instrument and parking rate measuring instrument and perform arithmetic processing, a switching control part 23 which switches the neural networks according to the parking tendency of large and small cars, and a parking condition decision part 25 which decides the parking condition from the signal from the information processing part.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a parking condition measuring device for determining the parking condition of a parking lot such as a parking area or a service area of a highway where a large number of small vehicles and small size vehicles are mixed.

[0002]

2. Description of the Related Art As a method for measuring the parking situation in the above-mentioned parking lot, Japanese Patent Application Laid-Open No. 4-158 filed by the present applicant.
The number of parking lots in the entrance / exit of a parking lot on a highway and in the representative area of a specific block, which is proposed in Japanese Patent Publication No. 00, is counted as one for both large vehicles and small vehicles, and is input to the input layer of the information processing unit using a neural network method. In addition, there is a method to measure the parking situation.

A neural network constructed by using neurons as nodes will be described below. That is, it is composed of a node that performs a non-linear process on a plurality of inputs and obtains a plurality of outputs, and at least one or more inputs in each node are weighted and added to be converted into the input of the node itself. Then, it has a function to generate an output for propagating to another node for this converted input, and it is possible to control the weights so that the final output result from these nodes becomes a desired output. By using the information processing unit described above, the parking status of the entire parking lot can be measured with high accuracy from the parking status of the representative block.

[0004]

By the way, since the learning by the neural network is based on the average value of the measured data for several days, it is impossible to absorb the change in the parking situation of the large vehicle and the small vehicle on weekdays and holidays. However, there was a problem that the correct answer rate of parking situation measurement was lowered.

The main reason for lowering the rate of correct answer is that the tendency of large vehicles and small vehicles to park greatly differs depending on weekdays and holidays. Especially, on holidays with many small vehicles, two small vehicles park in the parking space of the large vehicle. Therefore, the correlation between the passing rate and the parking rate is broken.

Even on weekdays, the number of small cars is larger than that of large cars between 11:00 and 19:00, and the parking space for large cars is full due to a nap truck between 22:00 and 4:00. Occupying the space causes the problem that the correlation between the passing rate and the parking rate is broken as in the above, and the accuracy rate of the parking situation measurement is lowered.

The present invention is intended to solve the above-mentioned problems, and it is an object of the present invention to obtain a high correct answer rate even on holidays and a specific time period when the correct answer rate of parking situation measurement is lowered. There is a parking condition measuring device.

[0008]

SUMMARY OF THE INVENTION The parking condition measuring device of the present invention is intended to achieve the above-mentioned object, and the means is a parking condition in a representative area of a specific block of a parking lot using a neural network method. In the parking condition measuring device for measuring the parking condition of the entire parking lot from, a passage rate measuring device for measuring a passing vehicle in or out,
A parking rate measuring device that measures a parked vehicle in the representative area, an information processing unit including a plurality of neural networks that input signals of the passing rate measuring device and the parking rate measuring device and perform arithmetic processing, and the plurality of neural networks. It comprises a switching control unit for switching the network according to the parking tendency of a large-sized vehicle and a small-sized vehicle, and a parking state determination unit for determining the parking state based on a signal from the information processing unit.

Further, the switching control section may be composed of a perpetual program capable of setting weekdays, holidays and time zones, or discriminating between a large vehicle and a small vehicle by the passage rate measuring device. It is also possible to obtain the ratio of the number of passing large-sized vehicles and small-sized vehicles from the signal obtained by measuring the passing vehicles and switch the neural network by this output.

[0010]

The parking condition measuring device of the present invention constructed as described above measures the passing rate of vehicles passing through an entrance or an exit of a parking lot per unit time and parks in a representative area of a parking block. The parking rate of the vehicle is measured, and the measurement result is subjected to arithmetic processing in an information processing unit formed of a neural network, and is processed on weekdays, holidays, and / or
The neural network is switched by the switching control unit according to the time zone to display the parking situation.

Further, the passing rate of vehicles passing through the entrance or the exit of the parking lot per unit time is measured, and the parking rate of the vehicles parked in the representative area of the specific block is also measured. The information processing unit including a neural network performs arithmetic processing, and the switching control unit switches the neural network according to the ratio of passing large vehicles and small vehicles to display the parking status.

[0012]

[First Embodiment] Hereinafter, a first parking condition measuring device according to the present invention will be described.
The embodiment will be described with reference to FIGS. FIG. 1 is a plan view of the entire parking lot, and FIG. 2 is a circuit block diagram of the parking condition measuring device of the present invention. In FIG. 1, 10 is a parking lot,
A parking space 11 for large vehicles and a parking space 12 for small vehicles are arranged.

Numeral 21 indicates television cameras 211 and 212 installed on the entrance and exit of the parking lot 10 and the maximum number of vehicles per unit time (not shown) per unit time (the number of measured values multiplied by a safety factor). A passing rate measuring device including a calculation device for calculating a ratio of the number of vehicles (passing rate), 22 is a television camera 22 installed above an illumination column or the like in a parking lot.
1 and 222, and a parking rate measuring device including a computing device that calculates a parking rate per unit time in a representative area (not shown).

Since a method of detecting a passing vehicle and a parked vehicle by using image processing is a well-known technique, description thereof will be omitted. Further, the method of detecting the vehicle is not limited to the method based on the image processing, and a known vehicle detector may be used.

Reference numeral 23 is a multiplexer for switching the connection of a neural network, which will be described later, in accordance with an execution procedure set in advance for a perpetual program (the date, day of the week, time, etc. can be set and the program is executed based on the setting). The switching control unit 24, which is composed of neural networks 241-244, is an information processing unit that is switched by the switching control unit 23. The neural networks 241 to 244 are composed of a three-layer neural network including an input layer A, an intermediate layer B, and an output layer C shown in FIG.

Next, the structure of the neural network will be described. The input layer A has inputs I _{1 to} I ₄ , that is,
A plurality of neurons i _{1 to} i ₄ corresponding to the respective inputs I of the passing rate (entry rate, exit rate) and the parking rate of the representative area, and the intermediate layer B outputs with _four neurons j _{1 to} j ₄ . The layer C is composed of _three neurons k _{1 to} k 3 corresponding to the outputs O _{1 to} O ₃ . The number of neurons in the intermediate layer B is 4
The number is not limited to one, and may be any number required to obtain the required accuracy rate.

Each neuron i _{1 to} i ₄ of the input layer A transmits an external input to each neuron j _{1 to} j ₄ of the intermediate layer B. Wij is a weighting coefficient that indicates the degree of connection between any neuron i in the input layer A and any neuron j in the intermediate layer B.

Each of the neurons j _{1 to} j ₄ of the intermediate layer B nonlinearly transforms the input from the input layer A and transmits it to the output layer C.
Wjk is a weighting coefficient that indicates the degree of connection between any neuron j in the intermediate layer B and any neuron k in the output layer C.

D is a learning loop, which corrects the values of the weight Wij between the input layer A and the intermediate layer B and the weight Wjk between the intermediate layer B and the output layer C by the backpropagation method. T
_1, T ₂ and T ₃ are each correct answer events determined by actual measurement, respectively at the output O _1, O _2, and each time zone on weekdays and weekends to O ₃ output layer C.

With the above configuration, by performing weighted addition of each layer A to C, the function is converted into the input of the neuron itself, and an output for propagating to this neuron is generated for this input, and It becomes possible to control the weights so that the final output result generated from the neuron becomes a desired value.

Numeral 25 is provided with indicators 251, 252 and 253 for numerically displaying the signals of the outputs O ₁ , O ₂ and O ₃ of the output layer C, and discriminating the output with the maximum numerical value, the discrimination result is "full". This is a parking condition determination unit that outputs a signal of "crowded" or "vacant".

The outputs O ₁ , O ₂ and O ₃ of the output layer C have a parking rate of, for example, 80% or more of the maximum parking rate, 79% to 41%, and 40% or less of 1 respectively. It outputs as a signal from 0.00 to 1.00 for each correct answer event. Reference numeral 3 is a large-scale display device which is installed in front of the entrance of the parking lot 10 and displays the characters "full", "crowded" or "empty" in response to a signal from the parking condition judging section 25.

Next, the operation will be described separately for the "learning operation" and the "parking condition measuring operation" based on the above-mentioned configuration. “Learning operation” When performing the learning operation, the passing number and the parking number in each day and time zone shown in FIG. 2 are measured, and the passing rate from the passing rate measuring device 21 per unit time is measured in each neural network. , The parking rate from the parking rate measuring device 22 is inputted and learned, and the learned neural network measures the parking situation.

In this embodiment, the correct event T ₁ is input to O ₁ which outputs the full vehicle status, the correct event T ₂ is input to O ₂ which outputs the congestion status, and the correct event T ₂ is output to O ₃ which outputs the empty status. Is the correct answer event T ₃ . That is, the correct answer event when the vehicle is full is [T ₁ , T ₂ , T ₃ ]: [1, 0, 0], that is, when the vehicle is crowded [T ₁ , T ₂ , T ₃ ]: [0, ₁ , 0], When there is an empty space, [T ₁ , T ₂ , T ₃ ]: [0, 0, 1].

Next, learning is performed using the back propagation method, and this learning is repeated many times to output O.
The values of the weighting factors Wij and Wjk between the input layer A, the intermediate layer B, and the output layer C are modified so that the error between the correct answer event T and the correct answer event T is minimized. Note that the method is not limited to the back propagation method, and many methods are known for neural networks, and any of these methods may be used.

[Parking status measurement operation] The measured number of passing vehicles per unit time on each day of the week and time zone and the number of parking vehicles in the representative area are counted and added to the neural network for which the learning operation has finished, and the maximum passing vehicle number is obtained. The parking rate of the entire parking lot is calculated by inputting the parking rate obtained from the ratio of the passing rate and the number of parking lots, and the output O ₁ , O ₂ or O ₃ is obtained, and the vehicle is full or crowded. The status and the vacant status are output, the parking status determination unit 25 determines the largest output value, and the corresponding “full” or “crowded”
Alternatively, it outputs a signal of "empty". Then, this output is sent to the display device 3 installed in front of the entrance of the parking lot 10 to display "full", "crowded" or "empty".

In the above-described operation, one large vehicle is counted as "2" and one small vehicle is counted as "1" to obtain the passing rate and the parking rate, thereby further increasing the accuracy rate of the parking condition measurement. be able to.

Next, another embodiment according to the present invention will be described with reference to FIG. In the above-mentioned embodiment, the switching of the neural network in the switching control unit 23 is explained by the perpetual program, but in the present embodiment, it is switched by the output from the passage rate measuring device 21. It is the one.

That is, in the passage rate measuring device 21 of this embodiment, a position sensor for detecting the front end of a passing vehicle, a television camera for picking up an image of a license plate by a detection signal of the position sensor, and the television camera are used. It consists of an image processing device that processes the captured video, detects the size of the license plate, adds a large car as "2" and a small car as "1", and outputs the passing rate per unit time It is a thing.

In the case of this embodiment, the number of passing large vehicles and small vehicles from the passing rate measuring device 21 is compared, and the input to the information processing unit 24 is switched based on the comparison result. By changing over in the same manner and performing the same control as in the above-described embodiment, it is possible to increase the accuracy rate of the parking condition measurement.

In the above embodiment, a method of detecting the size of the license plate is illustrated as a method of distinguishing a large vehicle from a small vehicle. However, a method of determining the size of the photographed vehicle and a vehicle detector such as a loop coil. It is possible to use any vehicle detection method according to the installation situation, such as the method using. Further, the switching control unit 23 may switch the neural network by a combination of the perpetual program and the determination of the number of passing vehicles which are the two embodiments described above.

Furthermore, in FIG. 4, a method using three types of neural networks is illustrated, but by further subdividing the number of small vehicles and large vehicles, similar to the method by the perpetual program, more accurate parking state determination can be performed. It is possible. In the above-described embodiment, the passage rate measuring device is provided on the entrance and exit of the parking lot. However, an approximate effect can be obtained by providing the passage ratio measuring device on either the entrance or the exit.

[0033]

As described above, according to the present invention, the passing rate of vehicles passing through the entrance or the exit of the parking lot, the parking rate of vehicles parked in the representative area, the weekday, the holiday and the time zone are set. Since the parking situation is determined by using a neural network method as data that distinguishes between and, it is possible to check whether there are many small cars on Saturdays, holidays and daytime, and many large cars on late nights on weekdays. It is possible to determine the parking situation with high accuracy, and by having a distinction between a large vehicle and a small vehicle in the measurement of the passage rate and the parking rate, it is possible to perform a measurement with a higher accuracy rate. Is.

[Brief description of drawings]

FIG. 1 is a plan view of an entire parking lot for explaining an embodiment of a parking condition measuring device according to the present invention.

FIG. 2 is a circuit block diagram showing a configuration of a parking condition measuring device.

FIG. 3 is an explanatory diagram showing a configuration of a neural network.

FIG. 4 is an explanatory diagram showing a configuration of another embodiment.

[Explanation of symbols]

 10 Parking Lot 21 Passage Rate Measuring Device 22 Parking Rate Measuring Device 23, Switching Control Section 24 Information Processing Section 25 Parking Status Discriminating Section 3 Display Device

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] March 1, 1994

[Procedure Amendment 1]

[Document name to be amended] Statement

[Name of item to be corrected] 0003

[Correction method] Change

[Correction content]

That is, it is composed of a node which performs a non-linear process on a plurality of inputs and obtains a plurality of outputs. By weighting and adding at least one or more inputs in each node, the node itself It has the function of converting to an input and producing an output for propagating to another node for this converted input, and the final output result from these nodes becomes a desired output,
By using the information processing unit capable of controlling the weight, the parking condition of the entire parking lot can be measured with high accuracy from the parking condition of the representative area. Hereinafter, description will be made by using a neural network configured by using neurons as nodes.

Claims (3)

[Claims] 1. A parking condition measuring device for measuring a parking condition of an entire parking lot from a parking condition in a representative area of a specific block of a parking lot by using a neural network method. A rate measurement device, a parking rate measurement device that measures a parked vehicle in the representative area, and an information processing unit including a plurality of neural networks that inputs signals of the passage rate measurement device and the parking rate measurement device and performs arithmetic processing. A switching control unit that switches the plurality of neural networks according to the parking tendency of a large vehicle and a small vehicle, and a parking state determination unit that determines the parking state based on a signal from the information processing unit. Parking condition measuring device. 2. The parking condition measuring device according to claim 1, wherein the switching control unit is composed of a perpetual program capable of setting weekdays, holidays and time zones. 3. The switching control unit determines the ratio of the number of passing large vehicles and small vehicles from the signal obtained by measuring the passing vehicles by distinguishing between the large vehicle and the small vehicle by the passage rate measuring device, and using the output. The parking condition measuring device according to claim 1, wherein the neural network is switched.