CN112927455B

CN112927455B - Intelligent monitoring method for parking lot and application

Info

Publication number: CN112927455B
Application number: CN202110099236.5A
Authority: CN
Inventors: 李子龙
Original assignee: Guangdong Hongyuan Xinke Automation Technology Development Co ltd
Current assignee: Guangdong Saide Automation Technology Co.,Ltd.
Priority date: 2021-01-25
Filing date: 2021-01-25
Publication date: 2022-03-04
Anticipated expiration: 2041-01-25
Also published as: CN112927455A

Abstract

The invention provides an intelligent monitoring method for a parking lot and application thereof, comprising the following steps: acquiring a real-time monitoring video image of a parking lot; detecting people and vehicles in the parking lot according to the real-time monitoring video image of the parking lot, and further detecting whether the parking space where the relevant vehicle is located is invaded; detecting a vehicle leaving from a parking space by tracking a moving target in the parking lot based on a real-time monitoring video image of the parking lot; and sending alarm information if the vehicle leaves the parking space and drives away from the invaded parking space. Compared with the prior art, the system can play a role in safety monitoring, can prevent vehicles from being invaded and then failing to check in time when leaving from the parking lot, has high intelligent degree, and can greatly reduce the monitoring and management cost of large parking lots.

Description

Intelligent monitoring method for parking lot and application

Technical Field

The invention relates to the technical field of intelligent monitoring, in particular to an intelligent monitoring method for a parking lot and application of the intelligent monitoring method.

Background

With the development of video monitoring technology, the existing monitoring system is applied to parking places, such as factories, markets, schools, hospitals and the like, and is mainly of two types aiming at the existing monitoring device or system of the parking lot, wherein the first type is to set a barrier gate and a camera at an entrance and count passing vehicles, so that the parking information in the parking lot is known, and the parking information is transmitted to a rear-end device. The second is that the inside camera or other sensors that set up in parking area to the parking condition in every parking stall is known to the accuracy, transmits parking information to rear end device at last. The parking information can only include vacant parking space information or non-vacant parking space information, and can also include both the vacant parking space information and the non-vacant parking space information, and the rear-end device can be a monitoring screen in a monitoring room or a display screen at an entrance of a parking lot.

That is, the parking lot monitoring technology in the prior art is mainly used for parking charging, vehicle access identification, and knowing the vacant parking space. For the safety problem possibly existing in the parking lot, for example, related vehicles on the parking spaces can be invaded, monitoring facilities in the parking lot only have a monitoring camera shooting function, camera shooting data are recorded in real time through camera shooting monitoring of fixed positions, so that the scene can be recovered through troubleshooting on-site monitoring after problems occur, but the monitoring device can only perform later-stage troubleshooting through images, cannot avoid the situation that the automobile is stolen in time, and cannot avoid property loss.

In order to solve the problem, in the prior art, the face of a driver is photographed and then compared with a prestored picture on a vehicle driving certificate to check whether the vehicle is invaded. However, the technology has many defects, such as that the process of recording and storing the driving license is complicated, and especially the technology is more complicated for occasions with many vehicles and frequent change of personnel. And because the difference between the owner's picture on many driver licenses and the actual image of the owner is large, the definition of the driver real-time image obtained by remote shooting through the camera is not high sometimes, and the like, the accuracy of the monitoring system is general.

Disclosure of Invention

In order to overcome the defects of the prior art, the invention provides an intelligent parking lot monitoring method and application, which are used for solving at least one of the technical problems, and the specific technical scheme is as follows:

an intelligent monitoring method for a parking lot comprises the following steps:

acquiring a real-time monitoring video image of a parking lot;

detecting people and vehicles in the parking lot according to the real-time monitoring video image of the parking lot, and further detecting whether parking spaces of related vehicles are invaded;

detecting a vehicle leaving from a parking space by tracking a moving target in the parking lot based on a real-time monitoring video image of the parking lot;

and sending alarm information if the vehicle leaves the parking space and drives away from the invaded parking space.

In a specific embodiment, the method for acquiring the real-time monitoring video image of the parking lot comprises the following steps:

the method comprises the steps that cameras are arranged at multiple positions of a parking lot respectively, and images of the parking lot are collected continuously according to preset time intervals to obtain multiple frames of continuous video images;

and tracking and matching moving targets in images acquired by different cameras.

In one specific embodiment, the method for detecting whether the parking space of the relevant vehicle is invaded or not according to the real-time monitoring video image of the parking lot comprises the following steps:

identifying people in the parking lot through the multi-frame continuous video images, and tracking and matching the motion trail of the people;

detecting whether a person enters a parking space of a related vehicle;

for the parking spaces where people enter the relevant vehicles, in each frame of image, if the proportion of the cross area between the people and the relevant parking spaces divided by the area of the people is larger than a preset proportion b1, judging that the frame of image has an intrusion footprint;

counting the total frame number of the images with the intrusion footprints, calculating the duration time of the intrusion of the related parking spaces by dividing the total frame number with the intrusion footprints by the frame rate of the multi-frame continuous video images, and judging that the related parking spaces are intruded if the duration time exceeds preset time t 1;

preferably, the value range of the preset proportion b1 is 45% -55%, and further preferably, the value range of the preset proportion b1 is 50%;

preferably, the preset time t1 ranges from 4 seconds to 7 seconds, and further preferably, the preset time t1 is 5 seconds.

In one particular embodiment, a method of tracking a moving object within a parking lot includes:

identifying all moving objects in the current frame image: locking all moving targets in the current frame image by comparing the difference between the previous preset number of X frame images and the current frame image;

tracking and classifying the same moving target, setting all the moving targets determined before the current frame image as known targets, setting all the moving targets in the current frame image as unclassified targets, tracking and matching the known targets with the unclassified targets one by one, and storing the historical moving track of each moving target:

s01: detecting the area S1 of a cross region between the previous frame image and the current frame image of an unclassified object of a known object and the area S of the unclassified object in the current frame image, and identifying whether the unclassified object meets the condition that the result of dividing the area S1 by the area S is larger than a preset proportion b 2;

s02: if there are a plurality of unclassified objects satisfying step S01 in the current frame,

s02_ 01: calculating the motion direction d1 of the known target and the motion direction d2 of the unclassified target one by one according to the historical motion trail, calculating the intersection angle between the motion direction d1 and the motion direction d2, when a preset number N of unclassified targets exist, each known target and each unclassified target have N intersection angles, and when the maximum value of the N intersection angles is larger than a preset angle, tracking the classified target according to the minimum intersection angle;

s02_ 02: if the step S02_01 fails to be executed, calculating the distances between the previous frame image and the non-classified targets of the known targets in the current frame image one by one, when there are K non-classified targets in a preset number, each of the known targets and the non-classified targets has K distances, calculating the maximum distance and the minimum distance of the K distances, and if the ratio obtained by dividing the difference value between the maximum distance and the minimum distance by the maximum distance is greater than a third preset ratio, tracking the classified targets according to the minimum distance;

s02_ 03: if the step S02_02 fails to be executed, calculating the area increment of the known target in the previous frame image and the unclassified target in the current frame image one by one, and tracking the classified target according to the minimum area increment of the target between the previous frame image and the current frame image;

preferably, the value range of the preset proportion b2 is 30% -50%, and further preferably, the preset proportion b2 is 40%;

preferably, the value range of the preset angle is 15-21 degrees, and further preferably, the preset angle is 18 degrees;

preferably, the value range of the third preset proportion is 15% -25%, and further preferably, the third preset proportion is 20%.

In one particular embodiment, a method of detecting a vehicle in a parking lot includes:

identifying a potential target existing in the multi-frame image for a time exceeding a preset time t2, and detecting a vehicle from the potential target by the following method L or method M:

the method L comprises the following steps: by a deep learning algorithm, comprising:

l1: identifying that the potential target is a vehicle through a deep learning algorithm;

l2: the ratio of the area of the potential target to the area of the parking space is larger than a preset proportion b 3;

l3: in the single frame image, if the potential target satisfies the step L1 and the step L2 at the same time, it is preliminarily determined that the potential target is a vehicle;

l4: continuously detecting the multi-frame images according to the method of the step L3, and outputting potential targets, which are continuously vehicles in the multi-frame images and have the time exceeding the preset time t3, as finally detected vehicles;

method M: by a size detection method, comprising:

m1: the aspect ratio of the potential target is larger than the preset proportion b 4;

m2: the ratio of the area of the potential target to the area of the parking space is larger than a preset proportion b 3;

m3: the area of the potential target > < ═ the maximum area of the vehicle, wherein, the minimum area of the vehicle is calculated according to the minimum size of all parking spaces, and the maximum area of the vehicle is calculated according to the maximum size of all parking spaces;

m4: in a single frame image, if the potential target simultaneously meets the steps M1, M2 and M3, preliminarily judging that the potential target is like a vehicle;

m5: continuously detecting the multi-frame images according to the method of the step M4, and outputting a potential target as a finally detected vehicle, wherein the ratio of the time of continuously imaging the vehicle in the multi-frame images to the preset time t2 exceeds the preset proportion b 5;

preferably, the preset time t2 is 3 seconds;

preferably, the preset ratio b3 is 90%;

preferably, the preset time t3 is 0.8 seconds;

preferably, the preset proportion b4 is 80%;

preferably, the preset ratio b5 is 90%.

In a specific embodiment, the deep learning algorithm of method L is enhanced as follows:

a size detection method introduced into the method M is combined with a deep learning algorithm of the method L, and a vehicle is detected from a potential target;

searching front, back, left and right respectively relative to the potential target for processing possible displacement of the vehicle;

and automatically saving the confused image in the deep learning algorithm of the method L to expand the deep learning database.

In one particular embodiment, a method of detecting a vehicle exiting a parking space includes:

n1: detecting a vehicle in a parking lot;

n2: the current position of the vehicle is further from the parking space than its birth position;

n3: the current position of the vehicle is not overlapped with the parking space;

n4: the vehicle is in the parking lot beyond a preset proportion b5 in the historical motion footprint at a previous preset time t4, preferably, the preset time t4 is 3 seconds, and the preset proportion b5 is 60%;

n5: if the vehicle satisfies steps N1-N4, it is determined that the corresponding vehicle is a vehicle that departs from the parking space.

In a specific embodiment, the method further comprises sending out prompt information for the vehicle leaving the parking space if the vehicle is not the vehicle leaving the invaded parking space.

An intelligent parking lot monitoring system for executing the intelligent parking lot monitoring method according to any one of the preceding embodiments, comprising:

the real-time monitoring video image acquisition module is used for acquiring real-time monitoring video images of the parking lot;

the personnel detection module is used for detecting the personnel in the parking lot according to the real-time monitoring video image of the parking lot;

the vehicle detection module is used for detecting vehicles in the parking lot according to the real-time monitoring video images of the parking lot;

the parking space detection module is used for detecting whether the parking space of the related vehicle is invaded;

the vehicle leaving detection module is used for monitoring a video image in real time based on the parking lot and detecting a vehicle leaving from the parking space by tracking a moving target of the parking lot;

and the alarm module is used for sending alarm information to the vehicle leaving from the parking space if the vehicle leaves from the invaded parking space.

A computer device, comprising:

one or more processors;

a memory for storing one or more programs;

when the one or more programs are executed by the one or more processors, the one or more processors implement the parking lot intelligent monitoring method according to any one of the foregoing embodiments.

The invention has at least the following beneficial effects:

according to the intelligent monitoring method and the application of the parking lot, people and vehicles in the parking lot are detected according to the real-time monitoring video images of the parking lot, whether parking spaces of related vehicles are invaded or not is further detected, vehicles leaving from the parking spaces are detected by tracking moving targets in the parking lot based on the real-time monitoring video images of the parking lot, and alarm information is sent out if the vehicles leaving from the parking spaces are vehicles leaving from the invaded parking spaces. Therefore, whether the vehicle is invaded or not is judged by detecting suspicious behaviors of the user through machine vision, whether the vehicle leaves from the parking space or not is further judged by combining with moving target identification, and then the separated invaded vehicle is alarmed. Compared with the prior art, the intelligent monitoring method and the application of the parking lot can play a role in safety monitoring, prevent vehicles from being incapable of being checked in time when the vehicles leave the parking lot after being invaded, have high intelligent degree, can greatly reduce the monitoring and management cost of a large parking lot, and do not need an operator to frequently input data.

In order to make the aforementioned and other objects, features and advantages of the present invention comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.

Fig. 1 is a general schematic diagram of an intelligent parking lot monitoring method according to embodiment 1;

fig. 2 is a schematic diagram of a method of detecting whether a parking space of an associated vehicle is invaded in embodiment 1;

fig. 3 is a schematic view of a method of tracking a moving object of a parking lot in embodiment 1;

FIG. 4 is a schematic diagram of method L in example 1;

FIG. 5 is a schematic diagram of method M in example 1.

Detailed Description

Example 1

As shown in fig. 1 to 5, the present embodiment provides an intelligent parking lot monitoring method, including:

s1: and acquiring a real-time monitoring video image of the parking lot. Specifically, cameras can be respectively arranged at one or more positions of the parking lot, and images of the parking lot are continuously acquired according to a preset time interval to obtain multi-frame images, namely real-time monitoring video images. The cameras can be respectively arranged at different positions of the parking lot and continuously acquire images of the parking lot to obtain multi-frame images so as to monitor the parking lot in real time, wherein the multi-frame images are formed by a plurality of single-frame images, a picture of the parking lot shot by each single-frame image is a static image, and when the single-frame images are continuously shot, the positions of a moving target on the frames of images are different, so that the visual tracking of the moving target can be realized.

In the case of multiple cameras, the orientations of the multiple cameras can be calibrated in advance. When the plurality of cameras respectively take pictures, matching the targets in the images acquired by different cameras, wherein the matching comprises the matching of a static target and the matching of a moving target. Specifically, although the positions of the cameras at different positions and the pictures obtained by inconsistent shooting visual angles are different, the targets in the real-time monitoring video images acquired by the different cameras can be matched through the calibrated positions and directions of the cameras.

S2: and detecting people and vehicles in the parking lot according to the real-time monitoring video image of the parking lot, and further detecting whether the parking spaces of the related vehicles are invaded.

The method for detecting the vehicles on the parking lot comprises the following steps:

the potential target existing in the multi-frame image for a time exceeding the preset time t2 is identified, wherein the preset time t2 is preferably 3 seconds, and may be other values around 3 seconds such as 2.9 seconds, 3.1 seconds, and the like, and compared with the prior art that the potential target is identified directly from a single-frame or multi-frame image, the present embodiment identifies the potential target existing in the multi-frame image for a time exceeding the preset time t2, thereby effectively eliminating noise and achieving higher identification accuracy of the potential target. And detecting a vehicle from the potential targets by the following method L or method M:

l1: and identifying that the potential target is the vehicle through a deep learning algorithm. The deep learning algorithm is used for learning the internal rules and the expression levels of sample data, the information obtained in the learning process is greatly helpful for explaining data such as images, and the final aim of the deep learning algorithm is to enable a machine to have the analysis and learning capacity like a human and to be capable of identifying images. Specifically, in this embodiment, training data may be generated by using images of a large number of vehicles, the training data may be trained to obtain a vehicle prediction model, and then, whether a potential vehicle is a vehicle may be identified by inputting the acquired potential target image into the vehicle prediction model. Illustratively, the present embodiment provides three types of deep learning models: a neural network system based on convolution operation, namely a convolution neural network; self-coding neural networks based on multi-layer neurons, including both self-coding and sparse coding, which has received much attention in recent years; and pre-training in a multilayer self-coding neural network mode, and further optimizing the deep confidence network of the neural network weight by combining the identification information. The above deep learning models can be used alone or in combination to improve the accuracy of the deep learning algorithm, and for the specific execution methods of the above deep learning models, reference may be made to the prior art, and details are not repeated in this example.

L2: the ratio of the area of the potential target to the area of the parking space is larger than the preset ratio b 3. The area of the potential target and the area of the parking space can be obtained by converting the size (such as pixels) on the image. Preferably, the preset ratio b3 is 90%, and may be other values around 90% such as 89%, 91%, etc.

L3: in the single-frame image, if the potential target satisfies both step L1 and step L2, it is preliminarily determined that the potential target is a vehicle. Compared with a pure deep learning algorithm, the deep learning algorithm is combined with the area ratio comparison to comprehensively judge whether the potential target is the vehicle, so that the vehicle identification accuracy is higher.

L4: the multi-frame images are successively detected in accordance with the method of step L3, and potential targets that are successively vehicles in the multi-frame images for a time exceeding the preset time t3 are output as finally detected vehicles.

Wherein the preset time t3 is 0.8 seconds. Therefore, on one hand, the potential target with the time of continuously being the vehicle in the multi-frame images exceeding the preset time t3 is output as the finally detected vehicle, so that the noise can be effectively eliminated, and the identification accuracy of the potential target is higher. On the other hand, because the preset time t3 is less than the preset time t2, the situation that the real vehicles possibly generated due to the reasons of excessive image definition, excessive image processing analysis and the like are mistakenly rejected because the time requirement is too strict can be avoided, and the balance of improving the detection accuracy and preventing missing of the vehicles is realized.

Method M: by a size detection method, comprising:

m1: the aspect ratio of the potential target is greater than the preset ratio b 4. However, as will be understood by those skilled in the art, the length of the potential target and the width of the potential target are not in a fixed direction, but in the image, the side with the larger length of the potential target is the length direction, and the side with the smaller length of the potential target is the width direction. The predetermined ratio b4 is preferably 80%, and may be other values around 80%, such as 79%, 81%, and the like.

M2: the ratio of the area of the potential target to the area of the parking space is larger than the preset ratio b 3. Preferably, the preset ratio b3 is 90%, and may be other values around 90% such as 89%, 91%, etc.

M3: the area of the potential target > < ═ the area of the potential target, either one of the minimum area of the vehicle and the area of the potential target, wherein the minimum area of the vehicle is calculated according to the minimum size of all parking spaces, and the maximum area of the vehicle is calculated according to the maximum size of all parking spaces.

M4: in the single frame image, if the potential target simultaneously satisfies the step M1, the step M2 and the step M3, the potential target is preliminarily determined to be like a vehicle. Therefore, the embodiment creatively provides a specific method for detecting the vehicle through the size, namely, the relationship among the aspect ratio of the potential target, the area ratio of the potential target to the parking space, the area of the potential target, the minimum area of the vehicle and the maximum area of the vehicle is provided.

M5: and continuously detecting the multi-frame images according to the method of the step M4, and outputting the potential target of which the ratio of the time of continuously imaging the vehicle in the multi-frame images to the preset time t2 exceeds the preset proportion b5 as the finally detected vehicle. Preferably, the preset ratio b5 is 90%, and may be other values around 90% such as 89%, 91%, etc. The embodiment outputs the potential target with the ratio of the time of continuously imaging the vehicle in the multi-frame image to the preset time t2 exceeding the preset proportion b5 as the finally detected vehicle, so that the noise can be effectively eliminated, and the identification accuracy of the potential target is higher.

Preferably, the deep learning algorithm of method L is enhanced as follows:

(1) the size detection method introduced into the method M is combined with the deep learning algorithm of the method L, and the vehicle is detected from the potential target. By combining the deep learning algorithm and the size detection method, for example, comparing the results of the two methods, if the detected target is determined to be the vehicle, the vehicle is finally identified, and if the results of the two methods are inconsistent, the image can be used as a confused image or output to be manually corrected, so that the accuracy redundancy of the detection method can be further improved.

(2) And searching forwards, backwards, leftwards and rightwards respectively relative to the potential target for processing possible displacement of the vehicle. For example, an X direction and a Y direction perpendicular to each other may be defined, a square detection area that frames a potential target is set, and the target recognition is performed by moving the square detection area in the X direction or the Y direction, so that a possible displacement of the vehicle when the vehicle is recognized may be handled.

(3) The confusing image in the deep learning algorithm of the method L is automatically saved to expand the deep learning database.

In this example, the method for detecting people in the parking lot according to the real-time monitoring video image of the parking lot comprises one or more of deep learning algorithm detection, size detection and human shape detection. For example, for deep learning algorithm detection, a deep learning algorithm in vehicle detection can be referred to, sample data of a human is trained firstly, and then an object to be detected is input into a training model for prediction to identify whether the object is a human or not. For example, for size detection, it can be detected whether the human is by analyzing the length and width size values, ratios, and other characteristics of the human. For example, for human shape detection, whether a human is detected by detecting the shape of the human, such as the head, the trunk, the limbs, and the like, is not described in detail in this embodiment.

In this embodiment, the method for detecting whether the parking space of the relevant vehicle is invaded according to the real-time monitoring video image of the parking lot includes:

identifying people in the parking lot through multi-frame continuous video images, and tracking and matching the motion trail of the people;

detecting whether a person enters a parking space of a related vehicle;

counting the total frame number of the images with the intrusion footprints, calculating the duration time of the related parking spaces being invaded by dividing the total frame number with the intrusion footprints by the frame rate (namely the frame number transmitted per second) of the continuous video images of multiple frames, and judging that the related parking spaces are invaded if the duration time exceeds the preset time t 1;

preferably, the value range of the preset proportion b1 is 45% -55%, and further preferably, the preset proportion b1 is 50%;

preferably, the preset time t1 ranges from 4 seconds to 7 seconds, and further preferably, the preset time t1 is 5 seconds. Therefore, when a person stays in the parking space of the related vehicle for more than a certain time, the person can be automatically judged to be suspicious behavior, namely, the related parking space is judged to be invaded, if the vehicle subsequently leaves from the parking lot, the manager of the parking lot can be prompted to take measures for prevention by sending alarm information, for example, the vehicle is detected.

S3: the method comprises the steps of detecting vehicles leaving from parking spaces by tracking moving targets in the parking lots based on real-time monitoring video images of the parking lots.

S4: and sending alarm information if the vehicle leaves the parking space and drives away from the invaded parking space.

In this embodiment, a method of tracking a moving object (person or vehicle) of a parking lot includes:

the method for tracking the moving target in the parking lot comprises the following steps:

identifying all moving objects in the current frame image: locking all moving targets in the current frame image by comparing the difference between the previous preset number of X frame images and the current frame image, wherein X can be optimally set according to the moving speed of the moving targets and the frame rate of the real-time monitoring video image, and exemplarily X is 25 frames;

s01: detecting the area S1 of a cross region between the previous frame image and the current frame image of the known target and the area S of the un-classified target in the current frame image, and identifying whether the un-classified target meets the condition that the result of dividing the area S1 by the area S is larger than a preset proportion b2, namely, an overlapping region must exist between the previous frame image and the current frame image of the same moving target and the overlapping region needs to be larger than a preset threshold value;

s02_ 01: calculating the motion direction d1 of the known target and the motion direction d2 of the unclassified target one by one through the historical motion trajectory, calculating the intersection angle between the motion direction d1 and the motion direction d2, when a preset number N of unclassified targets exist, each known target and each unclassified target have N intersection angles, when the maximum value of the N intersection angles is larger than a preset angle, tracking the classified target according to the minimum intersection angle, namely when the motion directions of the unclassified targets have obvious difference, preferentially selecting the motion direction for classification, classifying the unclassified target with the minimum intersection angle as the same motion target as the corresponding known target, and exemplarily, N is 5;

s02_ 02: if the step S02_01 fails to be executed, calculating the distances between the previous frame image and the non-classified targets of the known targets in the current frame image one by one, when there are K non-classified targets in a preset number, each of the known targets and the non-classified targets has K distances, calculating the maximum distance and the minimum distance of the K distances, and if the ratio obtained by dividing the difference value between the maximum distance and the minimum distance by the maximum distance is greater than a third preset ratio, tracking the classified targets according to the minimum distance; illustratively, K is 5.

preferably, the value range of the preset proportion b2 is 30% -50%, and further preferably, the value range of the preset proportion b2 is 40%;

preferably, the third preset proportion is 15% -25%, and further preferably, the third preset proportion is 20%.

In the present embodiment, a method of detecting a vehicle leaving a parking space includes:

n1: detecting a vehicle in a parking lot;

n2: the current location of the vehicle is further from the parking space than its birth location (first historical footprint);

In this embodiment, it is preferable that the method further includes, for a vehicle leaving the parking space, issuing a prompt message if the vehicle is not a vehicle leaving the parking space that has been invaded.

According to the intelligent monitoring method for the parking lot, the people and the vehicles in the parking lot are detected according to the image of the parking lot, whether the parking spaces of the related vehicles are invaded or not is further detected, the vehicles leaving from the parking spaces are detected by tracking the moving targets of the parking lot based on the image of the parking lot, and alarm information is sent out if the vehicles leaving from the parking spaces are the vehicles driving away from the invaded parking spaces. Therefore, whether the vehicle is invaded or not is judged by detecting suspicious behaviors of the user through machine vision, and whether the vehicle is invaded or not is further judged by combining with moving target identification.

Example 2

In this embodiment, for the intelligent parking lot monitoring method provided in embodiment 1, as shown in fig. 4, an intelligent parking lot monitoring system is configured to execute the intelligent parking lot monitoring method in embodiment 1, and includes:

the image acquisition module is used for acquiring real-time monitoring video images of the parking lot;

and the alarm module is used for sending alarm information to the vehicle leaving from the parking space if the vehicle leaves from the invaded parking space. Wherein, the alarm information comprises one or more of character information, flashing alarm lamp and alarm bell.

Example 3

The embodiment provides a computer device, including:

a computer device, comprising:

one or more processors;

a memory for storing one or more programs;

when the one or more programs are executed by the one or more processors, the one or more processors implement the parking lot intelligent monitoring method according to any one of the foregoing technical solutions. The computer device may be embodied in the form of a general purpose computing device or may be configured as a robot. Components of the computer device may include, but are not limited to: one or more processors or processing units, a system memory, and a bus connecting various system components, such as the system memory and the processing units. The computer device typically includes a variety of computer system readable media. Such media may be any available media that is accessible by the device computer and includes both volatile and nonvolatile media, removable and non-removable media. The system memory may include computer system readable media in the form of volatile memory.

The computer device may also communicate with one or more external devices, such as a keyboard, pointing device, display, etc., and may also communicate with one or more devices that enable a user to interact with the computer device, and/or with any devices that enable the computer device to communicate with one or more other computing devices.

The processing unit executes various functional applications and data processing by running programs stored in the system memory.

Those skilled in the art will appreciate that the figures are merely schematic representations of one preferred implementation scenario and that the blocks or flow diagrams in the figures are not necessarily required to practice the present invention.

Those skilled in the art will appreciate that the modules in the devices in the implementation scenario may be distributed in the devices in the implementation scenario according to the description of the implementation scenario, or may be located in one or more devices different from the present implementation scenario with corresponding changes. The modules of the implementation scenario may be combined into one module, or may be further split into a plurality of sub-modules.

The above-mentioned invention numbers are merely for description and do not represent the merits of the implementation scenarios.

The above disclosure is only a few specific implementation scenarios of the present invention, however, the present invention is not limited thereto, and any variations that can be made by those skilled in the art are intended to fall within the scope of the present invention.

Claims

1. An intelligent monitoring method for a parking lot is characterized by comprising the following steps:

acquiring a real-time monitoring video image of a parking lot;

for the vehicle leaving from the parking space, if the vehicle leaves from the invaded parking space, alarm information is sent;

s02_ 03: if the step S02_02 fails to be executed, the area increments of the known target in the previous frame image and the unclassified target in the current frame image are calculated one by one, and the classified target is tracked according to the minimum area increment of the target between the previous frame image and the current frame image.

2. The intelligent monitoring method for the parking lot according to claim 1, wherein the method for acquiring the real-time monitoring video image of the parking lot comprises the following steps:

3. The intelligent monitoring method for the parking lot according to claim 2, wherein the method for detecting whether the parking space of the relevant vehicle is invaded according to the real-time monitoring video image of the parking lot comprises the following steps:

detecting whether a person enters a parking space of a related vehicle;

counting the total frame number of the images with the invasion footprints, calculating the duration time of the invasion of the related parking spaces by dividing the total frame number with the invasion footprints by the frame rate of the multi-frame continuous video images, and judging that the related parking spaces are invaded if the duration time exceeds preset time t 1.

4. The intelligent parking lot monitoring method according to claim 3,

the value range of the preset proportion b1 is 45-55%;

the preset time t1 ranges from 4 seconds to 7 seconds.

5. The intelligent parking lot monitoring method according to claim 4, wherein the preset proportion b1 is 50%; the preset time t1 is 5 seconds.

6. The intelligent parking lot monitoring method according to claim 1,

the value range of the preset proportion b2 is 30-50%;

the value range of the preset angle is 15-21 degrees;

the value range of the third preset proportion is 15% -25%.

7. The intelligent parking lot monitoring method according to claim 6, wherein the preset proportion b2 is 40%;

the preset angle is 18 degrees;

the third predetermined proportion is 20%.

8. The intelligent monitoring method for a parking lot according to claim 1, wherein the method for detecting vehicles in the parking lot comprises:

method M: by a size detection method, comprising:

m5: and continuously detecting the multi-frame images according to the method of the step M4, and outputting the potential target of which the ratio of the time of continuously imaging the vehicle in the multi-frame images to the preset time t2 exceeds the preset proportion b5 as the finally detected vehicle.

9. The intelligent parking lot monitoring method according to claim 8,

the preset time t2 is 3 seconds;

the preset proportion b3 is 90%;

the preset time t3 is 0.8 second;

the preset proportion b4 is 80%;

the preset ratio b5 is 90%.

10. The intelligent parking lot monitoring method according to claim 9, wherein the deep learning algorithm of method L is enhanced according to the following method:

11. The intelligent parking lot monitoring method according to claim 1, wherein the method of detecting a vehicle leaving the parking space comprises:

n1: detecting a vehicle in a parking lot;

n4: the vehicle with the historical motion footprint exceeding the preset proportion b5 in the previous 3 seconds is in the parking lot, and preferably the preset proportion b5 is 60%;

12. The intelligent parking lot monitoring method according to claim 1, further comprising sending a prompt message for a vehicle leaving the parking space if the vehicle is not a vehicle that has driven away from the parking space that has been intruded.

13. An intelligent parking lot monitoring system for performing the intelligent parking lot monitoring method according to any one of claims 1 to 12, comprising:

the alarm module is used for sending alarm information to the vehicle leaving from the parking space if the vehicle leaves from the invaded parking space;

wherein, it specifically includes to leave car detection module:

14. A computer device, comprising:

one or more processors;

a memory for storing one or more programs;

when executed by the one or more processors, cause the one or more processors to implement the parking lot intelligent monitoring method of any one of claims 1-12.