Urban surveillance method and system

Abstract

The invention relates to an urban surveillance method, comprising a training phase (100), intended to create training data; a detection phase (200), in which at least one detection vehicle (20) equipped with at least one sensor is mobile in an area to be monitored (22), the at least one sensor records data relating to the area to be monitored (22), then the recorded detection data is processed by a computer detection device (28), configured to detect elements (24) present in the area to be monitored using the training data; and an analysis phase (300), in which the area to be monitored (22) and the elements (24) present in the area to be monitored (22) are analyzed by a computer analysis device (38); characterized in that the training phase (100) and the detection phase (200) are distinct;and in that during the analysis phase (300), the computer analysis device (38) establishes a score for the area to be monitored (22), taking into account various parameters, including the different elements (24) present in the area to be monitored (22). The invention also relates to an urban surveillance system. Abbreviated figure: 1;

Description

Title of the invention: Urban surveillance method and system technical field

[0001] The present invention relates to a method and system of urban surveillance, in particular to ensure the cleanliness of cities.

[0002] The field of the invention is that of urban surveillance systems, comprising a mobile vehicle in an area to be monitored, and a computer device configured for the detection of elements present in that area. Prior art

[0003] Prior art already provides urban surveillance systems, including vehicles or urban equipment fitted with sensors, such as cameras. The data can be processed by human operators, assisted by computer devices. However, these tasks are tedious, and the attention span of human operators is variable. Alternatively, the data can be processed without human operators, by an artificial intelligence model trained to detect different elements.

[0004] Document CN109165582B describes an example of a surveillance system, including an artificial intelligence model trained for the detection of various features. A street cleaning vehicle equipped with a high-resolution camera travels through the streets. The geolocated images are transmitted to a computer server for temporary storage. The server segments the images into 420x400 pixel segments for model training and feature detection. However, this method requires the transfer of large amounts of data to the server.

[0005] The object of the present invention is to propose an improved method and monitoring system.

[0006] To this end, the invention relates to a method of urban surveillance, comprising: - a training phase, in which at least one training vehicle equipped with at least one sensor is mobile in a training area, at least one sensor records data relating to the training area, then the recorded training data is processed by a computer training device, configured to detect elements present in the training area and create learning data;

[0007] - a detection phase, in which at least one detection vehicle equipped If at least one sensor is mobile within a monitored area, that sensor records data relating to the monitored area, and then the detection data recorded data are processed by a computer detection device, configured to detect elements present in the area to be monitored using training data;

[0008] - an analysis phase, in which the area to be monitored and the elements present in the area to be monitored is analyzed by a computer analysis device;

[0009] characterized in that the training phase and the detection phase are distinct; and in that during the analysis phase, the computer analysis device establishes a score for the area to be monitored, taking into account different parameters, including the different elements present in the area to be monitored.

[0010] Thus, the invention improves the efficiency of both the training and detection phases. The training data can be processed initially in a data processing center. The detection data can then be processed either in the data processing center or directly in the detection vehicle. Following the analysis phase, policymakers, municipal services, and sanitation operators can take the necessary measures to ensure the city's cleanliness.

[0011] According to other advantageous features of the invention, taken individually or in combination:

[0012] - The training phase and the detection phase are temporally distinct. This means that these phases can be carried out over different time intervals.

[0013] - The training phase and the detection phase are materially distinct. This means that these phases can be carried out with different equipment. In particular, these phases can be carried out with different computer devices.

[0014] - The computer training device includes an intelligence model artificially trained for the detection of various elements, including animal droppings.

[0015] - The computer training device includes an intelligence model artificial intelligence trained to detect various items, including bottles, cans, overflowing bins and plastic bags.

[0016] - The computer detection device includes an intelligence model artificially trained for the detection of different elements.

[0017] - The computer detection device includes the same intelligence model artificial than the computer training device, trained using training data and learning data.

[0018] - The training vehicle is equipped with at least two sensors in the form of a left camera and a right camera.

[0019] - The detection vehicle is equipped with at least two sensors in the form of a left camera and a right camera.

[0020] - The computer analysis device aggregates data from different sources to establish the score.

[0021] - The data sources aggregated by the computer analysis device include at least one internal source of data transmitted by the computer detection device and at least one external source of data relating to the area to be monitored.

[0022] - The area to be monitored is a city, and the score is established for the cleanliness of that city.

[0023] - The computer training device processes the training data and detects elements present in the training area with a delay relative to the recording by at least one sensor fitted to the training vehicle.

[0024] - Preferably, the computer detection device processes the detection data and detects the elements present in the area to be monitored in real time relative to the recording by at least one sensor equipping the detection vehicle.

[0025] - Alternatively, the computer detection device processes the detection data and detects the elements present in the area to be monitored with a delay relative to the recording by at least one sensor equipping the detection vehicle.

[0026] The invention also relates to an urban surveillance system, comprising:

[0027] - training equipment, comprising:

[0028] - at least one training vehicle equipped with at least one sensor and configured to move within a training area, while at least one sensor records data relating to the training area,

[0029] - a computer training device, configured to process data training data recorded by at least one sensor fitted to the training vehicle, detect elements present in the training area and create learning data;

[0030] - detection equipment, including:

[0031] - at least one detection vehicle equipped with at least one sensor and configured for to move within a monitored area, while at least one sensor records data relating to the monitored area,

[0032] - a computer detection device, configured to process data training data recorded by at least one sensor equipping the vehicle to detect and detect elements present in the area to be monitored using the learning data;

[0033] - analytical equipment, including a computer analysis device configured to analyze the area to be monitored and the elements present in the area to be monitored;

[0034] characterized in that the training equipment is configured to operate during a training phase and the detection equipment is configured to operate during a detection phase distinct from the training phase;

[0035] and in that the computer analysis device is configured to establish a score for the area to be monitored, taking into account various parameters, including the different elements present in the area to be monitored. Description of the figures

[0036] The invention will be better understood upon reading the following description, given solely by way of non-limiting example and made with reference to the accompanying drawings in which:

[0037] [Fig. 1] is a diagram illustrating the method, comprising the training phase, the detection phase and the analysis phase.

[0038] [Fig.2] is a diagram illustrating the training phase.

[0039] [Fig.3] is a diagram illustrating the detection phase.

[0040] [Fig.4] is a diagram illustrating the analysis phase. Detailed description of the invention

[0041] Figures 1 to 4 show a method and an urban surveillance system according to the invention.

[0042] The method comprises a successive training phase (100), a detection phase (200) and an analysis phase (300).

[0043] During the training phase (100), at least one training vehicle (10) equipped with at least one sensor is mobile within a training area (12). The sensors record data relating to the training area (12). The recorded training data is then processed by a computer training device (18), configured to detect elements (14) present in the training area (12) and create training data.

[0044] During the detection phase (200), at least one detection vehicle (20) equipped with at least one sensor is mobile within a monitored area (22). The sensors record data relating to the monitored area (22). The recorded detection data is then processed by a computer detection device (28), configured to detect elements (24) present in the monitored area (22) using training data.

[0045] During the analysis phase (300), the area to be monitored (22) and the elements (24) present in the area to be monitored (22) are analyzed by a computer analysis device (38). The computer analysis device (38) establishes a score for the area to be monitored (22), taking into account various parameters, including the different elements (24) present in the area to be monitored (22).

[0046] In [Fig.2], the training phase (100) is detailed in different stages (110, 120, 130, 140, 150, 160, 170, 180, 190).

[0047] Step (110) consists of preparing a training vehicle (10) equipped with at least one sensor. Step (110) is carried out at the operating base, for example, premises belonging to the operator of the urban surveillance system. In particular, step (110) consists of setting up and checking the equipment of the vehicle (10). This equipment may include the sensors, a geolocation device to identify the position of the vehicle (10), a hard drive to store the data recorded by the sensors and the associated geolocation data, etc. Preferably, the vehicle (10) is equipped with at least two sensors in the form of a left camera and a right camera. The vehicle (10) may also be equipped with a front-facing camera mounted on a mast to map the city streets.

[0048] Step (120) consists of moving the training vehicle (10) within a training area (12), while the vehicle's sensors (10) record data relating to the training area (12), including the presence of notable features (14). The vehicle (10) moves along pre-prepared routes, filming sidewalks and roadsides. When the training area (12) is a city, the vehicle (10) makes several circuits within the city to obtain as much data as possible for training.

[0049] Step (130) consists of storing the data recorded by the sensors on the hard drive. The data recorded by the sensors are, for example, high-resolution images. Advantageously, step (130) can be carried out continuously, simultaneously with step (120).

[0050] Step (140) consists of backing up the hard drive data to a storage server once the vehicle (10) has returned to the base of operations. When the data consists of high-resolution images, their resolution can be reduced to decrease the storage volume on the server. The data stored on the server can be displayed in step (320) by the mayor or municipal services.

[0051] Step (150) consists of encrypting the data stored on the hard drive, and then transmitting this hard drive to a processing center equipped with the computer drive device (18). This transmission of the hard drive can be carried out by mail, courier, or delivery person. The data volumes are very large, on the order of several terabytes. Thus, transmission via hard drive is preferred to online transmission.

[0052] Step (160) consists of decrypting and processing the data previously recorded on the hard drive using the computer training device (18). This data These constitute training data. The device (18) is trained to detect notable features (14) present in the training area (12). More specifically, the device (18) includes an artificial intelligence model trained using the training data. Depending on the training sessions performed, the device (18) can be trained to detect several dozen different features (14).

[0053] By way of non-limitation, the elements (14) detectable by the device (18) may include in particular animal droppings, bottles, cans, overflowing bins, plastic bags, cigarette butts, free parking spaces, damage to the urban environment (damaged bus shelter, degraded road surface, etc.).

[0054] Step (170) consists of saving the training results as learning data, usable in the detection phase (200).

[0055] In [Fig.3], the detection phase (200) is detailed in different stages (210, 220, 230, 240, 250).

[0056] Step (210) consists of preparing a detection vehicle (20) equipped with at least one sensor. Step (210) is carried out at the operating base, for example, premises belonging to the operator of the urban surveillance system. In particular, step (210) consists of setting up and checking the equipment of the vehicle (20). This equipment may include the sensors, a geolocation device to identify the position of the vehicle (20), a hard drive to store the data recorded by the sensors and the associated geolocation data, the detection computer (28), etc. Advantageously, installing the detection computer (28) directly in the vehicle (20) avoids data transmissions, allows for real-time data processing, and yields results more quickly. Preferably, the vehicle (20) is equipped with at least two sensors in the form of a left-hand camera and a right-hand camera.

[0057] Step (220) consists of moving the detection vehicle (20) into a monitored area (22), while the vehicle's sensors (20) record data relating to the monitored area (22), including the presence of notable features (24). The vehicle (20) moves along pre-prepared routes, filming sidewalks and roadsides.

[0058] Step (230) consists of storing the data recorded by the sensors on the hard drive. Advantageously, step (230) can be carried out continuously, simultaneously with step (220).

[0059] Step (240) consists of processing the data recorded on the hard drive using the computer detection device (28). Advantageously, the device (28) can be installed directly in the vehicle (20), in order to avoid data transmissions, process the data in real time, and obtain results more quickly. The device (28) is trained to detect notable features (24) present in the area to be monitored (22). More specifically, the device (28) includes an artificial intelligence model trained using training data. Depending on the training performed on the device (18) and the training data, the device (28) can be trained to detect several dozen different features (24).

[0060] By way of non-limitation, the elements (24) may include in particular animal droppings, bottles, cans, overflowing bins, plastic bags, cigarette butts, free parking spaces, damage to the urban environment (damaged bus shelter, degraded road surface, etc.).

[0061] Step (250) consists of saving the detection results as detection data, which can be used in the analysis phase (300). In particular, the hard drive data can be saved to a storage server once the vehicle (20) has returned to the base of operations. When the data consists of high-resolution images, their resolution can be reduced to decrease the storage volume on the server. The data saved on the server can be displayed during step (320) by the mayor or municipal services.

[0062] The analysis phase (300) includes the steps (310, 320).

[0063] Step (310) consists of making the detection results available in a form usable by an operator. According to a particular embodiment, following the training phase (100) but without the detection phase (200) having been carried out, step (310) may consist of making the training results available in a form usable by an operator. Indeed, the data resulting from the training or detection may require processing to be usable by the operator. This processing can be carried out using the computer analysis device (38).

[0064] Step (320) consists of displaying the detection and / or training results on an interface, allowing the output of a score per geographical unit: street segment, street, neighborhood, city, or metropolitan area comprising several cities. Advantageously, this display can be performed on a web portal accessible to the mayor and municipal services. The web portal can display images of notable features present in the area to be monitored. A city's mayor may be interested in a city-wide score, allowing them to compare their performance with other cities, either to highlight their positive policies or to continue their efforts to improve. The information can be transmitted to municipal services, particularly sanitation operators, so that they can plan their interventions. The score can be calculated using device (38).

[0065] According to a particular embodiment, the computer analysis device (38) can aggregate data from different sources to establish the score. In particular, the data sources aggregated by the device (38) can include internal and external data sources relating to the area to be monitored (22). An internal source can be the computer detection device (28). An external data source can be an application made available to citizens for reporting detected waste, or the monitoring system of another monitoring operator.

[0066] According to a preferred embodiment, the area to be monitored (22) is a city, and the score is established for the cleanliness of that city. In practice, the cleanliness of a city is not easy to determine and analyze. There is a need to conduct a cleanliness audit of the city and make this information accessible and easy to understand. Through a summary score and a user-friendly interface, the system according to the invention makes clear information available to policymakers, municipal services, and sanitation operators.

[0067] According to other embodiments, the area to be monitored (22) can be a street, a neighborhood, or a metropolis comprising several cities. Thanks to the scores per unit of territory, it is possible to compare the cleanliness of different areas, in order to better target cleaning interventions, as well as communication with residents, to promote virtuous behavior.

[0068] When the urban surveillance system is in operation, different modes of operation can be considered. Initially, it is important to complete phases 100, 200, and 300 to create the training data and train the model. Subsequently, once the training data is available for the detection phase, i.e., once the model is trained, it is possible to complete only steps 200 and 300 for each intervention of vehicle 20. Alternatively, it is possible to regularly repeat phase 100 to create new training data and continue training the model.

[0069] By way of non-limitation, the devices (18; 28) can be configured to detect the following items (14; 24): cigarette, leaf, group of leaves, cardboard, can, glass bottle, PET bottle, carton / box, food packaging, newspaper, broken glass, syringes, transparent plastic, opaque plastic, dog feces, small dog feces bags, masks, capsules. Other items (14; 24) may be considered within the scope of the invention.

[0070] Moreover, the method and the urban surveillance system may be different from figures 1 to 4 without departing from the scope of the invention, which is defined by the claims.

[0071] In practice, the training vehicle (10) and the detection vehicle (20) can be separate or identical. In other words, the training phase (100) and the detection phase (200) can be carried out with separate vehicles (10; 20), or with the same vehicle (10; 20).

[0072] Furthermore, the training area (12) and the monitoring area (22) may be separate or identical. For example, the same city may be used as both the training area (12) and the monitoring area (22). Alternatively, one city may be used as the training area (12), and other cities may be used as monitoring areas (22).

[0073] The computer training device (18), the computer detection device (28) and the computer analysis device (38) can form a single computer system.

[0074] Alternatively, the devices (18, 28, 38) may belong to separate computer systems. For example, device (18) is installed in one city, while devices (28, 38) are installed in a second city.

[0075] The training phase (100) and the detection phase (200) are temporally distinct. This means that these phases can be carried out over different time intervals.

[0076] The training phase (100) and the detection phase (200) are physically distinct. This means that these phases can be carried out with different equipment. In particular, these phases can be carried out with different computer devices (18, 28). Furthermore, the training phase (100) and the detection phase (200) can be carried out with different vehicles (10, 20).

[0077] When the training vehicle (10) and the detection vehicle (20) are different, they are nevertheless preferably equipped with the same sensors for the detection of the elements (14, 24), with the same configurations (physical characteristics, position and orientation on the vehicle, resolution).

[0078] The computer training device (18) processes the training data and detects the elements (14) present in the training area (12) with a delay relative to the recording by the sensors equipping the training vehicle (10).

[0079] Preferably, the computer detection device (28) processes the detection data and detects the elements (24) present in the area to be monitored in real time relative to the recording by the sensors equipping the detection vehicle (20). Thus, the operator can react in real time to send a team to the position of the detected elements (24).

[0080] Alternatively, the computer detection device (28) can process the detection data and detect the elements present in the area to be monitored (22) with a delay relative to the recording by the sensors equipping the detection vehicle (20).

[0081] Beyond the detection and collection of waste, other possible applications include parking management, monitoring of road conditions, and all detection of relevant road information.

[0082] The technical characteristics of the various embodiments and variants mentioned above can be combined, in whole or in part. Thus, the method and the system can be adapted in terms of cost, functionality, and performance.

Claims

Demands

1. Urban surveillance method, comprising: - a training phase (100), in which at least one training vehicle (10) equipped with at least one sensor is mobile in a training area (12), the at least one sensor records data relating to the training area (12), then the recorded training data is processed by a computer training device (18), configured to detect elements (14) present in the training area (12) and create training data;- a detection phase (200), in which at least one detection vehicle (20) equipped with at least one sensor is mobile in a monitored area (22), the at least one sensor records data relating to the monitored area (22), and then the recorded detection data is processed by a computer detection device (28), configured to detect elements (24) present in the monitored area (22) using the training data; - an analysis phase (300), in which the monitored area (22) and the elements (24) present in the monitored area (22) are analyzed by a computer analysis device (38); characterized in that the training phase (100) and the detection phase (200) are distinct;in that during the analysis phase (300), the computer analysis device (38) establishes a score for the area to be monitored (22), taking into account various parameters, including the different elements (24) present in the area to be monitored (22).

2. Method according to claim 1, characterized in that the computer training device (18) comprises an artificial intelligence model trained for the detection of different items (14), including animal droppings.

3. Method according to any one of claims 1 or 2, characterized in that the computer training device (18) comprises an artificial intelligence model trained for the detection of various items (14), including bottles, cans, overflowing bins and plastic bags.

4. A method according to any one of claims 1 to 3, characterized in that the training vehicle (10) is equipped of at least two sensors in the form of a left camera and a right camera.

5. Method according to any one of claims 1 to 4, characterized in that the detection vehicle (20) is equipped with at least two sensors in the form of a left camera and a right camera.

6. Method according to any one of claims 1 to 5, characterized in that the computer analysis device (38) aggregates data from different sources to establish the score.

7. Method according to any one of claims 1 to 6, characterized in that the data sources aggregated by the computer analysis device (38) include at least one internal data source transmitted by the computer detection device (28) and at least one external data source relating to the area to be monitored (22).

8. Method according to any one of claims 1 to 7, characterized in that the area to be monitored is a city, and the score is established for the cleanliness of that city.

9. Method according to any one of claims 1 to 8, characterized in that the computer drive device (18) processes the drive data and detects the elements (14) present in the drive zone (12) with a delay relative to the recording by at least one sensor equipping the drive vehicle (10).

10. Urban surveillance system, comprising: - training equipment, including: - at least one training vehicle equipped with at least one sensor and configured to move within a training area, while at least one sensor records data relating to the training area, - a training computer system, configured to process the training data recorded by at least one sensor equipping the training vehicle, detect elements present in the training area and create training data; - detection equipment, including: - at least one detection vehicle equipped with at least one sensor and configured to move within an area to be monitored, while at least one sensor records data relating to the area to be monitored, - a computer detection device, configured to process the training data recorded by at least one sensor equipping the detection vehicle and to detect elements present in the area to be monitored using the learning data; - analytical equipment, including a computer analysis device configured to analyze the area to be monitored and the elements present in the area to be monitored; characterized in that the training equipment is configured to operate during a training phase and the detection equipment is configured to operate during a detection phase separate from the training phase; and in that the computer analysis device is configured to establish a score for the area to be monitored, taking into account various parameters, including the different elements present in the area to be monitored.

Images