Low-power camera trap, featuring multiple sensors

Abstract

Low-power camera trap comprising a plurality of sensors. The invention relates to a camera trap (1) for tracking and monitoring animals, people, and / or vehicles, comprising a camera (6), a first processor (3) and a second processor (8), a passive infrared sensor (4), and an auxiliary presence sensor (5). The first processor is configured to analyze the measurements taken by the auxiliary presence sensor when the passive infrared sensor signals detection, so as to confirm the detection and trigger the camera. The second processor is configured to analyze the images obtained by the camera to identify the animals, people, and / or vehicles. Figure for the abstract: Fig. 1

Description

Title of the invention: Low-power photo trap, comprising a plurality of sensors Technical field

[0001] The present invention relates to the field of camera traps.

[0002] A camera trap is a device intended in particular for the observation or monitoring of animal diversity in the natural environment and allows for taking pictures without human intervention. It is typically a stand-alone device comprising an on-board shooting device, such as a camera, which is automatically triggered when movement is detected by the system itself or by a separate sensor.

[0003] The invention relates more specifically to a low-power consumption photographic trap comprising a plurality of sensors.

[0004] The invention also relates to a method for detecting animals, people and / or vehicles using this camera trap.

[0005] The main area of ​​application envisaged concerns animal monitoring and surveillance. This may concern in particular the management of a nature reserve or a regional or national park by carrying out a census of the species present with automatic counting and / or monitoring of sensitive animals such as the wolf or the lynx; the monitoring and protection of livestock farms, particularly in pastoral areas; the analysis of environmental impacts of industrial land and agro-industrial operations; motorway monitoring, particularly of eco-bridges and eco-ducts; a pest intrusion alert system coupled or not with a device of the hawk-scarer type. Prior art

[0006] Animal tracking and monitoring devices generally come in the form of camera traps or acoustic recorders.

[0007] Camera traps fall into two categories: those with local data storage and those with remote data transmission.

[0008] Photographic traps comprising local storage record all the data obtained, which may in particular include images and / or videos, on a storage medium integrated into the trap, typically a removable memory card such as an SD card integrated into the trap's camera. This recording means offers a substantial storage space, but has the major disadvantage of not uploading any information in real time. The readings of these photo traps are carried out manually periodically, and constitute a tedious and repetitive task. for the operators in charge of their deployment.

[0009] Traps comprising remote data transmission allow captured photographs or videos to be sent remotely. These devices are generally equipped with a mobile phone card such as a SIM card, in order to transmit the data via the cellular network, for example 3G or 4G networks. Unlike traps comprising local storage, it is possible here to know in real time the presence of wild animals near the trap.

[0010] This way of proceeding, however, has several drawbacks.

[0011] First, the deployment of traps is limited by cellular network coverage, which is often weak and unstable in remote natural areas.

[0012] Then, transmissions by cellular network are very energy-intensive and significantly degrade the autonomy of camera traps: sending photographs or videos by 3G or 4G cellular network is not suitable for loT systems (acronym for "Internet of Things"). In addition, each trap requires a SIM card and a subscription to a telephone operator, which hinders the deployment of traps on a large scale, typically of the order of several dozen traps in a nature reserve.

[0013] Currently commercially available camera traps use a so-called PIR (Passive InfraRed sensor) motion sensor to trigger the taking of photographs. Such a sensor has the advantage of consuming very little energy but is still subject to activation errors following false detections, for example caused by variations in brightness or the movement of the foliage of a tree. In addition, PIR sensors operate poorly when directly illuminated by sunlight. In addition, the field of vision of a PIR sensor is typically restricted: as a result, this type of sensor is not responsive enough to capture certain fast-moving animals.

[0014] It is possible to carry out a posteriori a digital processing of the images in order to eliminate false detections of the passive infrared sensor. This allows a saving of time thanks to an automatic sorting of the empty images, but it does not solve the problem of having activated the trap camera in an unjustified manner, thus degrading the energy autonomy of the photographic trap.

[0015] Acoustic recorders are less commonly used tools than camera traps but allow for the collection of different information on the biodiversity of a location. These devices continuously capture surrounding sounds: they are particularly suitable for the identification of birds, bats or small amphibians, which are very difficult to capture with a camera trap.

[0016] In any case, commercial animal tracking and monitoring devices Currently commercialized have real drawbacks, the main one being the lack of local processing of information and therefore a lack of responsiveness to subjects detected by the devices.

[0017] In addition, manually processing the results for each reading can be problematic, especially when several dozen devices are deployed.

[0018] Publication [1] discloses an animal classification device based on a Raspberry Pi single-board computer coupled with a commercially available camera trap, also comprising a module for transmitting data to a remote server with an alert system via a mobile application. This publication also describes a study of different neural network models adapted to animal classification from images. The processing of the images acquired by the trap is carried out locally, the results being transmitted to the remote server using an IoT protocol. The solution described is however very energy-intensive: a consumption of 106 Wh over 24 hours is mentioned for the single-board computer, without taking into account the energy consumption of the camera trap. In addition, the device described uses a commercially available camera trap and therefore remains subject to the problem of false detections of these traps.

[0019] Publication [2] makes a comparison between two audio and video methods of animal detection and discloses an automatic classification of recorded audio data, allowing animals to be classified according to their cries. This publication describes the use of deep learning algorithms applied to spectrograms. Audio classification is not performed in real time and therefore cannot be used as a means of detection. The audio and video analysis processes are performed separately and do not allow data to be cross-referenced in real time on the same computing unit.

[0020] Publication [3] studies the combination of an audio device (passive microphone in continuous listening) and a video device for animal surveillance and demonstrates the interest of coupling different types of data.

[0021] However, the solutions described above do not allow real-time classification of data obtained by several sensors.

[0022] There is therefore a need to propose an improved camera trap, in particular one which minimizes energy consumption while allowing real-time classification of the measured data.

[0023] The aim of the invention is to meet at least part of this need. Statement of the invention

[0024] To this end, the invention relates in one of its aspects to a photographic trap for tracking and monitoring animals, people and / or vehicles, comprising

[0025] - a shooting device;

[0026] - a first processor and a second processor;

[0027] - a passive infrared sensor configured to detect a triggering event and sending a first signal to the first processor when a trigger event is detected;

[0028] - an auxiliary presence sensor configured to carry out measurements allowing detection of animals, people and / or vehicles,

[0029] the first processor being configured to analyze the measurements made by the auxiliary presence sensor upon receipt of the first signal so as to confirm a detection and, when a detection is confirmed, trigger the shooting device and send a second signal to the second processor,

[0030] the second processor being configured to analyze the images obtained by the shooting device upon receipt of the second signal so as to identify the animals, people and / or vehicles in the images.

[0031] By "processor" is meant an electronic circuit or component capable of executing machine instructions. A processor may in particular be a microprocessor or an integrated circuit.

[0032] By "triggering event" is meant an event occurring in the field of vision of the passive infrared sensor and triggering the latter. A triggering event may in particular be the passage of an animal, a person or a vehicle in the field of vision of the infrared sensor but may also be a false detection for example due to the movement of vegetation or a variation in brightness.

[0033] Thus, according to the invention, a photographic trap comprising at least two distinct processors and two distinct presence sensors is implemented.

[0034] The presence of an auxiliary presence sensor, in addition to the passive infrared sensor, makes it possible to eliminate false detections and thus to avoid triggering the image-taking device, which may in particular be a camera or a photographic device, in the absence of actual detection of an animal, a person or a vehicle. Advantageously, this improves the energy autonomy of the camera trap and facilitates the sorting of the images obtained by eliminating empty images.

[0035] In addition, the auxiliary presence sensor can make it possible to anticipate the passage of an animal, giving the camera trap better responsiveness. This is important in the case of animals or vehicles moving quickly because the responsiveness of the infrared sensor may be insufficient in this case.

[0036] The use of two separate processors advantageously allows for significant energy savings. Indeed, the first processor manages tasks which do not do not require significant computing power. The first processor can therefore be a low-power processor. Image analysis, which requires high computing power, is carried out by the second processor. The latter may only operate when a detection is confirmed in order to analyze and classify the resulting images.

[0037] The second processor allows images acquired by the shooting device to be processed locally and immediately, unlike existing solutions offering remote processing either on a third-party system or on a remote server.

[0038] Immediate data processing has the particular advantage of enabling a real-time reaction: this may include remote transfer of information via communication protocol for the Internet of Things (IoT), an alert on a mobile phone or the activation of a scarer such as an audible alarm or a light flash.

[0039] According to an advantageous variant, the first processor is configured to power up the second processor when sending the second signal and to power down when the image analysis is complete. In this way, the second processor does not consume energy when no image analysis is in progress.

[0040] Advantageously, the first processor is a low or ultra-low power processor. In particular, the electrical current consumption of the first processor in a standby mode is less than 20 pA, preferably less than 10 pA, 200 nA, or 100 nA.

[0041] Indeed, this processor remains constantly powered when the camera trap is in operation. It is therefore preferable that its energy consumption is controlled. Advantageously, the first processor can remain in standby mode as long as the infrared sensor does not send it any detection signal to further limit its consumption.

[0042] According to an advantageous characteristic, the exit time from standby mode of the first processor is less than 50 ps, ​​preferably 10 ps, ​​more preferably 5 ps. This makes it possible to limit the latency between the passage of an object in front of the photographic trap and the triggering of the shooting device.

[0043] According to one embodiment, the auxiliary presence sensor comprises at least one microphone configured to pick up sounds emitted by animals, people and / or vehicles. Preferably, the auxiliary presence sensor comprises at least one microphone sensitive to audible frequencies and / or sensitive to ultrasound.

[0044] Advantageously, the microphones typically operate with particularly low power consumption, which contributes to the energy efficiency of the camera trap. In addition, the use of direct microphones tionals allows good prediction of the passage of animals, people or vehicles in the field of vision of the shooting device.

[0045] The use of other types of sensors is however conceivable. Thus, the auxiliary presence sensor may comprise at least one of a rangefinder, a lidar sensor, a photodetector array and a thermal camera. An example of a photodetector array is known from the manufacturer STMicroelectronics under the reference VL53L5CX.

[0046] Advantageously, these sensors provide data that requires significantly less processing than audio data in order to confirm the presence of an animal, human or vehicle.

[0047] According to an advantageous variant, the camera trap further comprises a long-distance communication device, preferably according to a communication protocol for the Internet of Things, in particular a LoRaWAN radio communication protocol. By "long-distance communication" is meant the usual meaning of the technology, namely communication over a possible distance of up to several thousand kilometers.

[0048] Preferably, the camera trap comprises an electric battery configured to power it. The capacity of the battery is preferably greater than 5000 mAh or 10,000 mAh in order to allow the trap to operate for a long time.

[0049] More preferably, the photo trap comprises a photovoltaic panel configured to electrically power the components of the photo trap.

[0050] The inventors have developed a camera trap consuming approximately 3 Wh in 24 hours. With a battery with a capacity of 10,000 mAh, the camera trap can achieve an autonomy of approximately one month under normal conditions of use, with a maximum of 20 triggers of the shooting device per day and a temperature between 10°C and 30°C.

[0051] The invention also relates to a method for detecting animals, people and / or vehicles implemented by at least one camera trap as described previously, comprising the steps consisting of:

[0052] a) powering up the camera trap, the second processor being kept powered down;

[0053] b) when the passive infrared sensor is triggered, sending the first signal to the first processor and analyzing by means of the first processor the measurements of the auxiliary presence sensor in order to confirm the detection of an animal, a person or a vehicle;

[0054] c) if the detection is confirmed, taking at least one image with the image-taking device, switching on the second processor, sending the second signal to the second processor and analyzing by means of the second processor the at least one image so as to identify an animal, a person or a vehicle present in the image.

[0055] Preferably, the measurements of the auxiliary presence sensor comprising audio data.

[0056] According to an advantageous characteristic, the first processor periodically performs, preferably at least 5 times, 10 times or 20 times per second, an acquisition and an analysis of the audio data in order to calculate the power spectral density of the audio data at different frequencies, the first processor performing an analysis of the most recent power spectral density when it receives the first signal from the passive infrared sensor to confirm a detection. Brief description of the drawings

[0057] [Fig-1] [Fig.l] is a functional diagram of a camera trap according to the invention.

[0058] [Fig.2] [Fig.2] represents a photographic trap according to the invention.

[0059] [Fig.3] [Fig.3] is a flowchart detailing the operation of a camera trap according to the invention. Detailed description

[0060] [Fig.l] is a functional diagram of the main elements of a camera trap 1 according to the invention.

[0061] In the embodiment of [Fig.l], the photographic trap 1 comprises a first electronic card 2 comprising a first processor 3.

[0062] A passive infrared sensor 4 is connected to the first electronic card 2 and sends a first signal to the first processor 3 in the event of detection of a triggering event. A triggering event may in particular be the passage of an animal, a person or a vehicle in the field of vision of the infrared sensor but may also be a false detection, for example due to the movement of vegetation or a variation in brightness.

[0063] The photographic trap 1 also includes an auxiliary presence sensor 5 connected to the first electronic card 2.

[0064] In the illustrated example, the auxiliary sensor 5 comprises a microphone sensitive to audible frequencies and a microphone sensitive to ultrasound. The microphones may be of the microelectromechanical system (MEMS) type, for example the IMP23ABSU microphones of a STEVAL-MIC007V1 electronic card from the manufacturer STMicroelectronics.

[0065] The microphones can pick up the sounds of animal or human footsteps, or vehicle noises and thus confirm good detection by discarding irrelevant trigger events from the infrared sensor.

[0066] Microphones, especially directional microphones, have the advantage of enable the prediction of the passage of an animal, person or vehicle in the field of vision of the camera by analyzing the direction of origin of the footsteps. This is particularly useful for animals or fast-moving vehicles, for which the passive infrared sensor may not be sufficiently responsive.

[0067] When activated by the infrared sensor 4, the first processor 3 analyses the data collected by the auxiliary sensor 5. If this analysis confirms the effective detection of an object, the first processor 3 triggers a shooting device 6 and sends a second signal to a second processor 8 of a second electronic card 7.

[0068] If the detection is not confirmed, unnecessary triggering of the shooting device and powering up of the second processor are avoided, thus improving the energy performance of the trap.

[0069] When it receives the second signal, the second processor analyzes the images taken by the shooting device 6 to recognize the detected object and classify the images accordingly.

[0070] Preferably, the second processor implements a neural network model suitable for recognizing animals, people and / or vehicles in images, making it possible to classify the images according to the photographed objects. Such models are for example described in publication [1].

[0071] The first processor is for example the ARM Cortex-M4 processor of an STM32L4 microcontroller from the manufacturer STMicroelectronics, having a clock frequency of 80 MHz, a standby consumption of the order of 100 nA and a standby exit time of 5 ps.

[0072] Such a processor typically does not have the computing power necessary to perform an analysis of the images taken by the shooting device, in particular to perform a classification calculation using convolutional neural networks.

[0073] The second processor makes it possible to compensate for the lack of power of the first processor and to carry out the analysis of the images.

[0074] In order to limit the power consumption of the trap, the second processor is only powered on when images are taken and need to be analyzed. The rest of the time, the second processor is kept off in order to minimize the power consumption of the trap.

[0075] The second processor may for example be the ARM Cortex-A72 processor of a Raspberry Pi 4 B single-board computer, having a clock frequency of 1.8 GHz, a standby consumption of the order of 400 mA and a standby exit time of 1 s.

[0076] The camera trap 1 further comprises a long-range telecommunications device 9, preferably using an loT protocol such as a LoRaWAN protocol to transmit data remotely. The LoRaWAN protocol is a very low-speed, very low-power telecommunications protocol.

[0077] The telecommunications device 9 is configured to transmit the images taken by the shooting device 6, the data collected by the auxiliary sensor 5 as well as the results of the analysis of the images by the second processor 8. Preferably, only the results of the analysis of the images by the second processor 8 are transmitted. Indeed, photographic or sound data are voluminous and therefore poorly suited to transmission in LoRaWAN: their transmission would consume a lot of energy and bandwidth.

[0078] The telecommunications device 9 is advantageously connected to the second electronic card 7 and controlled by the second processor 8.

[0079] The photographic trap 1 is electrically powered by an electric battery 10. Preferably, this battery has a capacity greater than 5000 mAh or 10,000 mAh so as to allow the operation of the photographic trap over a long period.

[0080] [Fig.2] is a view of a camera trap 1 according to the invention installed on a TREE.

[0081] The components of the trap 1 are contained in a housing 11 which preferably has a camouflage pattern intended to make the trap more discreet. The infrared sensor 4, the auxiliary sensor 5 and the shooting device 6 are at least partially arranged in openings of the housing 11.

[0082] In the embodiment of [Fig.2], the auxiliary presence sensor 5 comprises a microphone sensitive to audible frequencies 12 and an ultrasonic microphone 13.

[0083] The trap 1 advantageously comprises an illumination panel 14 emitting in the infrared. The illumination panel 14 may in particular comprise a plurality of light-emitting diodes emitting in the infrared. The wavelength(s) of the light emitted by the illumination panel 14 may be chosen so as not to frighten the animals that one seeks to monitor with the trap.

[0084] The illumination panel 14 is advantageously connected to the first electronic card 2 and controlled by the first processor 3.

[0085] The illumination panel 14 can produce an infrared light flash, invisible to the naked eye for humans and animals, so as to illuminate the surroundings of the camera trap in very low light conditions, typically at night or in very cloudy weather. The shooting device 6 can be sensitive to infrared light and then produces a sufficiently bright image even in complete darkness, thanks to the light generated by the illumination panel 14.

[0086] Trap 1 also includes a brightness sensor 15.

[0087] The brightness sensor 15 is advantageously connected to the first electronic card 2 and the first processor 3 processes the signals generated by this sensor.

[0088] The brightness sensor 15 measures the brightness at the level of the photographic trap 1. The measurement is carried out before the triggering of the shooting device 6.

[0089] By comparing the measured brightness value with a reference value, the first processor 3 activates or not the illumination panel 14. Thus, the illumination panel 14 is only activated to take a picture in a low-light environment. This advantageously allows energy savings.

[0090] A switch 16 accessible on the housing makes it possible to control the electrical power supply to the camera trap 1.

[0091] [Fig.3] is a flowchart showing the operation of a camera trap 1 as illustrated in [Fig.l].

[0092] Step 100 marks the initialization of the photographic trap 1, that is to say the powering up of the trap.

[0093] When first powered up, the trap is initialized (step 101). This step may in particular include a series of calibrations of the sensors. The passive infrared sensor 4 may in particular carry out a calibration allowing it to analyze its environment to make its triggers more reliable. During this phase, the first processor 3 may also initialize backup registers used for storing data from the auxiliary sensor 5 on a non-volatile storage memory.

[0094] Once the initialization is complete, all of the components of the trap, including the first processor 3, are put into standby mode (102). The first processor 3 is configured to periodically exit standby mode in order to carry out acquisitions of the data collected by the microphones of the auxiliary sensor 5 (103). This can be carried out at a sampling frequency determined so as to correspond to a spectral band comprising sufficient harmonics to process the signals associated with the stride of an animal. This frequency is for example between 10 Hz and 20 Hz, which corresponds to the maximum frequency of the stride of a wild animal.

[0095] Alternatively, the acquisition of data from the auxiliary sensor can be carried out continuously, to the detriment however of the electrical consumption of the trap.

[0096] Advantageously, the data acquired by the microphones are processed by the first processor in order to facilitate real-time audio analysis and recognition. The processing can take place as follows.

[0097] When a predetermined number N of data acquisitions are carried out, the first processor performs a Fourier transform, preferably of the FFT type, on these N measurements. When a predetermined number M of Fourier transforms are carried out, the first processor establishes a period gram, or periodo gram in English, from these M Fourier transforms.

[0098] The determination of the value of N is carried out in particular as a function of the desired frequency resolution AB (according to the equation AB = 2B / N where B represents the sampling frequency) and the maximum time that is accepted between two gram periods. If the value of N is too high, the time interval between two gram periods will be too long. Preferably, a gram period is established at least every second or every two seconds.

[0099] For example, N can be between 8 and 32, preferably equal to 16, for a value of B of the order of 20 Hz.

[0100] Preferably, the value of M is greater than 10 in order to allow the analysis of the evolution of the audio data over a sliding period of approximately 10 s. M may be between 10 and 20.

[0101] The period grams make it possible to estimate the power spectral density at different frequencies of the audio signals in the listening domains.

[0102] In a step 105, depending on whether or not a triggering event is detected (Y) by the infrared sensor 4, the first processor 3 is taken out of the standby state (106) or kept in the standby state (102).

[0103] When it is awakened, the first processor 3 carries out an analysis of the measurements of the auxiliary sensor 5 carried out in the background (106). In particular, this may consist of the analysis of the most recent gram period if the gram period calculations are carried out. The analysis of several consecutive gram periods, however, allows more precise applications. The analysis of these measurements (107) makes it possible to confirm (Y) or not (N) the actual detection of an animal, a person or a vehicle monitored by the camera trap (108).

[0104] The analysis of the measurements makes it possible, for example, to recognize known patterns combining frequency thresholds and temporal dynamics. These patterns are in particular patterns characteristic of certain animal species or certain categories of animals. For example, it is possible to discriminate between bipedal animals, quadrupedal animals, bicycles or four-wheeled vehicles.

[0105] Optionally, the first processor can also trigger the camera device when the calculated period gram exceeds a certain predefined threshold, even in the absence of triggering of the passive infrared sensor. A threshold exceeding corresponds to a sudden variation in the audio field and would therefore reflect the sudden arrival of an animal in the perimeter of the camera trap. Thus, in this operating mode, the camera device 6 can be triggered on the basis of the measurements of the auxiliary sensor 5 alone.

[0106] If the detection is confirmed, the shooting device 6 is triggered by the first processor 3 and takes one or more photographs (step 109). Then, the second processor 8, previously switched off to minimize the power consumption of the trap, is switched on by the first processor 3 (step 110).

[0107] The first processor 3 then returns to the standby state (step 111) for a determined duration, typically of the order of 1.5 minutes (step 112). This period is preferably determined to be slightly longer than the processing time of the photographs by the second processor 8. After this period, the first processor 3 comes out of standby to turn off the second processor 8 (step 131).

[0108] In parallel, the second processor 8 is initialized (step 120) and verifies that a storage memory such as an SD card is connected (step 121). If this is not the case (N), after a short wait, for example between 1 s and 5 s (step 122), a check is made that a total waiting time (timeout) has not been exceeded (step 123). If this total waiting time is not exceeded (N), we return to step 121. If it is exceeded (Y), the stop signal of the second processor 8 is activated (step 129) and the second processor 8 is deactivated (step 130) and powered down (step 131).

[0109] If the storage memory is properly connected (Y), the photographs taken by the shooting device 6 in step 109 are copied (step 124) in order to be pre-processed (step 125) and classified (step 126) by the second processor 8. The pre-processing provides various improvements to the image, for example a brightness adjustment and / or a correction of defects in the lens of the shooting device. The pre-processing is typically very fast and has a negligible cost in terms of computing power compared to the cost of classification.

[0110] Preferably, the second processor 8 implements a neural network model adapted to the recognition of animals, people and / or vehicles on images to classify the photographs according to the recognized objects.

[0111] The obtained images and data are then recorded in the storage memory (step 127) and can be transmitted to a predetermined recipient by the long-distance transmission device 9 (step 128). Preferably, only the result of the classification is transmitted in order to minimize the amount of data transmitted. The shutdown signal of the second processor is then activated (step 129), informing the first processor 3 that the second processor is ready to be powered down. The second processor is then deactivated (step 130) and powered down (step 131).

[0112] The capture and analysis cycle being completed, the camera trap 1 then returns to the waiting state (step 102), the first processor returning to the standby state.

[0113] Other variants and improvements may be provided without departing from the scope of the invention. List of cited documents

[0114] [1] I. A. Zualkeman et al., "Towards an loT-based Deep Learning Architecture for Caméra Trap Image Classification," 2020 IEEE Global Conférence on Artificial Intelligence and Internet of Things (GCAIoT), Dubai, United Arab Emirates, 2020, pp. 1-6, doi: 10.1109 / GCAIoT51063.2020.9345858

[0115] [2] Hiroto Enari, Haruka S. Enari, Kei Okuda, Tetsuya Maruyama, Kana N. Okuda, « An évaluation ofthe efficiency of passive acoustic monitoring in detecting deer and primates in comparison with caméra traps », Ecological Indicators, Volume 98, 2019, Pages 753-762, ISSN 1470-160X, https: / / doi.Org / 10.1016 / j.ecolind.2018.ll.062

[0116] [3] Rachel T. Buxton, Patrick E. Lendrum, Kevin R. Crooks, George Wittemyer, « Pairing caméra traps and acoustic recorders to monitor the ecological impact of human disturbance », Global Ecology and Conservation, Volume 16, 2018, e00493, ISSN 2351-9894, https: / / doi.Org / 10.1016 / j.gecco.2018.e00493

Claims

Claims

1. Camera trap (1) for tracking and monitoring animals, people and / or vehicles, comprising: - a camera device (6); - a first processor (3) and a second processor (8); - a passive infrared sensor (4) configured to detect a triggering event and send a first signal to the first processor when a triggering event is detected;- an auxiliary presence sensor (5) configured to carry out measurements allowing detection of animals, people and / or vehicles, the first processor being configured to analyze the measurements carried out by the auxiliary presence sensor upon receipt of the first signal so as to confirm a detection and, when a detection is confirmed, trigger the camera device and send a second signal to the second processor, the second processor being configured to analyze the images obtained by the camera device upon receipt of the second signal so as to identify the animals, people and / or vehicles in the images.;

2. A camera trap according to claim 1, the first processor being configured to turn the second processor on when sending the second signal and off when the analysis of the images is complete.

3. Photo trap according to one of the preceding claims, the electrical current consumption of the first processor in a standby mode being less than 200 nA, preferably 100 nA.

4. Photo trap according to claim 3, the exit time from sleep mode of the first processor being less than 50 ps, ​​preferably 10 ps, ​​more preferably 5 ps.

5. Photographic trap according to one of the preceding claims, the auxiliary presence sensor comprising at least one microphone (12, 13) configured to pick up sounds emitted by animals, people and / or vehicles.

6. Photographic trap according to the preceding claim, the auxiliary presence sensor comprising at least one microphone sensitive to audible frequencies (12) and / or to ultrasounds (13).

7. A photo trap according to one of the preceding claims, the auxiliary presence sensor comprising at least one of a rangefinder, a lidar sensor, a photodetector array and a thermal camera.

8. Photo trap according to one of the preceding claims, further comprising a long-distance communication device (9), preferably according to a communication protocol for the Internet of Things, in particular a LoRaWAN radio communication protocol.

9. Photographic trap according to one of the preceding claims, comprising an electric battery (10) configured to electrically power the phototrap.

10. Photo trap according to one of the preceding claims, comprising a photovoltaic panel configured to electrically power the components of the photo trap.

11. Method for detecting animals, people and / or vehicles implemented by at least one camera trap according to one of the preceding claims, comprising the steps of: a) switching on the camera trap, the second processor being kept off; b) when the passive infrared sensor is triggered, sending the first signal to the first processor and analyzing by means of the first processor the measurements of the auxiliary presence sensor in order to confirm the detection of an animal, a person and / or a vehicle; c) if the detection is confirmed, taking at least one image with the camera device, switching on the second processor, sending the second signal to the second processor and analyzing by means of the second processor the at least one image so as to identify an animal, a person and / or a vehicle present in the image.

12. Method according to the preceding claim, the measurements of the auxiliary presence sensor comprising audio data.

13. Method according to the preceding claim, the first processor periodically performing, preferably at least 5 times, 10 times or 20 times per second, an acquisition and an analysis of the audio data in order to calculate the power spectral density of the audio data at different frequencies, the first processor performing an analysis of the most recent power spectral density when it receives the first signal from the passive infrared sensor to confirm a detection.

Images