Object Detection System and Method

Abstract

The present invention discloses an object detection system and method. The object detection system includes: a PIR sensing module for performing first-stage detection of an object; when the PIR sensing module measures the appearance of an obscuring object, it will send an interrupt signal to inform the system and record the PIR signal after the appearance of the obscuring object; an autoencoder for performing binary classification detection on the PIR signal obtained by the PIR sensing module according to the PIR signal characteristics to filter out most false wake-up results; a training module for accurately obtaining the potential feature data of the object through an immediate access data recording method of the PIR signal and training the autoencoder in an unsupervised training manner so that it can reconstruct the object signal. The object detection system and method proposed by the present invention can quickly and accurately distinguish whether an object is approaching under the condition of low power consumption.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent security, and relates to an intelligent security system, and particularly to an object detection system and method. Background Art

[0002] Performing object detection on edge products requires meeting the low-power consumption requirement. Therefore, a multi-stage wake-up method is usually used. First, a PIR sensor is used for the first-stage wake-up, and then a camera is used for the second-stage object recognition. However, due to the high false triggering rate of the sensor, there will be additional power consumption in false wake-up and image shooting.

[0003] In view of this, there is an urgent need to design a new object detection method to overcome at least some of the above defects existing in the existing object detection methods. Summary of the Invention

[0004] The present invention provides an object detection system and method, which can quickly and accurately distinguish whether an object is approaching under the condition of low power consumption.

[0005] To solve the above technical problems, according to one aspect of the present invention, the following technical solution is adopted:

[0006] An object detection system, the object detection system includes:

[0007] A PIR sensing module for performing the first-stage object detection. When the PIR sensing module measures the appearance of an occluder, it will send an interrupt signal to inform the system and record the PIR signal after the appearance of the occluder;

[0008] An autoencoder for performing binary classification detection on the PIR signal obtained by the PIR sensing module according to the PIR signal characteristics to filter out most of the false wake-up results; and

[0009] A training module for accurately obtaining the potential feature data of the object through the data recording method of immediately accessing the PIR signal and training the autoencoder in an unsupervised training manner so that it can reconstruct the object signal.

[0010] As an implementation manner of the present invention, the training module is used to identify the PIR signal characteristics by using an AI model to reduce the false triggering of the first-stage wake-up, reduce the subsequent camera wake-up times and reduce the overall power consumption.

[0011] As an implementation manner of the present invention, the training module includes a reconstruction unit for inputting N pieces of data after the PIR sensing module sends back an interrupt signal when encountering an occluder into the autoencoder trained for a specific object, and the autoencoder reconstructs the signal;

[0012] If it belongs to the target object category, the reconstruction error is less than a threshold value; if it does not belong to the target object category, the signal cannot be reconstructed; the reconstruction error is greater than a threshold value.

[0013] Accurately separate the categories of target objects and non-target objects. If it belongs to the target object category, initiate subsequent actions to reduce incorrect camera wake-up and lower power consumption.

[0014] As an implementation of the present invention, when the PIR sensing module receives infrared rays within a specific wavelength range, it will affect the receiver element of the sensor and generate a signal with a potential difference. When the human body temperature is 36.5 °C, the human body will naturally emit infrared rays with a wavelength of approximately 9000 nm to 10000 nm; this is the natural emission wavelength range of the human body. Compared with other inanimate objects, the infrared rays emitted by the human body are distinguishable (warm-blooded animals will also emit a similar wavelength band). For this characteristic, after the PIR sensing module detects an occlusion (sends an interrupt signal), it will immediately record the subsequent N signals (taking 16 signals for humans as an example), and use this unique signal as a feature to determine whether it is a human.

[0015] As an implementation of the present invention, the autoencoder performs target object detection on the signal generated by the PIR. The autoencoder is trained using the obtained PIR features. The training objective function is to make the reconstruction error between the reconstructed signal and the input signal as small as possible. Eventually, the obtained autoencoder can accurately reconstruct the PIR features of the target object; conversely, if the PIR features are not of the target object, a reconstruction error will be obtained, and it can be determined whether the correct target object is detected by setting the threshold value of the reconstruction error.

[0016] The autoencoder includes an encoder and a decoder, which perform compression and decompression operations respectively; if the signal can be successfully reconstructed, it indicates that the signal has a mapping relationship similar to the target object and is regarded as the same category.

[0017] The autoencoder is used to separate the categories of target objects and non-target objects. If it belongs to the target object category, initiate subsequent actions to reduce incorrect camera wake-up and lower power consumption.

[0018] According to another aspect of the present invention, the following technical solution is adopted: A target detection method, the target detection method includes:

[0019] A PIR signal acquisition step; the PIR sensing module acquires the PIR signal after the appearance of an occlusion.

[0020] A classification detection step; the autoencoder performs binary classification detection on the PIR signal acquired by the PIR sensing module according to the PIR signal characteristics to filter out most of the incorrect wake-up results; and

[0021] Training steps: Accurately obtain the potential characteristic data of the target object through the data recording method of immediately accessing the PI R signal, and train the autoencoder using the unsupervised training method so that it can reconstruct the target object signal.

[0022] As an implementation manner of the present invention, in the training steps, an AI model is used to identify the PI R signal characteristics to reduce the mis-touch in the first-stage wake-up, reduce the subsequent camera wake-up times, and reduce the overall power consumption.

[0023] As an implementation manner of the present invention, the training steps include that after the PI R sensing module transmits an interrupt signal when encountering an obstacle, N pieces of data are fed back and input into the autoencoder trained for a specific target object, and the autoencoder reconstructs the signal;

[0024] If it belongs to the target object category, the reconstruction error is less than a threshold; if it does not belong to the target object category, the signal cannot be reconstructed; the reconstruction error is greater than a threshold;

[0025] Accurately separate the categories of the target object and non-target object. If it belongs to the target object category, subsequent actions are started to reduce the incorrect wake-up of the camera to reduce the power consumption.

[0026] As an implementation manner of the present invention, when the PI R sensing module receives infrared rays in a specific wavelength range, it will affect the receiver element of the sensor and generate a signal with a potential difference. When the human body temperature is 36.5 °C, the human body will naturally emit infrared rays, and the wavelength is about 9000 nm to 10000 nm; this is the natural emission wavelength range of the human body. Compared with other inanimate objects, the infrared rays emitted by the human body are discriminable (warm-blooded animals will also emit a similar wavelength band); for this characteristic, after the PI R sensing module detects an obstacle (transmits an interrupt signal), it will immediately record the subsequent N pieces of signals (taking 16 pieces for humans as an example), and use this unique signal as a feature for judging whether it is a human.

[0027] As an implementation manner of the present invention, the autoencoder performs target object detection on the signal generated by the PI R. The autoencoder is trained using the already obtained PI R characteristics (the target object is a human). The objective function of the training is to make the reconstruction error between the reconstructed signal and the input signal as small as possible. The finally obtained autoencoder can accurately reconstruct the PI R characteristics of the target object; on the contrary, if the PI R characteristics are non-target objects, a large reconstruction error will be obtained, and whether the correct target object is detected is obtained by setting a threshold of the reconstruction error;

[0028] The autoencoder includes an encoder and a decoder, which perform compression and decompression operations respectively; if the signal can be successfully reconstructed, it means that the signal has a mapping relationship similar to the target object and is regarded as the same category;

[0029] Separate the categories of the target object and the non-target object. If it belongs to the target object category, initiate subsequent actions to reduce the incorrect wake-up of the camera and thus lower the power consumption.

[0030] The beneficial effects of the present invention are as follows: The target detection system and method proposed by the present invention can quickly and accurately distinguish whether a target object is approaching under the condition of low power consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a schematic diagram of the composition of the target detection system in an embodiment of the present invention.

[0032] Figure 2 It is a flowchart of the target detection method in an embodiment of the present invention.

[0033] Figure 3 It is a schematic diagram of using a PIR sensing circuit system to extract the characteristic waveform of the target object in an embodiment of the present invention.

[0034] Figure 4 It is a schematic diagram of using an autoencoder for target object detection in an embodiment of the present invention.

[0035] Figure 5 It is a schematic diagram of the autoencoder architecture in an embodiment of the present invention.

[0036] Figure 6 It is a schematic diagram of using the PIR feature for the first-stage target object detection through the autoencoder in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0037] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0038] To further understand the present invention, the preferred implementation schemes of the present invention will be described below in conjunction with embodiments. However, it should be understood that these descriptions are only for further explaining the features and advantages of the present invention, rather than limiting the claims of the present invention.

[0039] The description of this part only focuses on several typical embodiments, and the present invention is not limited to the scope described in the embodiments. The mutual replacement of some technical features between the same or similar prior art means and the embodiments is also within the scope of the description and protection of the present invention.

[0040] The expressions of the steps in each embodiment in the specification are only for convenience of description, and the implementation manner of the present application is not limited by the order of step implementation. The "connection" in the specification includes both direct connection and indirect connection.

[0041] The present invention discloses a target detection system. Figure 1 It is a schematic diagram of the composition of the target detection system in an embodiment of the present invention; please refer to Figure 1, the target detection system includes: a PIR sensing module 1, an autoencoder 2, and a training module 3.

[0042] The PIR sensing module 1 is used for detecting the target object in the first stage. When the PIR sensing module measures the appearance of an obstacle, it will send an interrupt signal to inform the system and record the PIR signal after the appearance of the obstacle.

[0043] The autoencoder 2 is used to perform binary classification detection on the PIR signal obtained by the PIR sensing module according to the PIR signal characteristics to filter out most of the false wake-up results.

[0044] The training module 3 is used to accurately obtain the potential feature data of the target object through the data recording method of immediately accessing the PIR signal and train the autoencoder in an unsupervised training manner so that it can reconstruct the target object signal.

[0045] In an embodiment of the present invention, the training module 3 is used to identify the PIR signal characteristics by using an AI model to reduce the false touch of the first-stage wake-up, reduce the subsequent camera wake-up times, and reduce the overall power consumption.

[0046] In an embodiment of the present invention, the training module 3 includes a reconstruction unit, which is used to obtain N pieces of data after the PIR sensing module sends back an interrupt signal when encountering an obstacle and input them into the autoencoder trained for a specific target object. The autoencoder reconstructs the signal. If it belongs to the target object category, the reconstruction error is less than a threshold; if it does not belong to the target object category, the signal cannot be reconstructed; the reconstruction error is greater than a threshold. Accurately separate the categories of the target object and the non-target object. If it belongs to the target object category, start the subsequent actions to reduce the false wake-up of the camera and reduce the power consumption.

[0047] Figure 3 Schematic diagram of using a PIR sensing circuit system to extract the characteristic waveform of the target object in an embodiment of the present invention; please refer to Figure 3 , in an embodiment of the present invention, when the PIR sensing module 1 receives infrared rays in a specific wavelength range, it will affect the receiver element of the sensor and generate a signal with a potential difference. The human body temperature is about 36.5 °C. At this time, the human body will naturally emit corresponding infrared rays, and the wavelength is about 9000 nm to 10000 nm; this is the natural emission wavelength range of the human body. Compared with other inanimate objects, the infrared rays emitted by the human body are discriminable (warm-blooded animals will also emit a similar wavelength band); for this characteristic, after the PIR sensing module detects an obstacle (sends an interrupt signal), it will immediately record the subsequent N pieces of signals (taking 16 pieces for humans as an example) and use this unique signal as a feature to determine whether it is a human.

[0048] Figure 4Schematic diagram of using an autoencoder for target detection in an embodiment of the present invention; please refer to Figure 4 In an embodiment of the present invention, the autoencoder performs target detection on the signal generated by the PIR. The autoencoder is trained using the obtained PIR features (the target is a person), and the objective function of the training is to minimize the reconstruction error between the reconstructed signal and the input signal. The finally obtained autoencoder can accurately reconstruct the PIR features of the target; conversely, if the PIR features are not the target, a reconstruction error will be obtained, and it can be determined whether the correct target is detected by setting the threshold of the reconstruction error.

[0049] Figure 5 Schematic diagram of the autoencoder architecture in an embodiment of the present invention; please refer to Figure 5 In an embodiment of the present invention, the autoencoder includes an encoder and a decoder, which perform compression and decompression operations respectively; if the signal can be successfully reconstructed, it means that the signal has a mapping relationship similar to the target, and is regarded as the same category.

[0050] Figure 6 Schematic diagram of the first-stage target detection of PIR features through the autoencoder in an embodiment of the present invention; please refer to Figure 6 In an embodiment of the present invention, the autoencoder is used to separate the categories of targets and non-targets. If it belongs to the target category, subsequent actions are initiated to reduce the incorrect wake-up of the camera and lower the power consumption.

[0051] The present invention further discloses a target detection method. Figure 2 Flowchart of the target detection method in an embodiment of the present invention; please refer to Figure 2 The target detection method includes:

[0052]

Step S1

[0053] In an embodiment of the present invention, when the PIR sensing module receives infrared rays in a specific wavelength range, it will affect the receiver element of the sensor and generate a signal with a potential difference. The human body temperature is 36.5 °C, and at this time, the human body will naturally emit infrared rays with a wavelength of about 9000 nm to 10000 nm; this is the natural emission wavelength range of the human body, and compared with other inanimate objects, the infrared rays emitted by the human body are distinguishable (warm-blooded animals will also emit a similar wavelength band); for this characteristic, after the PIR sensing module detects the occluder (sends an interrupt signal), it will immediately record the subsequent N signals (taking 16 signals as an example for a person) and use this unique signal as a feature to determine whether it is a person.

[0054]

Step S2

[0055] In an embodiment of the present invention, the autoencoder performs target detection on the signals generated by the PIR. The autoencoder is trained using the obtained PIR characteristics (the target is a person). The objective function of the training is to make the reconstruction error between the reconstructed signal and the input signal as small as possible. The finally obtained autoencoder can accurately reconstruct the PIR characteristics of the target; conversely, if the PIR characteristics are not the target, a large reconstruction error will be obtained. Whether the correct target is detected is determined by setting a threshold of the reconstruction error;

[0056] The autoencoder includes an encoder and a decoder, which perform compression and decompression operations respectively; if the signal can be successfully reconstructed, it means that the signal has a mapping relationship similar to the target, and is regarded as the same category;

[0057] Separate the categories of the target and non-target. If it belongs to the target category, subsequent actions are initiated to reduce the incorrect wake-up of the camera and thus reduce power consumption.

[0058]

Step S3

[0059] In an embodiment of the present invention, in the training step, an AI model is used to identify the PIR signal characteristics to reduce the false touch in the first-stage wake-up, reduce the subsequent camera wake-up times, and reduce the overall power consumption.

[0060] In an embodiment of the present invention, the training step includes the PIR sensing module transmitting N pieces of data after sending an interruption signal when encountering an obstacle, and inputting them into the autoencoder trained for a specific target. The autoencoder reconstructs the signals.

[0061] If it belongs to the target category, the reconstruction error is less than a threshold; if it does not belong to the target category, the signal cannot be reconstructed; the reconstruction error is greater than a threshold.

[0062] Accurately separate the categories of the target and non-target. If it belongs to the target category, subsequent actions are initiated to reduce the incorrect wake-up of the camera and thus reduce power consumption.

[0063] In summary, the target detection system and method proposed by the present invention can quickly and accurately determine whether a target is approaching under the condition of low power consumption.

[0064] It should be noted that the present application can be implemented in software and / or a combination of software and hardware; for example, it can be implemented using an application specific integrated circuit (ASIC), a general purpose computer, or any other similar hardware device. In some embodiments, the software program of the present application can be executed by a processor to implement the above steps or functions. Similarly, the software program of the present application (including related data structures) can be stored in a computer-readable recording medium; for example, a RAM memory, a magnetic or optical drive, or a floppy disk and similar devices. Additionally, some steps or functions of the present application can be implemented using hardware; for example, as a circuit that cooperates with the processor to execute each step or function.

[0065] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0066] The description and application of the present invention here are illustrative and do not intend to limit the scope of the present invention to the above embodiments. The effects or advantages involved in the embodiments may not be reflected in the embodiments due to various factors. The description of the effects or advantages is not used to limit the embodiments. Modifications and changes to the disclosed embodiments here are possible, and substitutions and equivalents of various components for those of ordinary skill in the art are well-known. Those skilled in the art should clearly understand that the present invention can be implemented in other forms, structures, arrangements, proportions, and using other components, materials, and parts without departing from the spirit or essential characteristics of the present invention. Other modifications and changes can be made to the disclosed embodiments here without departing from the scope and spirit of the present invention.

Claims

1. A target detection system, characterized in that, The target detection system includes: A PIR sensing module for detecting a target object. When the PIR sensing module measures the appearance of an occlusion, it will send an interrupt signal to inform the system and record the PIR signal after the appearance of the occlusion; An autoencoder for performing binary classification detection on the PIR signal obtained by the PIR sensing module according to the PIR signal characteristics to filter out most false wake-up results; and A training module for accurately obtaining the potential feature data of the target object through an immediate access PIR signal data recording method and training the autoencoder using an unsupervised training method so that it can reconstruct the target object signal; The training module includes a reconstruction unit for obtaining several pieces of data after the PIR sensing module sends back an interrupt signal when encountering an occlusion and inputting them into the autoencoder trained for a specific target object, and the autoencoder reconstructs the signal; If it belongs to the target object category, the reconstruction error is less than a threshold; if it does not belong to the target object category, the signal cannot be reconstructed; the reconstruction error is greater than a threshold; Accurately separate the categories of target objects and non-target objects. If it belongs to the target object category, start subsequent actions to reduce false wake-up of the camera and reduce power consumption.

2. The target detection system according to claim 1, wherein: The training module is used to identify the PIR signal characteristics using an AI model to reduce false touches in the first-stage wake-up, reduce the subsequent camera wake-up times, and reduce the overall power consumption.

3. The target detection system according to claim 1, wherein: When the PIR sensing module receives infrared rays in a specific wavelength range, it will affect the receiver element of the sensor and generate a signal with a potential difference; After the PIR sensing module detects an occlusion, it will immediately record several subsequent signals and use this unique signal as a feature for judging whether it is a person.

4. The target detection system according to claim 1, wherein: The autoencoder performs target object detection on the signal generated by the PIR; trains the autoencoder using the obtained PIR characteristics. The objective function of the training is to make the reconstruction error between the reconstructed signal and the input signal as small as possible. The finally obtained autoencoder can accurately reconstruct the PIR characteristics of the target object; conversely, if the PIR characteristics are non-target objects, a reconstruction error will be obtained, and it is determined whether the correct target object is detected by setting the threshold of the reconstruction error; The autoencoder includes an encoder and a decoder, which perform compression and decompression operations respectively; if the signal can be successfully reconstructed, it means that the signal has a mapping relationship similar to the target object and is regarded as the same category; The autoencoder is used to separate the categories of target objects and non-target objects. If it belongs to the target object category, start subsequent actions to reduce false wake-up of the camera and reduce power consumption.

5. A target detection method, characterized in that, The target detection method includes: A PIR signal acquisition step; the PIR sensing module acquires the PIR signal after the appearance of an occlusion; A classification detection step; the autoencoder performs binary classification detection on the PIR signal obtained by the PIR sensing module according to the PIR signal characteristics to filter out most false wake-up results; and Training steps; accurately obtain the potential characteristic data of the target object through the data recording method of immediately accessing the PIR signal, and use the unsupervised training method to train the autoencoder so that it can reconstruct the target object signal; The training steps include several pieces of data sent back by the PIR sensing module after receiving an interruption signal when encountering an obstacle, and input them into the autoencoder trained for a specific target object, and the autoencoder reconstructs the signal; If it belongs to the target object category, the reconstruction error is less than a threshold; if it does not belong to the target object category, the signal cannot be reconstructed; the reconstruction error is greater than a threshold; Accurately separate the categories of the target object and non-target object. If it belongs to the target object category, start subsequent actions to reduce the incorrect wake-up of the camera and reduce power consumption.

6. The target detection method according to claim 5, wherein: In the training steps, an AI model is used to identify the PIR signal characteristics to reduce the false touch in the first-stage wake-up, reduce the subsequent camera wake-up times and reduce the overall power consumption.

7. The target detection method according to claim 5, wherein: When the PIR sensing module receives infrared rays in a specific wavelength range, it will affect the receiver element of the sensor and generate a signal with a potential difference; After the PIR sensing module detects an obstacle, it will immediately record several subsequent signals, and use this unique signal as a feature for judging whether it is a person.

8. The target detection method according to claim 6, wherein: The autoencoder performs target object detection on the signal generated by the PIR, uses the obtained PIR characteristics to train the autoencoder, and the training objective function is that the smaller the reconstruction error between the reconstructed signal and the input signal, the better. Finally, the obtained autoencoder can accurately reconstruct the PIR characteristics of the target object; on the contrary, if the PIR characteristics are non-target objects, a large reconstruction error will be obtained, and whether the correct target object is detected is obtained by setting a threshold of the reconstruction error; The autoencoder includes an encoder and a decoder, which perform compression and decompression operations respectively; if the signal can be successfully reconstructed, it means that the signal has a mapping relationship similar to the target object and is regarded as the same category; Separate the categories of the target object and non-target object. If it belongs to the target object category, start subsequent actions to reduce the incorrect wake-up of the camera and reduce power consumption.

Images