A hierarchical energy-saving control method and device, electronic equipment and storage medium

Abstract

Embodiments of the present application disclose a hierarchical energy-saving control method and device, electronic equipment and storage medium, which belong to the technical field of low-power embedded systems. The method comprises the following steps: dividing the power supply of a camera into a micro-power domain, an imaging domain, an inference domain and a radio frequency domain; only the micro-power domain is always on after power-on; collecting a sensor trigger signal, a battery state of charge value and a solar photovoltaic power; combining a return difference threshold to obtain a power consumption state permitted by the power; powering the camera after power detection when the sensor triggers; collecting images at a low frame rate and generating motion confirmation evidence through frame difference; powering the inference domain after power detection when the evidence is valid; performing target detection on the images to obtain a category and a confidence and generate target confirmation evidence; calculating the maximum allowed duty cycle and powering the radio frequency domain to control the average current of the radio frequency transmitter to be less than the upper limit after power detection when the evidence is valid. The present application solves the problems of high silent power consumption, high cost of false triggering and short weak energy compensation endurance in the prior art.

Description

Technical Field

[0001] This invention relates to the field of low-power embedded systems and power management technology, and in particular to a hierarchical energy-saving control method, device, electronic device and storage medium. Background Technology

[0002] With the widespread application of outdoor security and IoT sensing devices, mains-free cameras are becoming increasingly important in scenarios such as courtyards, perimeters, and agricultural and forestry areas. These devices typically rely on a combination of batteries and solar panels for power, and their average power consumption directly determines the battery capacity and the peak power of the solar panels, thus influencing material costs and installation size. However, existing mains-free cameras still face many challenges in power consumption control.

[0003] Traditional solutions typically employ a system-wide always-on approach, or power on the main control, imaging, inference, and communication modules simultaneously upon triggering. A system-wide always-on approach results in high static current during periods of inactivity, rapidly draining the battery. While waking up the entire module shortens the duration of high power consumption, it fails to truly reduce daily energy consumption due to the tendency for the communication module to become permanently connected and the high cost of accidental triggering. Although low-frame-rate continuous encoding schemes can maintain low-frame-rate imaging, their imaging links cannot be truly powered down, making it difficult to reduce silent power consumption to the milliwatt level.

[0004] Furthermore, existing methods suffer from a sharp reduction in battery life during cloudy or rainy weather with weak power supply, forcing developers to increase the capacity of solar panels and batteries. Additionally, if the peak current of communication transmission is not constrained, it can easily cause battery undervoltage reset, affecting system stability.

[0005] Therefore, there is an urgent need for a graded energy-saving control method that can effectively reduce silent power consumption, extend battery life, reduce solar panel specifications, and limit the average current of communication. Summary of the Invention

[0006] The embodiments of this invention provide a graded energy-saving control method to solve the problems of high silent power consumption, high cost of false triggering, short battery life due to weak power replenishment, and easy undervoltage caused by peak current in existing technologies. The technical solution is as follows: According to one aspect of the present invention, a graded energy-saving control method is provided, the method comprising: continuously acquiring state parameters of sensors through a low-power MCU in a normally open low-power domain; calculating remaining energy and predicted replenishment energy based on the state parameters; calculating budget tightness based on recent moving average power consumption and target range; and mapping the state parameters and hysteresis threshold to obtain a power consumption state permissible for the battery; the state parameters include a trigger signal, the state of charge value of a battery fuel gauge, and the photovoltaic power value and charging current value of a solar sampling circuit; when a sensor is triggered, detecting whether the current battery level can enter a listening state; if so, supplying power to a camera and an image processor, controlling the camera to acquire images at a set low frame rate and calculating frame difference or... Background subtraction yields motion amplitude. When the motion amplitude reaches a set requirement, motion confirmation evidence is generated. When the motion confirmation evidence is established, the current battery level is checked to see if the analysis state can be entered. If so, power is supplied to the inference domain, and the inference domain is controlled to infer the image to obtain the category of the detected target in the image and the corresponding confidence score. When the confidence score reaches a set requirement, target confirmation evidence is generated. When the target confirmation evidence is established, the current battery level is checked to see if the return state can be entered. If so, the maximum allowable duty cycle is calculated based on the power consumption of the currently powered non-RF domain, the remaining energy, and the predicted replenishment energy. After powering the RF domain, the average current of the RF transmission in the RF domain is controlled to not exceed the maximum allowable duty cycle.

[0007] In one embodiment, the remaining energy and predicted replenishment energy are calculated based on the state parameters, and the budget tension is calculated based on the recent moving average power consumption and the target range. The power consumption state that allows for power is obtained by combining the state parameters and the hysteresis threshold. This is achieved through the following steps: multiplying the state of charge value by a preset rated capacity value and coulomb efficiency coefficient to calculate the remaining energy; multiplying the photovoltaic power value or charging current value by the bus voltage value to obtain the instantaneous charging power; integrating and accumulating the instantaneous charging power within the predicted horizon using a sliding time window to obtain the predicted replenishment energy; calculating the average power consumption within the recent sliding time window to obtain the recent moving average power consumption; multiplying the recent moving average power consumption by the preset target range to obtain the total energy required for the range; dividing the total energy required for the range by the sum of the remaining energy and the predicted replenishment energy to obtain the budget tension; using a preset hysteresis mechanism based on the budget tension and the preset threshold range into which the state of charge value falls, determining the threshold to be used when switching power consumption states and mapping it to obtain the current power consumption state that allows for power; the power consumption states include a silent state, a listening state, an analysis state, and a feedback state.

[0008] In one embodiment, when the sensor is triggered, the current battery level is checked to determine if the system can enter a listening state. If so, power is supplied to the camera and image processor, and the camera is controlled to capture images at a set low frame rate. Frame differences or background differences are calculated on consecutive frames to obtain the motion amplitude. When the motion amplitude reaches a set requirement, motion confirmation evidence is generated through the following steps: When the sensor's trigger signal is valid, the current event level is set to the lowest level and compared with the power consumption state allowed by the battery. If the power consumption state meets the listening state requirements, power is supplied to the camera and image processor, and a preset stabilization delay is waited after power supply. After the stabilization delay ends, the camera is controlled to capture images at a set low frame rate. Pixel-by-pixel difference calculations are performed on the captured consecutive frames, or background difference calculations are performed with a preset background reference frame to obtain the motion amplitude, and the motion amplitude is continuously monitored to see if it exceeds a preset motion threshold. When the motion amplitude exceeds the motion threshold and the number of consecutive frames meeting the target reaches a preset requirement, motion confirmation evidence is generated. If the motion amplitude falls below the motion threshold, or the number of consecutive frames meeting the target is interrupted, or the sensor's trigger signal disappears, the power supply is cut off and the system returns to a silent state.

[0009] In one embodiment, when the motion confirmation evidence is established, it is detected whether the current power level allows entry into the analysis state. If so, power is supplied to the inference domain, and the inference domain is controlled to infer the image to obtain the category of the detected target in the image and the corresponding confidence score. When the confidence score reaches a set requirement, target confirmation evidence is generated through the following steps: when the motion confirmation evidence is established, the event level is raised from the lowest trigger level to the intermediate level and compared with the power consumption state allowed by the power level. If the power consumption state meets the analysis state requirements, power is supplied to the inference domain, and after power supply, a preset time is waited for... Readiness delay; after the readiness delay ends, the acquired image frames are transmitted to the inference domain. The inference domain loads a preset target detection model to perform inference operations on the image frames, and outputs the category labels and corresponding confidence scores of each detected target; the confidence scores are between 0 and 1; preset categories of interest are selected from the category labels. When the confidence score corresponding to the category of interest reaches a preset threshold and the number of consecutive hit frames reaches a preset requirement, target confirmation evidence is generated. Otherwise, the power supply to the inference domain, camera and image processor is cut off in reverse order, and the system returns to the listening state or silent state.

[0010] In one embodiment, when the target confirmation evidence is established, it is detected whether the current power level can enter the backhaul state. If so, the maximum allowable duty cycle is calculated based on the power consumption of the currently powered non-RF domain, the remaining energy, and the predicted replenishment energy through the following steps: When the target confirmation evidence is established, the event level is raised from the intermediate level to the highest level and compared with the power consumption state allowed by the power level. If the power consumption state meets the backhaul state requirements, the real-time power consumption of the powered non-RF domain is obtained as the base power consumption value; the remaining energy is added to the predicted replenishment energy to obtain the future available total energy; the future available total energy is divided by a preset time window and multiplied by a preset safety factor to obtain the upper limit of the allocable average power; the base power consumption value is deducted from the upper limit of the allocable average power to obtain the RF allocable power quota; the RF allocable power quota is divided by the system bus voltage to obtain the upper limit of the RF average current; the upper limit of the RF average current is divided by the RF transmit peak current and the smaller value between it and the maximum allowable duty cycle of the RF module is taken to obtain the maximum allowable duty cycle.

[0011] In one embodiment, controlling the average current of the radio frequency domain's radio frequency transmission to not exceed the maximum allowable duty cycle after powering the radio frequency domain is achieved through the following steps: powering the radio frequency domain and waiting for a preset network dwell time delay; after the network dwell time delay ends, periodically switching the power supply of the radio frequency domain on and off according to the heartbeat duty cycle or the feedback duty cycle, or issuing sleep and wake-up commands to the radio frequency domain to make the radio frequency domain alternately enter the working state and the sleep state; measuring the average current of the radio frequency domain's transmission using a sliding window; cutting off the power supply of the radio frequency domain when the average current exceeds the upper limit; and cutting off all power supply in reverse order and returning to the silent state when the feedback is completed, the event triggering condition disappears, or the power consumption state decreases.

[0012] In one embodiment, the method further includes the following steps: after cutting off all power supply, recording the breakpoint position of the currently incomplete backhaul data, and caching the incomplete backhaul data in the local storage medium, continuing to perform state parameter acquisition and budget update, and when the budget tension improves and the power consumption state rises back to the backhaul state, re-powering the RF domain to send the incomplete backhaul data; periodically waking up or the trigger signal of the sensor to wake up the low-power MCU to perform event processing operations, and after completing the processing operations, updating the acquired data to the internal register or low-power memory, re-entering the sleep state, and waiting for the next real-time clock interrupt to wake up again to perform the next acquisition operation.

[0013] According to one aspect of the present invention, a graded energy-saving control device is provided, the device comprising: a resident acquisition and budget mapping module, configured to continuously acquire sensor status parameters through a low-power MCU in a normally open low-power domain, calculate remaining energy and predicted replenishment energy based on the status parameters, calculate budget tension based on recent moving average power consumption and target range, and map the power consumption state to an allowable level by combining the status parameters and a hysteresis threshold; the status parameters include a trigger signal, a state of charge value of a battery fuel gauge, and a photovoltaic power value and charging current value of a solar sampling circuit; and an imaging power-on and motion confirmation module, configured to detect whether the current power level can enter a listening state when the sensor is triggered, and if so, to supply power to a camera and an image processor, control the camera to acquire images at a set low frame rate and calculate frame difference or backscatter for consecutive frames. The motion amplitude is obtained through scene difference analysis. When the motion amplitude reaches a set requirement, motion confirmation evidence is generated. The inference power-on and target confirmation module is used to detect whether the current power level can enter the analysis state when the motion confirmation evidence is established. If so, it supplies power to the inference domain and controls the inference domain to infer the image to obtain the category of the detected target in the image and the corresponding confidence score. When the confidence score reaches a set requirement, target confirmation evidence is generated. The radio frequency power-on and duty cycle current limiting module is used to detect whether the current power level can enter the feedback state when the target confirmation evidence is established. If so, it calculates the maximum allowable duty cycle based on the power consumption of the currently powered non-radio frequency domain, the remaining energy, and the predicted replenishment energy. After supplying power to the radio frequency domain, it controls the average current of the radio frequency transmission in the radio frequency domain to not exceed the maximum allowable duty cycle.

[0014] According to one aspect of the present invention, an electronic device includes at least one processor and at least one memory, wherein computer-readable instructions are stored on the memory; the computer-readable instructions are executed by one or more of the processors to cause the electronic device to implement the graded energy-saving control method as described above.

[0015] According to one aspect of the present invention, a storage medium has computer-readable instructions stored thereon, which are executed by one or more processors to implement the graded energy-saving control method as described above.

[0016] The beneficial effects of the technical solution provided by this invention are: In the above technical solution, this invention first divides the camera power supply into a normally-on low-power domain, an imaging domain, an inference domain, and a radio frequency domain. After power-on, only the normally-on low-power domain is enabled. A low-power MCU continuously collects sensor trigger signals, battery state of charge (SOC), and solar photovoltaic power and charging current values ​​to accurately assess the current available energy and budget constraints. Based on this, the power consumption state is obtained by mapping the SOC to the hysteresis threshold. When the sensor triggers, the power level is checked to see if it allows entering the listening state. If so, power is supplied to the camera and image processor to acquire images at a low frame rate and generate motion confirmation evidence through frame difference or background difference. When the motion confirmation evidence is established, the power level is checked to see if it allows entering the analysis state. If so, power is supplied to the inference domain to perform target detection on the image. The system infers the category and confidence score and generates target confirmation evidence. When the target confirmation evidence is valid, it checks whether the power level allows the system to enter the backhaul state. If so, it calculates the maximum allowable duty cycle based on the non-RF domain power consumption, remaining energy, and predicted power replenishment. After powering the RF domain, it controls the average current of RF transmission to not exceed this upper limit according to the duty cycle. When the backhaul is completed or the conditions deteriorate, it reverses the power supply and returns to the silent state. This achieves domain-based, stepped management of the power consumption of cameras without mains power. It not only significantly reduces silent power consumption but also effectively extends the battery life and reduces the size of the solar panel. This allows the device to complete monitoring tasks more stably and continuously in weak power replenishment scenarios, thus effectively solving the problems of high silent power consumption, high cost of false triggering, short battery life in weak power replenishment, and easy undervoltage caused by peak current in existing technologies. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a graded energy-saving control method according to an exemplary embodiment; Figure 2 This is a block diagram illustrating a graded energy-saving control device according to an exemplary embodiment; Figure 3 This is a hardware structure diagram of an electronic device according to an exemplary embodiment; Figure 4 This is a block diagram illustrating an electronic device according to an exemplary embodiment. Detailed Implementation

[0019] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0020] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this disclosure means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0021] This invention provides a hierarchical energy-saving control method. Through tiered power supply across power domains, dual-trigger constraints (event and power consumption), and RF duty cycle current limiting, it achieves a significant reduction in silent power consumption and a substantial extension of battery life. This solves the problems of high silent power consumption, short battery life with weak power replenishment, and undervoltage caused by peak current in existing technologies. This hierarchical energy-saving control method is applicable to hierarchical energy-saving control devices, which can be electronic devices. The hierarchical energy-saving control method in this invention can be applied to various scenarios, such as low-power control of power domains for cameras without mains power.

[0022] Please see Figure 1 This invention provides a graded energy-saving control method applicable to electronic devices.

[0023] In the following method embodiments, for ease of description, the execution subject of each step of the method is an electronic device, but this does not constitute a specific limitation.

[0024] like Figure 1 As shown, the method may include the following steps: Step 110: The sensor's state parameters are continuously collected by a low-power MCU in the normally open low-power domain. The remaining energy and predicted replenishment energy are calculated based on the state parameters. The budget tightness is calculated based on the recent moving average power consumption and the target range. The power consumption state that allows the power is obtained by combining the state parameters and the hysteresis threshold.

[0025] In one possible implementation, the remaining energy is calculated by multiplying the state of charge value by the preset rated capacity value and the coulomb efficiency coefficient. The instantaneous charging power is obtained by multiplying the photovoltaic power value or charging current value by the bus voltage value. The instantaneous charging power is then integrated and accumulated within the predicted horizon using a sliding time window to obtain the predicted replenishment energy.

[0026] In one possible implementation, the average power consumption within the recent sliding time window is calculated to obtain the recent sliding average power consumption. Multiplying this by the preset target range time yields the total energy required for the range. The total energy required for the range is then divided by the sum of the remaining energy and the predicted replenishment energy to obtain the budget tension. A preset hysteresis mechanism is used to determine the threshold to be used when switching power consumption states based on the budget tension and the preset threshold range in which the state of charge value falls, and this threshold is mapped to obtain the power consumption state that is currently allowed by the current power level.

[0027] The state parameters include trigger signals, the state of charge value of the battery fuel gauge, and the photovoltaic power value and charging current value of the solar sampling circuit, etc. The power consumption states include silent state, listening state, analysis state, and feedback state, etc., which are not limited here.

[0028] Specifically, the low-power MCU operates in the always-on low-power domain, maintaining uninterrupted operation with milliwatt-level power consumption. It continuously monitors the trigger signal of the PIR vibration sensor using a combination of interrupts and polling. Simultaneously, it reads the battery state-of-charge (SOC) value via the fuel gauge interface at a set sampling period, and reads the photovoltaic power and charging current values ​​of the solar sampling circuit via the analog-to-digital converter interface. These three types of data respectively represent "whether an event has occurred," "how much power remains," and "how much more power can be added," forming the data foundation for all subsequent decisions.

[0029] Furthermore, the MCU multiplies the state of charge (SOC) value by the battery's rated capacity and coulombic efficiency coefficient to obtain the current actual available remaining energy. Simultaneously, it multiplies the photovoltaic power value by the bus voltage value, or directly multiplies the charging current value by the bus voltage value, to obtain the instantaneous charging power. This instantaneous charging power is then integrated and accumulated within a preset prediction horizon over a sliding time window to estimate the energy that solar energy can replenish in the future, i.e., the predicted replenishment energy. Further, the MCU calculates the average power consumption within the recent sliding time window, multiplies it by the preset target range, and obtains the total energy required to maintain the target range. This total energy is divided by the sum of the remaining energy and the predicted replenishment energy to obtain the budget tension. Subsequently, a hysteresis mechanism is used to determine different thresholds for entering and exiting various power consumption states based on the budget tension range and the threshold range within which the SOC value falls, mapping and outputting the power consumption states permitted by the current power level, including silent state, listening state, analysis state, and feedback state.

[0030] In the above process, the embodiments of the present invention continuously collect trigger signals, state of charge and photovoltaic power through the normally open low power domain and continuously calculate the budget tension, so that the system can monitor the energy status in real time at the milliwatt level during the event-free period, providing a dynamic mapping basis for the power allowable power state, and achieving a significant reduction in silent power consumption and a significant improvement in battery life.

[0031] Step 120: When the sensor is triggered, the current power level is checked to see if it can enter the listening state. If it can, power is supplied to the camera and image processor, and the camera is controlled to acquire images at a set low frame rate and calculate the frame difference or background difference of consecutive frames to obtain the motion amplitude. When the motion amplitude reaches the set requirements, motion confirmation evidence is generated.

[0032] In one possible implementation, when the sensor's trigger signal is valid, the current event level is set to the lowest level and compared with the power consumption state that is permissible. If the power consumption state meets the requirements of the listening state, power is supplied to the camera and image processor, and a preset stabilization delay is waited after power supply. After the stabilization delay ends, the camera is controlled to acquire images at a set low frame rate. The motion amplitude is obtained by performing pixel-by-pixel difference calculation on the acquired continuous frames or by performing background difference calculation with a preset background reference frame, and the motion amplitude is continuously monitored to see if it exceeds a preset motion threshold.

[0033] In one possible implementation, when the motion amplitude exceeds the motion threshold and the number of consecutive frames meeting the preset requirement is reached, motion confirmation evidence is generated. If the motion amplitude falls back below the motion threshold, or the number of consecutive frames meeting the requirement is interrupted, or the sensor's trigger signal disappears, the power supply is cut off and the system returns to a silent state.

[0034] Specifically, when the PIR vibration sensor trigger signal is valid, the MCU sets the current event level to the lowest trigger level and compares it with the mapped power allowance state. If the power allowance state reaches or exceeds the listening state level, it determines that entering the listening state is allowed; if it is lower than the listening state level, it refuses to power on, the system remains silent and continues to collect data and update the budget, avoiding the waste of valuable energy due to false triggers when the battery is low.

[0035] Furthermore, after the power pre-check passes, the MCU enables the power supply rail of the imaging ISP domain via the load switch of the power management chip, supplying power to the image sensor and image signal processor. Since the device requires startup time to stabilize and output a valid image after power-off, the MCU waits for a preset stabilization delay before controlling the camera to start acquiring images at a low frame rate. Low frame rate acquisition is only used to obtain reference frames required for motion detection, and its power consumption is far lower than the normal recording frame rate.

[0036] Furthermore, the MCU performs pixel-by-pixel differential calculations on the acquired consecutive frames, or performs background differential calculations with a preset background reference frame, to obtain the motion amplitude of each region in the image. To avoid misjudgment caused by single-frame noise, the MCU requires the motion amplitude to exceed a preset motion threshold and the number of consecutive qualifying frames to reach a preset requirement before generating pixel-level motion confirmation evidence. If the motion amplitude falls below the threshold, the number of consecutive qualifying frames is interrupted, or the trigger signal disappears, the power supply to the imaging ISP domain is immediately cut off and the image returns to a silent state.

[0037] In the above process, the embodiments of the present invention check the power level before powering on after triggering, and use low frame acquisition combined with frame difference or background difference to confirm motion, so that PIR false triggering will not directly wake up high-power modules, providing the ability to complete motion authenticity verification at extremely low cost, and realizing controllable false triggering cost and significant compression of power consumption in the listening stage.

[0038] Step 130: When motion confirmation evidence is established, check whether the current power level can enter the analysis state. If so, power is supplied to the inference domain, and the inference domain is controlled to infer the image to obtain the category of the detected target in the image and the corresponding confidence score. When the confidence score reaches the set requirement, target confirmation evidence is generated.

[0039] In one possible implementation, when motion confirmation evidence is established, the event level is raised from the lowest trigger level to an intermediate level and compared with the power consumption state that is permissible. If the power consumption state meets the analysis state requirements, the inference domain is powered and a preset ready delay is waited for after power supply. After the ready delay ends, the acquired image frame is transmitted to the inference domain. The inference domain loads a preset target detection model to perform inference operations on the image frame and outputs the category label of each detected target and the corresponding confidence score. The confidence score is between 0 and 1.

[0040] In one possible implementation, preset categories of interest are selected from the category labels. When the confidence score corresponding to the category of interest reaches a preset threshold and the number of consecutive hit frames reaches a preset requirement, target confirmation evidence is generated. Otherwise, the power supply to the inference domain, camera, and image processor is cut off in reverse order, and the system returns to the listening state or silent state.

[0041] Specifically, once pixel-level motion confirmation evidence is established, the MCU raises the event level from the lowest trigger level to an intermediate level and compares it with the power-permitted state. If the power-permitted state reaches or exceeds the analysis state level, it determines that entry into the analysis state is allowed; if it is lower than the analysis state level, the system remains in the listening state to continue data acquisition, or falls back to the silent state after a period of no events, ensuring that the inference domain is not woken up when the power is insufficient.

[0042] Furthermore, after the power pre-check passes, the MCU enables the power supply rail of the inference domain via the load switch to supply power to the NPU or high-performance processor core, and waits for a preset ready delay. Once the inference domain is ready, the MCU transmits the image frames acquired and buffered during the listening phase to the inference domain. The inference domain loads a preset target detection model to perform inference operations on the image frames, outputting the category label and corresponding confidence score for each detected target. The confidence score is between 0 and 1, representing the model's degree of certainty regarding the judgment result.

[0043] Furthermore, the MCU filters preset categories of interest from the inference results, focusing only on target types relevant to the monitoring requirements. When the confidence score corresponding to the category of interest reaches a preset threshold and the number of consecutive hit frames reaches a preset requirement, target confirmation evidence is generated. If the confidence score does not meet the standard, the number of consecutive hit frames is interrupted, or the inference results do not contain the category of interest, the power supply to the inference domain is cut off in reverse order, followed by the power supply to the imaging ISP domain, returning to the listening state or silent state.

[0044] In the above process, the embodiments of the present invention first check the power level after motion evidence is established and then power on the inference domain, and use the target detection model to confirm the category and confidence of the cached frame. This ensures that the inference domain is only awakened when there is actual motion and the power level allows, providing a secondary confirmation mechanism between pixel motion and semantic target, and effectively suppressing the power consumption of invalid inference caused by false alarms.

[0045] Step 140: When the target confirmation evidence is established, check whether the current power level can enter the backhaul state. If it can, calculate the maximum allowable duty cycle based on the power consumption of the currently powered non-RF domain, the remaining energy, and the predicted replenishment energy. After powering the RF domain, control the average current of the RF transmission in the RF domain to not exceed the maximum allowable duty cycle.

[0046] In one possible implementation, when the evidence confirms the target, the event level is raised from the intermediate level to the highest level and compared with the power consumption state that is allowed by the power supply. If the power consumption state meets the feedback state requirements, the real-time power consumption of the powered-on non-RF domain is obtained as the base power consumption value. The remaining energy is added to the predicted replenishment energy to obtain the total future available energy. The total future available energy is divided by a preset time window and multiplied by a preset safety factor to obtain the upper limit of the allocable average power. The base power consumption value is deducted from the upper limit of the allocable average power to obtain the RF allocable power quota.

[0047] In one possible implementation, the upper limit of the average RF current is obtained by dividing the RF allocable power limit by the system bus voltage, and the upper limit of the average RF current is then divided by the RF transmit peak current and the smaller value is taken between this upper limit and the maximum allowable duty cycle of the RF module to obtain the maximum allowable duty cycle.

[0048] In one possible implementation, power is supplied to the RF domain and a preset network dwell time is waited for. After the network dwell time ends, the power supply to the RF domain is periodically switched on and off according to the heartbeat duty cycle or the feedback duty cycle, or sleep and wake-up commands are issued to the RF domain to make the RF domain alternately enter the working state and sleep state. The average current transmitted by the RF domain is measured by a sliding window. When the average current exceeds the upper limit, the power supply to the RF domain is cut off. When the feedback is completed, or the event triggering condition disappears, or the power consumption state decreases, all power supplies are cut off in reverse order and the RF domain returns to the silent state.

[0049] In one possible implementation, after cutting off all power, the location of the breakpoint of the currently incomplete data transmission is recorded, and the incomplete data transmission is cached in the local storage medium. Status parameter acquisition and budget update continue. When the budget tension improves and the power consumption returns to the transmission state, the RF domain is powered on again to send the incomplete data transmission. Periodic wake-up or sensor trigger signal wakes up the low-power MCU to perform event processing operations. After completing the processing operation, the acquired data is updated in the internal register or low-power memory, and the MCU re-enters the sleep state, waiting for the next real-time clock interrupt to wake up and perform the next acquisition operation.

[0050] Specifically, once the target confirmation evidence is established, the MCU raises the event level from the middle level to the highest level and compares it with the power-permitted state. If the power-permitted state reaches or exceeds the return state level, it is determined that entering the return state is allowed; if it is lower than the return state level, only the image data is stored locally without radio frequency transmission, and processing is carried out after the power conditions improve, so as to avoid the peak current of radio frequency transmission causing battery undervoltage when the power is insufficient.

[0051] Furthermore, after the power pre-detection passes, the MCU obtains the real-time power consumption of the currently powered non-RF domain as the base power consumption value. The remaining energy is added to the predicted replenishment energy to obtain the total future available energy. This is then divided by a preset time window and multiplied by a preset safety factor to obtain the upper limit of the allocable average power. The base power consumption value is subtracted from this upper limit to obtain the RF allocable power allowance, which is then divided by the system bus voltage to obtain the upper limit of the RF average current. This upper limit is divided by the RF transmit peak current and the smaller value is taken between this upper limit and the maximum allowable duty cycle of the RF module to obtain the maximum allowable duty cycle.

[0052] Furthermore, the MCU enables the RF domain power supply rail via a load switch and waits for a preset network setup delay. After network setup is complete, the MCU periodically switches the RF domain power supply on and off according to the heartbeat duty cycle or the feedback duty cycle, or issues sleep and wake-up commands to the RF domain, causing the RF domain to alternate between working and sleep states, thereby limiting the average current of RF transmission to within the average current value corresponding to the maximum allowable duty cycle. Simultaneously, a sliding window measurement is performed on the average current; if the upper limit is exceeded, the power supply is immediately cut off to perform hard limiting.

[0053] Furthermore, when the backhaul is completed, the event triggering condition disappears, or the permitted power consumption state decreases due to energy degradation, the MCU sequentially cuts off the power supply to the RF domain, inference domain, and imaging ISP domain in reverse order, and the system returns to a silent state where only the normally-on low-power domain operates. If the backhaul is cut off before completion, the breakpoint location is recorded and the incomplete data is cached to the local storage medium. When the budget constraints improve and the permitted power consumption state returns to the backhaul state, the RF domain is powered back on and transmission resumes from the breakpoint location.

[0054] In the above process, the embodiments of the present invention calculate the duty cycle before transmission, convert the peak current into a controllable average current using periodic on / off or sleep wake-up commands, hard limit the shutdown when the limit is exceeded and gradually decrease it in reverse order, so that radio frequency transmission no longer becomes the cause of battery undervoltage, providing the budgeted closed-loop control capability of radio frequency average current, and realizing the balance between reliable data transmission and safe battery operation under weak power conditions.

[0055] Through the above process, this embodiment of the invention divides the camera power supply into a normally-on low-power domain, an imaging domain, an inference domain, and a radio frequency domain. After power-on, only the normally-on low-power domain is enabled. The low-power MCU continuously collects sensor trigger signals, battery state of charge values, and solar photovoltaic power and charging current values ​​to calculate remaining energy, predict replenishment, and budget tightness. Combined with hysteresis threshold mapping, the permissible power consumption state is obtained. When the sensor is triggered, the imaging domain is powered after power pre-detection, and images are acquired at a low frame rate. Motion confirmation evidence is generated through frame difference or background difference. After the motion evidence is established, the inference domain is powered after power pre-detection, and target detection inference is performed on the image to obtain the category and confidence score and generate target confirmation evidence. After the target evidence is established, the maximum permissible duty cycle is calculated after power pre-detection, and the radio frequency domain is powered. The average current of radio frequency transmission is controlled according to the duty cycle to not exceed the upper limit. When the backhaul is completed or the conditions deteriorate, the power supply is cut off in reverse order and the system returns to a silent state. Uncompleted data is buffered and will be transmitted after the conditions improve. This invention enables tiered management of power consumption for cameras without mains power, which not only significantly reduces silent power consumption but also effectively extends battery life and reduces the size of the solar panel. This allows the device to complete monitoring tasks more stably and continuously in weak power supply scenarios, thereby effectively solving the problems of high silent power consumption, high cost of false triggering, short battery life under weak power supply, and easy undervoltage caused by peak current in existing technologies.

[0056] In an exemplary embodiment, the hierarchical energy-saving control method of the present invention is used to perform low-power control of power domains in an outdoor courtyard scene where a camera without mains power is installed. The power supply system consists of a solar panel with a peak power of 5W, a lithium battery with a capacity of 10000mAh, and a corresponding charging management circuit. The camera's main control board is equipped with four independently controllable power domains: normally open low-power domain D0, imaging ISP domain D1, AI inference domain D2, and radio frequency domain D3.

[0057] The D0 domain contains a low-power MCU (such as the STM32U5 series), a real-time clock chip, a PIR vibration sensor, a fuel gauge interface circuit, and drive circuits for various load switches. The D0 domain is directly powered by the battery through a low-dropout linear regulator, without any load switches, ensuring it is always powered. The MCU's quiescent current in sleep mode is approximately 2 kW. The real-time clock running current is approximately 1 The standby current of the PIR sensor is approximately 5 kW. The standby current of the power meter is approximately 1. The total static current in the D0 domain is approximately 9 At a bus voltage of 3.7V, the corresponding power consumption is approximately 33. With the power management chip's own static power consumption, the overall system's silent power consumption is controlled between 1 and 20 kWh. Within the range.

[0058] Domain D1 contains the image sensor (CIS), image signal processor (ISP), and optional IR-CUT switch and fill light sub-switch, with its power rail connected to the battery via a first load switch. Domain D2 contains the NPU or main control high-performance core, with its power rail connected to the battery via a second load switch. Domain D3 contains the Wi-Fi or cellular module and power amplifier, with its power rail connected to the battery via a third load switch. The enable pins of the three load switches are driven by the general purpose input / output ports of the MCU in domain D0. The MCU controls the power-on / off of each domain by pulling the corresponding pin high or low.

[0059] The solar sampling circuit includes a voltage divider resistor network and an analog-to-digital converter interface. The MCU reads the solar panel output voltage and charging current through this interface. The fuel gauge chip communicates with the MCU via an I2C interface, reporting the battery state-of-charge (SOC) value and the bus voltage value. The PIR sensor outputs a digital trigger signal, which is connected to the MCU's external interrupt pin. Simultaneously, the MCU also periodically polls the status of this pin using an internal timer for redundant detection.

[0060] After the system powers on, the MCU first executes the initialization program, configuring the polarity and power-on timing parameters of the load switches for each power domain from D0 to D3, and setting the enable pins of all three load switches to low level to ensure that only the D0 domain is powered. Subsequently, the MCU configures the real-time clock interrupt cycle, PIR external interrupt trigger mode, fuel gauge communication parameters, and analog-to-digital conversion sampling channel, and enters a silent state after completing the initialization.

[0061] In silent mode, the MCU spends most of its time in deep sleep mode, only waking up once at a set period (e.g., every 30 seconds) by the real-time clock to perform the following operations: read the state of charge value and bus voltage value reported by the fuel gauge through the I2C interface; read the output voltage and charging current value of the solar panel through the analog-to-digital converter interface; write the collected data into the sliding window buffer in the internal low-power memory; and then re-enter sleep mode to wait for the next real-time clock interrupt.

[0062] When the PIR sensor detects an infrared change and outputs a valid trigger signal, the MCU is woken up by an external interrupt, immediately records the time of the event, and sets the current event level to the lowest trigger level. At this time, the MCU does not immediately perform a power-on operation; instead, it first reads the most recently calculated power-permitted state and compares it with the required level for the listening state.

[0063] After each real-time clock wake-up to acquire data, or after a PIR-triggered wake-up, the MCU performs a budget tightness calculation and power-permitted power state mapping. The specific process is as follows: First, the MCU reads the State of Charge (SOC) value reported by the fuel gauge, multiplies it by the energy value E_nom corresponding to the battery's rated capacity and the coulombic efficiency coefficient η, and calculates the remaining energy E_rem = η·E_nom·SOC. For example, if the battery's rated energy is 37Wh, the coulombic efficiency is 0.98, and the current SOC is 60%, then the remaining energy is approximately 21.76Wh.

[0064] Secondly, the MCU reads the photovoltaic power or charging current values ​​collected in the recent period from the sliding window buffer, multiplies the charging current value by the bus voltage value to obtain the instantaneous charging power, and then integrates and accumulates the instantaneous charging power over the predicted horizon T_h (e.g., 24 hours) using a sliding time window to obtain the predicted energy replenishment Ê_pv. For example, if the average charging power over the past 24 hours is 0.5W, then the predicted energy replenishment is approximately 12Wh.

[0065] Furthermore, the MCU calculates the average power consumption within the recent sliding time window. Multiply this by the target battery life T_target (e.g., 48 hours) to get the total energy required for the remaining battery life. For example, if the recent average power consumption is 0.8W and the target battery life is 48 hours, then the total energy required is 38.4Wh. Divide this total energy by the sum of the remaining energy and the predicted replenishment energy (21.76Wh + 12Wh = 33.76Wh), and you get the budget tightness ρ, which is approximately 1.14.

[0066] The MCU maps the allowable power consumption state S_max based on the numerical range of the budget tension ρ and the preset threshold range into which the current state of charge (SOC) falls. In this embodiment, the following settings are defined: if ρ≤1 and SOC≥50%, then S_max=S3 (feedback state); if 1<ρ≤1.5 or 20%≤SOC<50%, then S_max=S2 (analysis state); if ρ>1.5 or SOC<20%, then S_max=S1 (listening state) or forced S0 (silent state). Simultaneously, a hysteresis mechanism is used for SOC: entering the low-allowance zone uses SOC_lo=20%, and exiting the low-allowance zone uses SOC_lo+Δ_hys=25%, to prevent frequent jumps in the allowable power consumption state when the SOC value fluctuates slightly near the threshold.

[0067] When the PIR trigger signal is valid and the power consumption state S_max ≥ S1 is enabled, the MCU determines that it is allowed to enter the listening state. The MCU enables the power supply rail of the imaging ISP domain D1 by pulling the enable pin of the first load switch high, thereby powering the image sensor and image signal processor. After power is supplied, the MCU starts the first timer and waits for a preset stabilization delay T_d1 (e.g., 100ms) to ensure that the image sensor and ISP complete initialization and output a stable image.

[0068] After the stabilization delay ends, the MCU sends a low frame rate acquisition configuration to the image sensor via the I2C interface, controlling it to start acquiring images at a set low frame rate (e.g., 2 frames per second). The acquired image frames are processed by the ISP and stored in the MCU's internal buffer. The MCU performs pixel-by-pixel difference calculations on two consecutive frames to obtain the brightness change value of each pixel, and then statistically analyzes the difference values ​​across the entire image to obtain the motion amplitude. Alternatively, the MCU can pre-acquire and store a background reference frame during a period without events, and then perform background difference calculations between the current frame and the background reference frame to obtain the motion amplitude in the same way.

[0069] The MCU continuously monitors whether the motion amplitude exceeds a preset motion threshold θ_pix and requires a continuous number of frames meeting the threshold to reach N_pix frames (e.g., 3 frames). If the condition is met, pixel-level motion confirmation evidence e_pix is ​​generated. If the motion amplitude falls below the threshold, the continuous number of frames meeting the threshold is interrupted, or the PIR trigger signal disappears for more than a preset time window, the MCU determines that the motion is invalid, cuts off the power supply to the D1 domain by pulling down the first load switch enable pin, and the system falls back to a silent state.

[0070] When pixel-level motion confirmation evidence e_pix is ​​established, the MCU raises the event level from the lowest trigger level to an intermediate level and reads the power allowance state S_max again for comparison. If S_max ≥ S2, it determines that entering the analysis state is allowed. The MCU enables the power supply rail of the AI ​​inference domain D2 by pulling high the enable pin of the second load switch, thereby powering the NPU or high-performance processor core, and starts the second timer to wait for the preset ready delay T_d2 (e.g., 80ms).

[0071] After the readiness delay ends, the MCU transmits the image frames buffered during the listening phase to the NPU via a high-speed communication interface. The NPU loads a preset target detection model (such as a detection model based on a convolutional neural network), performs inference operations on the image frames, and outputs the category labels and corresponding confidence scores for each detected target. In this embodiment, the preset category set includes three categories: people, vehicles, and animals, which can be set according to the actual application scenario. The confidence scores are between 0 and 1.

[0072] The MCU filters preset categories of interest (e.g., people and vehicles) from the inference results, extracts the corresponding confidence scores, and checks whether they reach a preset category confidence threshold θ_c (e.g., 0.7). Simultaneously, it requires a consecutive hit count of N_ai frames (e.g., 2 frames). If the conditions are met, AI target confirmation evidence e_ai is generated. If not, the MCU reverses the process by first pulling down the second load switch enable pin to cut off power to the D2 domain, then pulling down the first load switch enable pin to cut off power to the D1 domain, causing the system to fall back to a listening or silent state.

[0073] When the AI ​​target confirmation evidence e_ai is established, the MCU raises the event level from the middle level to the highest level and reads the power allowance state S_max again for comparison. If S_max ≥ S3, it is determined that entering the feedback state is allowed. The MCU first obtains the real-time power consumption of the currently powered non-RF domains (D1 domain and D2 domain) as the base power consumption value P_base.

[0074] Further, the MCU adds the remaining energy E_rem to the predicted replenishment energy Ê_pv to obtain the total future available energy, divides it by a preset time window T_h, and multiplies it by a preset safety factor α (e.g., 0.8) to obtain the upper limit of the allocable average power P_allow. The base power consumption value P_base is subtracted from P_allow to obtain the RF allocable power allowance. This allowance is divided by the system bus voltage V_bus to obtain the upper limit of the RF average current I_cap. I_cap is divided by the RF transmit peak current I_pk (e.g., 1.5A), and the smaller value is taken with the maximum allowable duty cycle δ_hw of the RF module (e.g., 20%) to obtain the maximum allowable duty cycle δ_max.

[0075] The MCU enables the power supply rail of the RF domain D3 by pulling the enable pin of the third load switch high, thus powering the Wi-Fi or cellular module and starting the third timer to wait for the preset network setup delay T_d3 (e.g., 3 seconds). After network setup ends, the MCU selects either the heartbeat duty cycle δ_hb (e.g., 1%) or the feedback duty cycle δ_ul (e.g., 10%) based on the current service scenario, neither of which exceeds δ_max. The MCU drives the enable pin of the third load switch through the timer's PWM waveform output, causing the RF domain to alternately turn on and off according to the duty cycle δ within the period T_rf, or sends sleep and wake-up commands to the module through the communication interface, causing the module to alternately enter working and sleep states.

[0076] Meanwhile, the MCU performs a sliding window measurement on the average current of the RF transmitter. Once the average current exceeds I_cap, the MCU immediately pulls down the third load switch enable pin to perform a hard-limit shutdown, avoiding software delays that could cause a battery undervoltage reset.

[0077] When the backhaul is complete, or the event trigger condition disappears beyond the fallback delay T_off (e.g., 10 seconds), or the power consumption state decreases due to worsening budget constraints, the MCU triggers the fallback process. The MCU pulls the third load switch enable pin low in reverse order to cut off the power supply to the D3 domain, pulls the second load switch enable pin low to cut off the power supply to the D2 domain, and pulls the first load switch enable pin low to cut off the power supply to the D1 domain. The system returns to a silent state where only the D0 domain is active.

[0078] If the data transmission is interrupted before completion, the MCU records the breakpoint of the incomplete data transmission and caches the incomplete data to local storage media (such as flash memory or a secure digital card). The MCU continues to collect status parameters and update the budget. When the budget pressure improves and the power consumption state returns to the transmission state, the MCU re-enables the D3 domain power supply and waits for the network latency before resuming transmission of the incomplete data from the breakpoint.

[0079] In silent mode, the MCU is periodically woken up by a real-time clock interrupt to perform data acquisition. After acquisition, the data is updated to internal registers or low-power memory, and then the MCU returns to sleep mode. When the PIR trigger signal is valid, the MCU is woken up by an external interrupt, immediately responds, and executes the event handling process. After the process is completed, the MCU returns to sleep mode to wait for the next trigger. This sleep and wake-up mechanism ensures that the MCU is always in a minimum power consumption state during periods without events, and is only briefly woken up when necessary to perform data acquisition or event handling, thereby keeping the overall power consumption in silent mode at the milliwatt level.

[0080] This embodiment divides the camera power supply into a normally-on low-power domain, an imaging ISP domain, an AI inference domain, and a radio frequency (RF) domain. After power-on, only the normally-on low-power domain is enabled. A low-power MCU continuously collects PIR trigger signals, battery state of charge values, and solar photovoltaic power and charging current values ​​to calculate remaining energy, predict replenishment, and budget constraints. Combined with a hysteresis threshold mapping, the permissible power consumption state is obtained. When PIR is triggered, power is supplied to the imaging domain after power pre-check, and images are acquired at a low frame rate. Motion confirmation evidence is generated through frame difference or background difference. After motion evidence is established, power is supplied to the inference domain after power pre-check, and target detection inference is performed on the image to obtain the category and confidence score and generate target confirmation evidence. After target evidence is established, the maximum permissible duty cycle is calculated after power pre-check, and power is supplied to the RF domain. The average RF transmission current is controlled according to the duty cycle to not exceed the upper limit. When the backhaul is completed or the conditions deteriorate, the power supply is cut off in reverse order and the system returns to a silent state. Uncompleted data is buffered and will be transmitted again when conditions improve. This embodiment achieves domain-based stepped power management, reducing silent power consumption to the milliwatt level, making the cost of false triggering controllable, effectively suppressing RF peak current, maintaining target battery life under weak power replenishment and rainy conditions, and significantly reducing the specification requirements of solar panels and batteries.

[0081] The following are embodiments of the apparatus of the present invention, which can be used to execute the graded energy-saving control method involved in the present invention. For details not disclosed in the embodiments of the apparatus of the present invention, please refer to the method embodiments of the graded energy-saving control method involved in the present invention.

[0082] Please see Figure 2 This invention provides a graded energy-saving control device 800.

[0083] The graded energy-saving control device 800 includes, but is not limited to: a resident acquisition and budget mapping module 810, an imaging power-on and motion confirmation module 830, an inference power-on and target confirmation module 850, and a radio frequency power-on and duty cycle current limiting module 870.

[0084] Among them, the resident acquisition and budget mapping module 810 is used to continuously acquire the sensor's status parameters through a low-power MCU in the normally open low-power domain, calculate the remaining energy and predict the replenishment energy based on the status parameters, calculate the budget tension based on the recent moving average power consumption and the target range, and obtain the power consumption state that allows the power supply by combining the status parameters and the hysteresis threshold. The status parameters include the trigger signal, the state of charge value of the battery fuel gauge, and the photovoltaic power value and charging current value of the solar sampling circuit.

[0085] The imaging power-on and motion confirmation module 830 is used to detect whether the current power level can enter the listening state when the sensor is triggered. If so, it supplies power to the camera and image processor, controls the camera to acquire images at a set low frame rate, and calculates the frame difference or background difference of consecutive frames to obtain the motion amplitude. When the motion amplitude reaches the set requirements, motion confirmation evidence is generated.

[0086] The inference power-on and target confirmation module 850 is used to detect whether the current power level can enter the analysis state when motion confirmation evidence is established. If it can, it supplies power to the inference domain and controls the inference domain to infer the image to obtain the category of the detected target in the image and the corresponding confidence score. When the confidence score reaches the set requirements, it generates target confirmation evidence.

[0087] The RF power-on and duty cycle current limiting module 870 is used to detect whether the current power level can enter the feedback state when the target confirmation evidence is established. If it can, it calculates the maximum allowable duty cycle based on the power consumption of the currently powered non-RF domain, the remaining energy and the predicted replenishment energy. After powering the RF domain, it controls the average current of the RF transmission of the RF domain to not exceed the maximum allowable duty cycle.

[0088] It should be noted that the hierarchical energy-saving control provided in the above embodiments is only an example of the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed. That is, the internal structure of the hierarchical energy-saving control device will be divided into different functional modules to complete all or part of the functions described above.

[0089] Furthermore, the embodiments of the graded energy-saving control device and the graded energy-saving control method provided in the above embodiments belong to the same concept, and the specific way in which each module performs its operation has been described in detail in the method embodiments, and will not be repeated here.

[0090] Figure 3 A schematic diagram of the structure of an electronic device according to an exemplary embodiment is shown.

[0091] It should be noted that this electronic device is merely an example adapted to the present invention and should not be construed as providing any limitation on the scope of use of the present invention. Furthermore, this electronic device should not be interpreted as requiring or depending on having... Figure 3 One or more components of the exemplary electronic device 2000 shown.

[0092] The hardware structure of electronic devices 2000 can vary significantly due to differences in configuration or performance, such as... Figure 3 As shown, the electronic device 2000 includes: a power supply 210, an interface 230, at least one memory 250, and at least one central processing unit (CPU) 270.

[0093] Specifically, power supply 210 is used to provide operating voltage for various hardware devices on electronic device 2000.

[0094] Interface 230 includes at least one wired or wireless network interface 231 for interacting with external devices. Of course, in other examples adapted to this invention, interface 230 may further include at least one serial-to-parallel conversion interface 233, at least one input / output interface 235, and at least one USB interface 237, etc. Figure 3 As shown, this does not constitute a specific limitation.

[0095] The memory 250 serves as a carrier for resource storage and can be a read-only memory, random access memory, disk, or optical disk, etc. The resources stored on it include the operating system 251, application programs 253, and data 255, etc., and the storage method can be temporary storage or permanent storage.

[0096] The operating system 251 is used to manage and control the various hardware devices and application programs 253 on the electronic device 2000, so as to enable the central processing unit 270 to perform calculations and processing on the massive data 255 in the memory 250. It can be Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.

[0097] Application 253 is a computer-readable instruction based on operating system 251 that performs at least one specific task, and may include at least one module ( Figure 3 (Not shown), each module may contain computer-readable instructions for the electronic device 2000. For example, the graded energy-saving control device can be considered as an application program 253 deployed on the electronic device 2000.

[0098] Data 255 may be signal information, etc., and is stored in memory 250.

[0099] The central processing unit 270 may include one or more processors and is configured to communicate with the memory 250 via at least one communication bus to read computer-readable instructions stored in the memory 250, thereby performing operations and processing on the massive amounts of data 255 stored in the memory 250. For example, a graded energy-saving control method can be implemented by the central processing unit 270 reading a series of computer-readable instructions stored in the memory 250.

[0100] Furthermore, the present invention can also be implemented through hardware circuits or a combination of hardware circuits and software. Therefore, the implementation of the present invention is not limited to any specific hardware circuit, software, or combination thereof.

[0101] Please see Figure 4 This invention provides an electronic device 4000, which may include: a desktop computer, a laptop computer, a server, etc., with sensor recognition capabilities.

[0102] exist Figure 4 In this context, the electronic device 4000 includes at least one processor 4001 and at least one memory 4003.

[0103] The data interaction between the processor 4001 and the memory 4003 can be achieved through at least one communication bus 4002. This communication bus 4002 may include a path for transmitting data between the processor 4001 and the memory 4003. The communication bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The communication bus 4002 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0104] Optionally, the electronic device 4000 may further include a transceiver 4004, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 4004 is not limited to one type, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present invention.

[0105] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this invention. Processor 4001 may also be a combination that implements computing functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0106] The memory 4003 may be a ROM (Read Only Memory) or other type of static storage device capable of storing static information and instructions, RAM (Random Access Memory) or other type of dynamic storage device capable of storing information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program instructions or code in the form of instructions or data structures and accessible by the electronic device 4000, but not limited thereto.

[0107] The memory 4003 stores computer-readable instructions, and the processor 4001 can read the computer-readable instructions stored in the memory 4003 through the communication bus 4002.

[0108] The computer-readable instructions are executed by one or more processors 4001 to implement the graded energy-saving control method in the above embodiments.

[0109] Furthermore, this embodiment of the invention provides a storage medium storing computer-readable instructions, which are executed by one or more processors to implement the graded energy-saving control method described above.

[0110] This invention provides a computer program product, which includes computer-readable instructions stored in a storage medium. One or more processors of an electronic device read the computer-readable instructions from the storage medium, load and execute the computer-readable instructions, thereby enabling the electronic device to implement the graded energy-saving control method as described above.

[0111] Compared with related technologies, the beneficial effects of the present invention are: 1. This invention can significantly reduce the silent power consumption of cameras without mains power and extend the overall battery life. By dividing the camera power supply into four independent power supply domains: normally open low-power domain, imaging domain, inference domain, and radio frequency domain, only the normally open low-power domain is enabled after power-on. The low-power MCU maintains sensor acquisition, fuel gauge reading, and power switch driving at milliwatt-level power consumption. The imaging domain, inference domain, and radio frequency domain are all in a completely powered-off state during non-event periods. This reduces the overall silent power consumption from the watt level of traditional normally open solutions to the milliwatt level, thereby significantly reducing the average daily energy consumption during non-event periods and effectively extending the battery life under battery power conditions.

[0112] 2. This invention can effectively reduce the cost of false triggering and avoid waking up high-power modules due to PIR false alarms. After being triggered by the sensor, a power pre-check is performed. Power is only supplied to the camera and image processor when the power consumption state meets the requirements of the listening state. Images are acquired at a set low frame rate. The motion amplitude is obtained by calculating the frame difference or background difference of consecutive frames. Motion confirmation evidence is generated only when the motion amplitude exceeds the threshold and the number of consecutive qualified frames meets the requirements. If the motion evidence is not valid, the power supply to the imaging domain is immediately cut off and the system returns to the silent state. This ensures that invalid triggers such as rustling grass only stay in the low-power listening stage and will not rashly wake up the inference domain and RF domain, thereby controlling the energy waste caused by false triggering to an extremely low level.

[0113] 3. This invention avoids ineffective inference power consumption and RF transmission, and achieves dual gating of event evidence and power budget. After motion confirmation evidence is established, the power is checked before powering the inference domain. The target detection model infers the category label and confidence score of the cached image frame. Target confirmation evidence is generated only when the confidence of the category of interest reaches the threshold and the number of consecutive hit frames meets the requirements. After power pre-check, it can enter the return state. If the power is insufficient, it is only stored locally without RF transmission, thereby ensuring that only valid events that have been confirmed by both pixel motion and semantic target and have sufficient power trigger high-power operation, avoiding inference and RF energy waste caused by false alarms.

[0114] 4. This invention can suppress the impact of RF transmission peak current on the battery, avoid undervoltage reset, and ensure stable system operation. Before entering the backhaul state, the upper limit of the allocable average power is calculated based on the current power consumption of the non-RF domain, the remaining energy, and the predicted replenishment energy. Then, the upper limit of the RF average current is solved and converted into the maximum allowable duty cycle. After the RF domain is powered on, the system performs periodic on / off switching or issues sleep / wake-up commands according to the heartbeat duty cycle or backhaul duty cycle, so that the average current of RF transmission does not exceed the upper limit. At the same time, the average current is measured by a sliding window and a hard-limited shutdown is immediately executed when the limit is exceeded. Thus, the peak current of the RF power amplifier is converted into a predictable average current, effectively suppressing battery undervoltage reset and board-side over-matching problems.

[0115] 5. This invention can reduce the specification requirements of solar panels and batteries, thereby reducing material costs and installation volume. Through domain-based tiered power supply, dual-trigger constraints of events and power consumption, and radio frequency duty cycle current limiting, the entire unit can operate at milliwatt-level during non-event periods, significantly reducing the average daily energy demand. Under the premise of meeting the constraint of continuous multi-day weak power supplementation in cloudy and rainy weather, smaller peak power solar panels and smaller capacity batteries can be used to maintain the target range, thereby reducing the overall material cost and installation volume. At the same time, the data integrity and system reliability in weak power supplementation scenarios are further improved through backhaul breakpoint buffering and condition improvement subsequent transmission mechanism.

[0116] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0117] The above description is only a partial embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A graded energy-saving control method, characterized in that, The method includes: The system continuously collects the sensor's status parameters using a low-power MCU in the normally open low-power domain. Based on these status parameters, it calculates the remaining energy and predicts the replenishment energy. It also calculates the budget tightness based on the recent moving average power consumption and the target range. Finally, it maps the power consumption state to the allowed power consumption state by combining the status parameters with the hysteresis threshold. The status parameters include the trigger signal, the state of charge value of the battery fuel gauge, and the photovoltaic power value and charging current value of the solar sampling circuit. When the sensor is triggered, it detects whether the current power level can enter the listening state. If so, it supplies power to the camera and image processor, controls the camera to acquire images at a set low frame rate, and calculates the frame difference or background difference of consecutive frames to obtain the motion amplitude. When the motion amplitude reaches the set requirement, motion confirmation evidence is generated. When the motion confirmation evidence is established, the current power level is checked to see if the analysis state can be entered. If it can, power is supplied to the inference domain, and the inference domain is controlled to infer the image to obtain the category of the detected target in the image and the corresponding confidence score. When the confidence score reaches the set requirement, target confirmation evidence is generated. When the target confirmation evidence is established, it is detected whether the current power level can enter the backhaul state. If so, the maximum allowable duty cycle is calculated based on the power consumption of the currently powered non-RF domain, the remaining energy, and the predicted replenishment energy. After powering the RF domain, the average current of the RF transmission in the RF domain is controlled to not exceed the maximum allowable duty cycle.

2. The graded energy-saving control method as described in claim 1, characterized in that, The process of calculating remaining energy and predicted replenishment energy based on the state parameters, calculating budget tightness based on recent moving average power consumption and target range, and mapping the state parameters and hysteresis threshold to obtain the permissible power consumption state includes: The remaining energy is calculated by multiplying the state of charge value with the preset rated capacity value and coulomb efficiency coefficient. The instantaneous charging power is obtained by multiplying the photovoltaic power value or charging current value with the bus voltage value. The instantaneous charging power is integrated and accumulated within the predicted horizon using a sliding time window to obtain the predicted energy replenishment. The average power consumption within the recent sliding time window is calculated. The recent sliding average power consumption is multiplied by the preset target battery life to obtain the total energy required for the battery life. The total energy required for the battery life is divided by the sum of the remaining energy and the predicted replenishment energy to obtain the budget tightness. A preset backtracking mechanism is used to determine the threshold to be used when switching power consumption states based on the budget tightness and the preset threshold range in which the state of charge value falls, and to map the power consumption state that is currently allowed by the power supply. The power consumption states include silent state, listening state, analysis state and feedback state.

3. The graded energy-saving control method as described in claim 1, characterized in that, When the sensor is triggered, the current battery level is checked to determine if the system can enter a listening state. If so, power is supplied to the camera and image processor, and the camera is controlled to acquire images at a set low frame rate. The frame difference or background difference is calculated on consecutive frames to obtain the motion amplitude. When the motion amplitude reaches a set requirement, motion confirmation evidence is generated, including: When the sensor’s trigger signal is valid, the current event level is set to the lowest level and compared with the power consumption state that is allowed by the power. If the power consumption state meets the requirements of the listening state, the camera and image processor are powered, and a preset stabilization delay is waited after power is supplied. After the stabilization delay ends, the camera is controlled to acquire images at a set low frame rate. The acquired continuous frames are subjected to pixel-by-pixel difference calculation or background difference calculation with a preset background reference frame to obtain the motion amplitude, and the motion amplitude is continuously monitored to see if it exceeds a preset motion threshold. When the motion amplitude exceeds the motion threshold and the number of consecutive frames meeting the preset requirement is reached, pixel-level motion confirmation evidence is generated. If the motion amplitude falls back below the motion threshold, or the number of consecutive frames meeting the requirement is interrupted, or the trigger signal of the sensor disappears, the power supply is cut off and the system returns to a silent state.

4. The graded energy-saving control method as described in claim 1, characterized in that, When the motion confirmation evidence is established, the current battery level is checked to determine if the analysis state can be entered. If so, power is supplied to the inference domain, and the inference domain is controlled to infer the image to obtain the category of the detected target in the image and its corresponding confidence score. When the confidence score reaches a set requirement, target confirmation evidence is generated, including: When the motion confirmation evidence is established, the event level is raised from the lowest trigger level to the intermediate level and compared with the power consumption state that is allowed by the power. If the power consumption state meets the analysis state requirements, the inference domain is powered and a preset ready delay is waited after power is supplied. After the readiness delay ends, the acquired image frames are transmitted to the inference domain. The inference domain loads a preset target detection model to perform inference operations on the image frames and outputs the category labels and corresponding confidence scores of each detected target; the confidence scores are between 0 and 1. The system filters preset categories of interest from the category labels. When the confidence score corresponding to the category of interest reaches a preset threshold and the number of consecutive hit frames reaches a preset requirement, target confirmation evidence is generated. Otherwise, the power supply to the inference domain, camera, and image processor is cut off in reverse order, and the system returns to the listening state or silent state.

5. The graded energy-saving control method as described in claim 2, characterized in that, When the target confirmation evidence is established, the step of detecting whether the current power level can enter the backhaul state, and if so, calculating the maximum allowable duty cycle based on the currently powered non-RF domain power consumption, remaining energy, and predicted replenishment energy, includes: When the target confirmation evidence is established, the event level is raised from the middle level to the highest level and compared with the power consumption state of the power permission. If the power consumption state meets the feedback state requirements, the real-time power consumption of the powered-on non-RF domain is obtained as the base power consumption value. The remaining energy is added to the predicted replenishment energy to obtain the total future available energy. The total future available energy is divided by a preset time window and multiplied by a preset safety factor to obtain the upper limit of the allocable average power. The base power consumption value is subtracted from the upper limit of the allocable average power to obtain the radio frequency allocable power quota. The upper limit of the average RF current is obtained by dividing the RF allocable power limit by the system bus voltage. The upper limit of the average RF current is then divided by the RF transmit peak current and the smaller value between the upper limit of the average RF current and the maximum allowable duty cycle of the RF module is obtained.

6. The graded energy-saving control method as described in claim 5, characterized in that, The step of controlling the average current of radio frequency transmission in the radio frequency domain to not exceed the maximum allowable duty cycle after powering the radio frequency domain includes: Power the radio frequency domain and wait for a preset network dwell time delay. After the network dwell time delay ends, periodically switch the power supply of the radio frequency domain on and off according to the heartbeat duty cycle or the feedback duty cycle, or issue sleep and wake-up commands to the radio frequency domain so that the radio frequency domain alternately enters working state and sleep state. The average current emitted by the radio frequency domain is measured using a sliding window. When the average current exceeds the upper limit, the power supply to the radio frequency domain is cut off. When the backhaul is completed, or the event triggering condition disappears, or the power consumption state decreases, all power supplies are cut off in reverse order and the system returns to a silent state.

7. The graded energy-saving control method as described in claim 1, characterized in that, The method further includes: After cutting off all power, record the breakpoint of the currently incomplete backhaul data and cache the incomplete backhaul data in the local storage medium. Continue to perform status parameter acquisition and budget update. When the budget tension improves and the power consumption state rises back to the backhaul state, re-power the RF domain and send the incomplete backhaul data. The low-power MCU is awakened by a periodic wake-up or a trigger signal from the sensor to perform event processing operations. After completing the processing operations, the collected data is updated to the internal register or low-power memory, and the MCU re-enters sleep mode, waiting for the next real-time clock interrupt to wake it up again to perform the next acquisition operation.

8. A graded energy-saving control device, characterized in that, The device includes: The resident data acquisition and budget mapping module is used to continuously acquire the sensor's status parameters through a low-power MCU in the normally open low-power domain, calculate the remaining energy and predict the replenishment energy based on the status parameters, calculate the budget tension based on the recent moving average power consumption and the target range, and combine the status parameters and hysteresis threshold to map the power permissible state; the status parameters include the trigger signal, the state of charge value of the battery fuel gauge, and the photovoltaic power value and charging current value of the solar sampling circuit; The imaging power-on and motion confirmation module is used to detect whether the current power level can enter the listening state when the sensor is triggered. If it can, it supplies power to the camera and image processor, controls the camera to acquire images at a set low frame rate, and calculates the frame difference or background difference of consecutive frames to obtain the motion amplitude. When the motion amplitude reaches the set requirement, motion confirmation evidence is generated. The inference power-on and target confirmation module is used to detect whether the current power level can enter the analysis state when the motion confirmation evidence is established. If it can, it powers the inference domain and controls the inference domain to infer the image to obtain the category of the detected target in the image and the corresponding confidence score. When the confidence score reaches the set requirement, it generates target confirmation evidence. The RF power-on and duty cycle current limiting module is used to detect whether the current power level can enter the feedback state when the target confirmation evidence is established. If it can, the module calculates the maximum allowable duty cycle based on the power consumption of the currently powered non-RF domain, the remaining energy, and the predicted replenishment energy. After powering the RF domain, the module controls the average current of the RF transmission in the RF domain to not exceed the maximum allowable duty cycle.

9. An electronic device, characterized in that, include: At least one processor and at least one memory, wherein, The memory stores computer-readable instructions; The computer-readable instructions are executed by one or more of the processors, causing the electronic device to implement the graded energy-saving control method as described in any one of claims 1 to 7.

10. A storage medium having computer-readable instructions stored thereon, characterized in that, The computer-readable instructions are executed by one or more processors to implement the graded energy-saving control method as described in any one of claims 1 to 7.

Images