Ultrasonic infrared positioning sorting method

Abstract

The invention relates to the technical field of renewable resource sorting, in particular to an ultrasonic infrared positioning sorting method which comprises the following steps that a detection module is arranged above a sorting area of a conveying belt, and the detection module comprises an ultrasonic sensor unit and an infrared sensor unit; the two sensor units are mechanically and rigidly connected and the detection directions are kept in spatial synchronization; controlling the detection module to cooperatively detect a target object on the conveyor belt, and synchronously acquiring a first detection signal from an ultrasonic sensor and a second detection signal from an infrared sensor; performing real-time event detection on the first detection signal and the second detection signal, and determining target object positioning information according to a detection result; and the target object positioning information is sent to the sorting execution mechanism, the sorting execution mechanism is controlled to sort the target objects, and the sorting accuracy and sorting efficiency of the renewable resources can be improved.

Description

Technical Field

[0001] This invention relates to the field of recycled resource sorting technology, specifically to an ultrasonic infrared positioning and sorting method. Background Technology

[0002] In dynamic and unstructured recycling resource sorting scenarios, objects on conveyor belts exhibit complex characteristics such as varying materials, irregular shapes, and potential stacking. Single ultrasonic or infrared sensor positioning solutions have limitations: ultrasonic sensors are sensitive to surface materials and dust, leading to jitter or loss of positioning accuracy; infrared sensors are sensitive to object color and ambient light, potentially distorting contour information. Existing fusion solutions often involve fixed-weight data overlay or simple sensor switching, making it difficult to handle specific events such as stacking or single-sensor failure. This results in poor positioning accuracy under complex conditions, impacting the accuracy and efficiency of recycling resource sorting.

[0003] Therefore, we propose an ultrasonic infrared positioning and sorting method to solve the above problems. Summary of the Invention

[0004] The purpose of this invention is to provide an ultrasonic infrared positioning and sorting method to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an ultrasonic infrared positioning and sorting method, the method comprising the following steps: A detection module is set above the conveyor belt sorting area. The detection module includes an ultrasonic sensor unit and an infrared sensor unit. The two sensor units are mechanically rigidly connected and their detection directions are kept spatially synchronized. The detection module is controlled to perform coordinated detection of target objects on the conveyor belt, and simultaneously acquires a first detection signal from an ultrasonic sensor and a second detection signal from an infrared sensor. Real-time event detection is performed on the first and second detection signals, and the target object positioning information is determined based on the detection results; The target object's location information is sent to the sorting execution mechanism to control it to sort the target object. While the sorting execution mechanism is performing a grasping operation on the current object, overlapping motion control is performed.

[0006] Preferably, the step of controlling the detection module to perform coordinated detection of target objects on the conveyor belt and simultaneously acquiring a first detection signal from an ultrasonic sensor and a second detection signal from an infrared sensor includes: A unified detection cycle clock is generated to control the sensor module to detect objects on the conveyor belt at a fixed frequency. Each detection cycle is divided into a synchronization trigger phase, a data acquisition phase, and a processing and tracking phase. In the synchronous trigger phase, ultrasonic emission pulses and infrared scanning / ranging start signals are generated simultaneously, enabling the two sensors to detect the state of the object at the same time. During the data acquisition phase, the following actions are performed in parallel: capturing the complete time-voltage waveform of the ultrasonic echo as the first detection signal; and capturing the raw output data stream of the infrared sensor as the second detection signal. In processing the tracking phase; At the start of the next detection cycle, the sensor module has already performed detection according to the adjusted orientation and parameters.

[0007] Preferably, the step of processing the tracking phase includes: The two signals are timestamped using a unified clock; the measurement point of the second detection signal is mapped to the same spatial coordinate system as the first detection signal. Verify the integrity of the two signals; if the verification fails, trigger a re-probe. In the current detection period and the preceding N-1 consecutive periods, the effective first detection signal and the second detection signal are correlated according to the time sequence; Based on the correlated multi-period data, a motion trajectory model of the target object is established; based on the motion trajectory model, the expected position of the target object in the next detection cycle is predicted. Based on the predicted position, the pointing adjustment parameters of the sensor module are calculated; before the end of the current detection cycle, the pointing adjustment parameters are sent to the drive mechanism of the sensor module.

[0008] Preferably, the step of performing real-time event detection on the first detection signal and the second detection signal, and determining the location information based on the detection result, includes: If multiple reflection peaks are detected in the first detection signal and the contour of the second detection signal is spatially discontinuous, a stacking event is triggered, and layered positioning is performed to determine the target object's location information. If no stacking event is triggered, the first and second detection signals are evaluated in real time, and the dominant positioning signal source is dynamically selected based on the evaluation results. The cross-sensor correction strategy is determined based on the dominant positioning signal source to obtain the target object positioning information.

[0009] Preferably, the step of triggering a stacking event and performing layered positioning to determine positioning information if multiple reflection peaks are detected in the first detection signal and the contour of the second detection signal is spatially discontinuous includes: The boundary of the uppermost object is identified using continuous contour segments of the second detection signal; Simultaneously, the time series of each reflection peak in the first detection signal is analyzed, the height of the lower object is calculated by inversion, and layered positioning information containing layer height and outline is generated.

[0010] Preferably, the step of performing real-time quality assessment on the first and second detection signals if no stacking event is triggered, and dynamically selecting the dominant signal source based on the quality assessment results, includes: When the quality parameter of the first detection signal is lower than the first reliability threshold and the quality parameter of the second detection signal is higher than the second reliability threshold, it is determined to be an infrared reliable working area, and an infrared sensor is selected as the dominant signal source for positioning. When the quality parameter of the second detection signal is lower than the second reliability threshold and the quality parameter of the first detection signal is higher than the first reliability threshold, it is determined to be an ultrasonic reliable working area, and the ultrasonic sensor is selected as the dominant signal source for positioning. When the quality parameters of both the first and second detection signals are higher than their respective reliability thresholds, it is determined to be a reliable working area for dual sensors, and multi-source data fusion is performed.

[0011] Preferably, the step of determining the cross-sensor correction strategy based on the dominant positioning signal source to obtain the target object positioning information includes: When the infrared sensor is used as the primary signal source for positioning, the contour positioning data of the object is obtained based on the infrared sensor. At the same time, reliable distance reference information is extracted from the ultrasonic signal, which includes the arrival time reference point of the ultrasonic signal envelope. Based on the distance reference information, the contour positioning data is calibrated and compensated to generate target object positioning information; When an ultrasonic sensor is used as the primary signal source for positioning, distance positioning data of an object is obtained based on the ultrasonic sensor. At the same time, reliable spatial feature information is extracted from the infrared signal, including the intensity distribution pattern or edge features of the infrared signal. Based on the spatial feature information, the distance positioning data generated by the ultrasonic wave is spatially calibrated to correct the deviation of the ranging point caused by the geometric characteristics of the reflecting surface, and to generate the positioning information of the target object. When performing multi-source data fusion, the measurement data from ultrasonic and infrared sensors are directly fused and calculated to generate the fused target object positioning information.

[0012] Preferably, the step of performing overlapping motion control while the sorting execution mechanism performs a grasping operation on the current object includes: Predict the arrival time and location of the next object to be sorted based on the conveyor belt speed; plan the movement trajectory of the sensor module based on the prediction results; While the sorting robot occupies the workspace, the control sensor module moves ahead of time along the avoidance trajectory to the next detection starting position.

[0013] Compared with the prior art, the beneficial effects of the present invention are: By combining two types of sensors, rather than simply using one of them, the sensors can actively adapt to changes in the target state and environment. This allows for continuous and stable locking and tracking of high-speed moving targets, improving the utilization rate of data collected by both types of sensors, increasing the accuracy of target object positioning, and thus improving the success rate of grasping in the sorting process of recyclable resources under complex working conditions. Ultimately, this enhances the accuracy and efficiency of recyclable resource sorting. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example: Please refer to Figure 1 This invention provides a technical solution for an ultrasonic infrared positioning and sorting method: an ultrasonic infrared positioning and sorting method comprising the following steps: S1: A detection module is set above the conveyor belt sorting area. The detection module includes an ultrasonic sensor unit and an infrared sensor unit. The two sensor units are mechanically rigidly connected and their detection directions are kept spatially synchronized. S2: Control the detection module to perform cooperative detection of target objects on the conveyor belt, and simultaneously collect the first detection signal from the ultrasonic sensor and the second detection signal from the infrared sensor; The steps of controlling the detection module to collaboratively detect target objects on the conveyor belt and synchronously acquire a first detection signal from an ultrasonic sensor and a second detection signal from an infrared sensor include: generating a unified detection cycle clock; controlling the sensor module to detect objects on the conveyor belt at a fixed frequency; each detection cycle is divided into a synchronization trigger phase, a data acquisition phase, and a processing tracking phase; in the synchronization trigger phase, ultrasonic emission pulses and infrared scanning / ranging start signals are generated synchronously, enabling both sensors to detect the object's state at the same time; in the data acquisition phase, the following are performed in parallel: capturing the complete time-voltage waveform of the ultrasonic echo as the first detection signal; and capturing the raw output data stream of the infrared sensor as the second detection signal; in the processing tracking phase; at the start of the next detection cycle, the sensor module has already detected according to the adjusted direction and parameters; It should be noted that the synchronous triggering specifically includes: connecting the ultrasonic sensor's emission enable signal and the infrared sensor's sampling start signal to the same digital output port; at the beginning edge of the first time window, the digital output port generates a rising edge pulse, simultaneously starting ultrasonic emission and infrared sampling; The steps for processing the tracking phase include: assigning the same timestamp to both signals based on a unified clock; mapping the measurement point of the second detection signal to the same spatial coordinate system as the first detection signal; verifying the integrity of the two signals, and triggering re-detection if the verification fails; associating the valid first detection signal and the second detection signal in a time sequence within the current detection cycle and the preceding N-1 consecutive cycles; establishing a motion trajectory model of the target object based on the associated multi-cycle data; predicting the expected position of the target object in the next detection cycle based on the motion trajectory model; calculating the pointing adjustment parameters of the sensor module according to the predicted position; sending the pointing adjustment parameters to the drive mechanism of the sensor module before the end of the current detection cycle; and dynamically optimizing the sensor parameters for the next detection cycle, including ultrasonic emission power, infrared exposure time, or scanning frequency, based on the distance, speed, and surface characteristics of the target object. It should be noted that multi-cycle data association includes motion consistency filtering: calculating the motion vectors of associated points in different cycles; removing outliers whose motion direction or amplitude deviates significantly from the mainstream direction of the conveyor belt; constructing an accurate motion model of the object using a point set with good spatiotemporal continuity; predicting the expected position of the target object in the next detection cycle, and selecting the prediction model based on the object type and surface characteristics; dynamic parameter optimization follows the principle of maximizing the signal-to-noise ratio: for ultrasonic sensors: increasing the transmission power when the object distance increases or the surface absorbs sound; adjusting the transmission frequency to avoid noise bands when ambient noise increases; for infrared sensors: increasing the exposure time or illumination intensity when the object surface has low reflectivity or strong ambient light; increasing the scanning frequency to reduce motion blur when the object is moving at high speed. Specifically, the sensor module includes rigidly connected and optically parallel ultrasonic and infrared sensors, which can be driven by a servo mechanism to adjust their spatial orientation. The main controller generates a highly stable detection cycle clock signal (e.g., with a period of T milliseconds). Each detection cycle is logically divided into three consecutive phases: a synchronization trigger phase, a data acquisition phase, and a processing tracking phase. The entire system operates in a coordinated and orderly manner based on this clock. At the beginning of each detection cycle (i.e., the start edge of the synchronization trigger phase), the system generates a rising edge pulse through a dedicated digital output port. This pulse signal is simultaneously connected via physical lines to: the transmit enable pin of the ultrasonic sensor, used to trigger it to emit a set of ultrasonic pulses; and the sampling start pin of the infrared sensor, used to trigger it to begin a scan or ranging sampling. By directly connecting and synchronously triggering, the system ensures that both sensors detect the object's state at the same instant in physical time, guaranteeing the accuracy of the data obtained from subsequent data fusion. After the trigger pulse, the system immediately enters the data acquisition phase, simultaneously activating two independent signal acquisition channels: The ultrasonic signal channel receives the ultrasonic echo signal, amplifies it analogally, and filters it with a bandpass filter before capturing the complete time-voltage waveform using a high-speed analog-to-digital converter (ADC), denoted as the first detection signal. The infrared signal channel receives the raw data stream output from the infrared sensor (e.g., a one-dimensional range array based on time-of-flight, or two-dimensional point cloud data based on triangulation / structured light), directly buffers it, and denoted as the second detection signal. Based on the generated unified clock, the data blocks of the first and second detection signals are timestamped with the same timestamp. According to the pre-calibrated spatial relative position and attitude relationship between the ultrasonic and infrared sensors, all measurement points (such as infrared point clouds) in the second detection signal are uniformly transformed into a spatial coordinate system with the ultrasonic sensor as the origin. The two signals are then rapidly checked. For example, the waveform of the first detection signal is checked to see if a valid echo peak exceeding a threshold amplitude appears within the expected flight time window; the number of valid measurement points in the second detection signal is checked to ensure its spatial distribution is continuous (without area holes exceeding a preset area threshold). If either signal fails verification, the normal process is immediately interrupted, a re-detection subroutine is initiated (e.g., a brief increase in transmission power followed by an immediate retry), and the data for that period is marked as low confidence. The valid detection signal in the current period (period k) is correlated with the valid signals of the previous N-1 consecutive periods (periods k-N+1 to k-1) to form a cross-time observation sequence. For the correlated feature points (such as the main echo point of ultrasound and stable corner points in the infrared point cloud), their motion vectors are calculated during the consecutive periods. By setting a threshold, abnormal points whose motion direction or amplitude is significantly inconsistent with the mainstream direction of the conveyor belt (known) are removed. For example, if the conveyor belt moves uniformly to the right, a point moving rapidly to the left is likely noise or a misjudgment and should be removed. Using the filtered set of points with good spatiotemporal continuity, a motion trajectory model of the current target object (such as position-time curve and velocity vector) is established. Based on the established object motion model, the expected position of the target object at the start of the next detection period (period k+1) is determined. According to the predicted position and the current pointing of the sensor module, the pointing adjustment parameters required to align the center of the sensor's optical axis with the predicted position are calculated. Before the processing and tracking phase of the current detection period (period k) ends, the calculated pointing adjustment parameters are sent to the servo motor of the sensor module. The system drives the mechanism to begin moving towards a new orientation; ultrasonic parameter optimization: if the currently detected object is far away, or the object surface absorbs sound (manifested as weak echo amplitude), the ultrasonic transmission power of the next cycle is appropriately increased; the noise spectrum in the environment is monitored in real time, and if severe interference is found in a specific frequency band, the ultrasonic transmission frequency of the next cycle is adjusted to avoid the noise band; infrared parameter optimization: if the object surface has low reflectivity (such as black rubber) or the ambient light is very strong, resulting in a weak infrared signal, the exposure time or supplementary light intensity of the infrared sensor in the next cycle is increased; if the object moves very fast, the scanning frequency of the infrared sensor in the next cycle is increased to reduce motion blur. These optimized parameters, along with the pointing adjustment parameters, will be set for the next detection cycle. When the (k+1)th detection cycle begins, the servo mechanism of the sensor module has completed the pointing adjustment, and the ultrasonic and infrared sensors are ready according to the optimized new parameters, enabling the sensors to continuously and stably lock onto and track high-speed moving target objects. By ensuring the simultaneity of data through hardware synchronous triggering, the sensors can actively adapt to changes in target state and environment, improving the accuracy of target object positioning during recyclable resource sorting, thereby improving the efficiency of recyclable resource sorting.

[0018] S3: Perform real-time event detection on the first and second detection signals, and determine the target object's location information based on the detection results; The steps of performing real-time event detection on the first and second detection signals and determining the positioning information based on the detection results include: if multiple reflection peaks are detected in the first detection signal and spatial discontinuity exists in the contour of the second detection signal, a stacking event is triggered, and layered positioning is performed to determine the positioning information of the target object; if no stacking event is triggered, real-time quality assessment is performed on the first and second detection signals, and the dominant positioning signal source is dynamically selected based on the quality assessment results; a cross-sensor correction strategy is determined based on the dominant positioning signal source to obtain the positioning information of the target object. It should be noted that when both stacking events and reliability degradation events are detected simultaneously, stacking events are processed first. After generating the layered positioning information, the corresponding reliability discrimination positioning process is applied layer by layer based on the signal quality evaluation results of each layer of objects. If multiple reflection peaks are detected in the first detection signal and the contour of the second detection signal is spatially discontinuous, a stacking event is triggered, and the steps of performing layered positioning to determine the positioning information include: using the continuous contour segments of the second detection signal to identify the boundary of the uppermost object; at the same time, parsing the time series of each reflection peak in the first detection signal, inverting and calculating the height of the lower object, and generating layered positioning information containing the layer height and contour. The steps of analyzing the time series of each reflection peak in the first detection signal, inverting and calculating the height of the lower object, and generating layered positioning information containing layer height and contour include: determining the distance of the uppermost object corresponding to the first reflection peak; calculating the time difference between the subsequent reflection peaks and the first reflection peak; calculating the height difference of each lower object relative to the upper object based on the propagation speed and time difference of ultrasonic waves in the medium; and mapping the spatial position of each object in the layer by combining the contour of the upper object identified by the second signal. If no stacking event is triggered, the first and second detection signals are subjected to real-time quality assessment. The step of dynamically selecting the dominant positioning signal source based on the quality assessment results includes: when the quality parameter of the first detection signal is lower than the first reliability threshold and the quality parameter of the second detection signal is higher than the second reliability threshold, it is determined to be an infrared reliable working area, and an infrared sensor is selected as the dominant positioning signal source; when the quality parameter of the second detection signal is lower than the second reliability threshold and the quality parameter of the first detection signal is higher than the first reliability threshold, it is determined to be an ultrasonic reliable working area, and an ultrasonic sensor is selected as the dominant positioning signal source; when the quality parameters of both the first and second detection signals are higher than their respective reliability thresholds, it is determined to be a dual-sensor reliable working area, and multi-source data fusion is performed. The steps for determining the target object positioning information based on the dominant positioning signal source and implementing a cross-sensor correction strategy include: when an infrared sensor is used as the dominant positioning signal source, acquiring the object's contour positioning data based on the infrared sensor, and simultaneously extracting reliable distance reference information from the ultrasonic signal, the distance reference information including the arrival time reference point of the ultrasonic signal envelope; calibrating and compensating the contour positioning data based on the distance reference information to generate the target object positioning information; when an ultrasonic sensor is used as the dominant positioning signal source, acquiring the object's distance positioning data based on the ultrasonic sensor, and simultaneously extracting reliable spatial feature information from the infrared signal, the spatial feature information including the intensity distribution pattern or edge features of the infrared signal; calibrating the spatial position of the ultrasonic-generated distance positioning data based on the spatial feature information to correct the ranging point deviation caused by the geometric characteristics of the reflecting surface, and generating the target object positioning information; when performing multi-source data fusion, directly fusing and calculating the measurement data from the ultrasonic and infrared sensors to generate the fused target object positioning information. It should be noted that the specific content of the reliable distance reference information extracted from the ultrasonic signal is as follows: In the event of ultrasonic reliability degradation, even if the overall signal quality deteriorates, at least one of the following reliable information can still be extracted from the ultrasonic signal: the time interval between the ultrasonic emission pulse and the first stable echo pulse, the position of the main peak in the ultrasonic signal envelope that directly corresponds to the front surface of the object, and the energy distribution characteristics of the ultrasonic signal in different time windows. The reliable information is used to establish an absolute distance reference and calibrate the relative depth measurement of the infrared sensor. It should be noted that the specific content of extracting reliable spatial feature information from infrared signals includes: in infrared reliability degradation events, even if the overall signal strength is insufficient, at least one of the following reliable information can still be extracted from the infrared signal: the first or second spatial derivative features of the infrared intensity distribution, used to identify object edges; the regional intensity distribution pattern of the infrared signal, used to identify the orientation or posture of the object; and the relative intensity relationship of the infrared signal between different spatial locations; the reliable information is used to provide geometric constraints on the object and correct the spatial ambiguity of single-point ultrasonic measurements; Specifically, the first and second input detection signals are analyzed in parallel to detect specific events, including stacking event detection and reliability degradation event detection (performed in parallel). Stacking events take precedence over reliability degradation events. For stacking event detection: the waveform of the first detection signal is analyzed to determine if there are multiple obvious reflection peaks (e.g., two or more echo peaks exceeding a threshold are identified using a peak detection algorithm). Simultaneously, the object contour reconstructed from the second detection signal is analyzed to determine if there are spatial discontinuities (e.g., contour breaks or protrusions that cannot be explained by a single object). When both conditions are met, a stacking event is determined to be triggered. The quality parameters of the first detection signal (e.g., signal-to-noise ratio, main peak amplitude stability) and the quality parameters of the second detection signal (e.g., average intensity, contour gradient consistency) are calculated. When any parameter is lower than its corresponding dynamic reliability threshold, a potential reliability degradation event flag is recorded. If a stacking event is detected, regardless of whether a reliability degradation flag is present, the stacking layer positioning process is immediately initiated. If no stacking event is detected, a primary-secondary collaborative positioning process based on reliability is used, and specific decisions are made based on the corresponding quality parameters. When entering the stacked layer localization process, the target is identified as multiple objects stacked vertically, and the localization goal is to obtain the positional information of each layer of objects. The second detection signal (infrared) with relatively more stable quality is processed first. From the discontinuous contours, the topmost continuous contour segment is extracted to identify the complete two-dimensional boundary of the top layer object. The first detection signal (ultrasonic wave) is processed simultaneously. The time series of its multiple reflection peaks is analyzed: the flight time corresponding to the first reflection peak is determined, and the distance D1 from the sensor to the surface of the top layer object is calculated. The time differences Δt2, Δt3, etc., between the second, third, ... reflection peaks and the first reflection peak are calculated. Based on the propagation speed v of ultrasonic waves in air, the height difference between the upper surface of each lower layer object and the lower surface of the upper layer object is calculated: ΔH2 = v * Δt2 / 2, ΔH3 = v * Δt3 / 2 … (dividing by 2 is due to the round-trip path). Spatial mapping is performed by combining the contour (planar position) of the top layer object and the height differences of each layer. Assuming the height of the topmost object's bottom surface is H0 (known or calibrated), the 3D position of the first (topmost) object is defined by its contour and height H0; the height of the second-layer object is H0 + ΔH2, and its planar position can be estimated by projecting the topmost contour in the vertical direction and combining it with possible offset models (or marked as approximately directly below); the same logic applies to lower layers. Finally, layered positioning information containing the height and planar contour information of each layer is generated. When no stacking event is detected, a reliability-based master-slave collaborative positioning process is executed, making dynamic decisions based on signal quality. If the quality parameter of the first detection signal is lower than the first reliability threshold and the quality parameter of the second detection signal is higher than the second reliability threshold, it is determined that the system has entered the infrared reliable operating zone. The infrared sensor is selected as the dominant positioning signal source. If the quality parameter of the second detection signal is lower than the second reliability threshold and the quality parameter of the first detection signal is higher than the first reliability threshold, it is determined that the system has entered the ultrasonic reliable operating zone. The ultrasonic sensor is selected as the dominant positioning signal source. If the quality parameters of both signals are higher than their respective thresholds, it is determined that the system has entered the dual-sensor reliable operating zone. No single dominant source is specified; fusion is performed.

[0019] Scenario A: Infrared-dominated correction (infrared reliable operating area), using the second detection signal as the primary source, generates high-precision contour positioning data (two-dimensional plane coordinate set) for the target object. Despite the poor overall quality of the ultrasonic signal, an attempt is made to extract reliable distance reference information. Specifically, one or more of the following can be extracted: the time interval between the ultrasonic emission pulse and the first stable echo pulse. This time point usually corresponds to the front surface of the object, is less affected by internal reflections, and is relatively reliable. The main peak position directly corresponding to the front surface of the object is identified from the signal envelope. The energy distribution of the signal within different time windows is analyzed to find the time delay corresponding to the region of concentrated and stable energy. Using the extracted reliable distance reference information (e.g., the absolute distance L calculated from the first echo time), the depth dimension of the infrared-generated contour positioning data is calibrated and compensated. For example, the relative depth distribution observed by infrared, with its centroid as a reference, is translated as a whole onto the plane determined by the absolute distance L, thereby correcting the systematic bias of the infrared ranging system caused by the object's color or material, and generating the final three-dimensional positioning information.

[0020] Scenario B: Ultrasonic-Dominated Correction (Reliable Ultrasonic Operating Area). Using the first detection signal as the primary source, core distance positioning data (one or more 3D points containing accurate depth information) is generated for the target object. Even with insufficient overall infrared signal strength or high noise, attempts are made to extract still reliable spatial feature information. Specifically, this includes: calculating the first or second spatial derivative of the infrared intensity distribution, whose extreme points can robustly indicate object edges, even with low overall contrast; identifying intensity distribution patterns in different regions of the infrared image (e.g., which side is brighter) to infer the approximate orientation or tilt of the object; and recognizing the relative intensity relationships between different spatial locations, which are more stable than absolute intensity values. The extracted spatial feature information (e.g., identified left and right edges of the object) is used to spatially calibrate the ultrasonic single-point (or sparse-point) ranging results. For example, if the ultrasonic ranging point hits an off-center edge due to the object's tilt, edge information identified by infrared can be used to correct the ranging point to the estimated object center or grasping point, correcting spatial offset errors caused by the geometric characteristics of the reflecting surface, and generating more reasonable 3D positioning information.

[0021] Scenario C: High-efficiency direct fusion (reliable operating range of dual sensors). When both sensors are reliable, a computationally efficient fixed-weight fusion algorithm (such as weighted average) or a decision-level fusion based on preset rules is used to directly fuse the measurement data of the two sensors to generate the final positioning information. This process does not require complex auxiliary information extraction and correction.

[0022] For stacked scenarios, determine the generated layered positioning information. For non-stacked scenarios, determine the final positioning information generated by the corresponding strategy. This positioning information can be sent to the robot trajectory planning module, along with confidence labels (such as the number of stacked layers and the dominant sensor type), for subsequent grasping decisions. If a stacking event is detected accompanied by a reliability degradation flag, a reliability-based master-slave collaborative positioning process needs to be executed separately for each layer of the object after completing layered positioning. For example, for a certain layer in the stack, if its surface material results in a poor infrared signal, an ultrasonic dominant correction strategy is used when calculating the position of that layer; if another layer has a smooth surface that causes chaotic ultrasonic echoes, an infrared dominant correction strategy is used for that layer. By first detecting events, then prioritizing arbitration, and finally processing layer by layer, the two types of sensors can be combined, rather than simply using one of them. The complementary information of the two is specifically corrected, improving the utilization rate of data collected by both types of sensors, adapting to the highly uncertain object states in recyclable resource sorting, improving the grasping success rate under complex working conditions, and thus improving the accuracy and efficiency of recyclable resource sorting.

[0023] S4: Send the target object positioning information to the sorting execution mechanism to control it to perform sorting operations on the target object. While the sorting execution mechanism performs the gripping operation on the current object, it performs overlapping motion control. While the sorting actuator is performing a grasping operation on the current object, the steps of performing overlapping motion control include: predicting the arrival time and position of the next object to be sorted based on the conveyor belt speed; planning the movement trajectory of the sensor module based on the prediction results; and controlling the sensor module to move in advance to the next detection starting position along the avoidance trajectory while the sorting robot occupies the workspace. It should be noted that the overlapping motion control is achieved through a dual-cycle timing sequence. Specifically, the first cycle is a detection-positioning-decision cycle, executed by the sensor module and the central controller; the second cycle is a grasping-movement cycle, executed by the drive mechanism of the sorting robot and the sensor module. The timing of the two cycles overlaps when the robot enters the grasping phase, so that the movement time of the sensor module is hidden within the grasping time of the robot. The movement trajectory planning of the sensor module follows the dynamic avoidance principle: real-time acquisition of the position and attitude of the robot's end effector; calculation of the shortest collision-free path from the current position to the next detection position of the sensor module; the path ensures that the sensor module maintains a safe distance from the robot, the grasped object, and the fixed structure of the conveyor belt during the movement; The first cycle is a detection-localization-decision cycle, which runs at a fixed high frequency (e.g., 100Hz) and is primarily executed by the sensor module and central controller. Each cycle sequentially completes: data acquisition, signal processing, event detection, localization calculation, and generation of a grasping decision command for the currently locked target object. The output of this cycle is the grasping point coordinates, posture, and grasping command sent to the robot arm. The second cycle is a grasping-movement cycle. This cycle is executed collaboratively by the sorting robot arm and the drive mechanism of the sensor module, and its cycle is determined by the physical time of a single grasping action. The core task of this cycle is to simultaneously perform the grasping, transferring, and releasing actions on the current object while driving the sensor module to move in advance to the detection starting position for the next object. The two cycles are not completely independent but have a precise phase overlap on the timeline. The key synchronization point is set at the moment when the first cycle completes the localization and decision-making for the current target object A and issues the grasping command to the robot arm, and the robot arm begins to execute the grasping action (e.g., accelerating from the preparation position to the grasping point). At the very moment the robotic arm begins its grasping action on object A, the second cycle of the movement task is immediately triggered. Based on the predicted information about the next object B to be sorted, acquired in the first cycle, the central controller begins planning and executing the trajectory of the sensor module from its position after detecting object A to its starting position to detect object B. From a temporal perspective, the entire movement process of the sensor module (acceleration, constant speed, deceleration) is completely covered within the time window of the robotic arm's grasping, lifting, and possible small-scale transfer of object A. Since the robotic arm's grasping action is physically necessary and cannot be shortened, hiding the sensor's movement within this time is equivalent to eliminating this movement time from the total cycle time. During the execution of the second cycle of the movement task, to ensure safety, the movement of the sensor module is planned for real-time obstacle avoidance. The central controller obtains the precise position (X, Y, Z) and attitude (Rx, Ry, Rz) of the robotic arm's end effector (gripper / suction cup) in real time through servo feedback. Simultaneously, the current position of the sensor module itself, the target position (next detection point), and the geometric models of all fixed obstacles in the workspace (such as conveyor belt supports and camera pillars) are known. The planner starts from the current position of the sensor module and ends at the predicted starting position of the next detection.The calculated path must ensure that the sensor module (considered a moving rigid body) maintains a distance greater than a safety threshold from the following three elements at every moment during its movement: the moving end effector (the primary dynamic obstacle); the object A already grasped by the robot (object A may be large, altering the envelope volume of the end effector); and all fixed structures. Real-time planning algorithms based on dynamic artificial potential fields or velocity obstacle methods can be used. For example, the end effector and the grasped object can be considered sources of repulsive force, while the target point can be considered a source of attractive force, calculating a real-time avoidance speed command for the sensor module. Alternatively, a time-varying no-entry zone can be defined within the overlapping area of ​​the robot's workspace and the sensor module's workspace, with the planner ensuring the sensor module's path always remains outside this zone. The planner outputs an avoidance trajectory (satisfying the sensor module's drive mechanism's acceleration and speed limits). This trajectory may not be a straight line but a curve bypassing the robot's workspace. The central controller decomposes this trajectory into servo commands for the drive mechanism, initiating movement of the sensor module. Because the planning is real-time, even if the robot arm's trajectory deviates slightly from the expected path due to unexpected grasping events (such as object slippage), the avoidance plan can be dynamically adjusted to ensure safety. At time t0, the sensor module completes the detection of object A, and the first loop generates a grasping command. At time t1 (≈t0), the robot arm receives the command and begins moving towards object A (the grasping phase begins). Simultaneously, the system starts the second loop, calculating and controlling the sensor module to move along the avoidance trajectory towards the pre-detection point of object B based on the predicted position of object B. At time t2, the robot arm reaches the position of object A and performs the grasping, clamping, and lifting actions. At time t3, the sensor module reaches the predetermined detection starting position for object B ahead of schedule and stabilizes. At time t4, the robot arm completes the grasping of object A and removes it from the conveyor belt area. At time t5, when object B enters the effective detection range, the sensor module immediately starts detecting B (the first loop begins processing B), while the robot arm may be transferring object A to the unloading area at this time. At this point, an efficient parallel cycle is completed. By setting up a double-loop timing overlap, the idle movement time of the sensor module and the effective grasping time of the robot are parallelized, allowing the sensor and the robot to work collaboratively in a shared space, improving space utilization and adaptability, thereby increasing sorting efficiency. By combining two types of sensors, rather than simply using one of them, the sensors can actively adapt to changes in the target state and environment. This allows for continuous and stable locking and tracking of high-speed moving targets, improving the utilization rate of data collected by both types of sensors, increasing the accuracy of target object positioning, and thus improving the success rate of grasping in the sorting process of recyclable resources under complex working conditions. Ultimately, this enhances the accuracy and efficiency of recyclable resource sorting.

[0024] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0025] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An ultrasonic infrared positioning and sorting method, characterized in that, Includes the following steps: A detection module is set above the conveyor belt sorting area. The detection module includes an ultrasonic sensor unit and an infrared sensor unit. The two sensor units are mechanically rigidly connected and their detection directions are kept spatially synchronized. The detection module is controlled to perform coordinated detection of target objects on the conveyor belt, and simultaneously acquires a first detection signal from an ultrasonic sensor and a second detection signal from an infrared sensor. Real-time event detection is performed on the first and second detection signals, and the target object positioning information is determined based on the detection results; The target object's location information is sent to the sorting execution mechanism to control it to sort the target object. While the sorting execution mechanism is performing a grasping operation on the current object, overlapping motion control is performed.

2. The ultrasonic infrared positioning and sorting method according to claim 1, characterized in that: The step of controlling the detection module to perform coordinated detection of target objects on the conveyor belt and synchronously acquiring a first detection signal from an ultrasonic sensor and a second detection signal from an infrared sensor includes: A unified detection cycle clock is generated to control the sensor module to detect objects on the conveyor belt at a fixed frequency. Each detection cycle is divided into a synchronization trigger phase, a data acquisition phase, and a processing and tracking phase. In the synchronous trigger phase, ultrasonic emission pulses and infrared scanning / ranging start signals are generated simultaneously, enabling the two sensors to detect the state of the object at the same time. During the data acquisition phase, the following actions are performed in parallel: capturing the complete time-voltage waveform of the ultrasonic echo as the first detection signal; and capturing the raw output data stream of the infrared sensor as the second detection signal. In processing the tracking phase; At the start of the next detection cycle, the sensor module has already performed detection according to the adjusted orientation and parameters.

3. The ultrasonic infrared positioning and sorting method according to claim 2, characterized in that: The step of processing the tracking phase includes: The two signals are timestamped using a unified clock; the measurement point of the second detection signal is mapped to the same spatial coordinate system as the first detection signal. Verify the integrity of the two signals; if the verification fails, trigger a re-probe. In the current detection period and the preceding N-1 consecutive periods, the effective first detection signal and the second detection signal are correlated according to the time sequence; Based on the correlated multi-period data, a motion trajectory model of the target object is established; based on the motion trajectory model, the expected position of the target object in the next detection cycle is predicted. Based on the predicted position, the pointing adjustment parameters of the sensor module are calculated; before the end of the current detection cycle, the pointing adjustment parameters are sent to the drive mechanism of the sensor module.

4. The ultrasonic infrared positioning and sorting method according to claim 1, characterized in that: The step of performing real-time event detection on the first and second detection signals and determining the location information based on the detection results includes: If multiple reflection peaks are detected in the first detection signal and the contour of the second detection signal is spatially discontinuous, a stacking event is triggered, and layered positioning is performed to determine the target object's location information. If no stacking event is triggered, the first and second detection signals are evaluated in real time, and the dominant positioning signal source is dynamically selected based on the evaluation results. The cross-sensor correction strategy is determined based on the dominant positioning signal source to obtain the target object positioning information.

5. The ultrasonic infrared positioning and sorting method according to claim 4, characterized in that: The step of triggering a stacking event and performing layered positioning to determine positioning information if multiple reflection peaks are detected in the first detection signal and spatial discontinuity exists in the profile of the second detection signal includes: The boundary of the uppermost object is identified using continuous contour segments of the second detection signal; Simultaneously, the time series of each reflection peak in the first detection signal is analyzed, the height of the lower object is calculated by inversion, and layered positioning information containing layer height and outline is generated.

6. The ultrasonic infrared positioning and sorting method according to claim 4, characterized in that: If no stacking event is triggered, the step of performing real-time quality assessment on the first and second detection signals, and dynamically selecting the dominant signal source based on the quality assessment results, includes: When the quality parameter of the first detection signal is lower than the first reliability threshold and the quality parameter of the second detection signal is higher than the second reliability threshold, it is determined to be an infrared reliable working area, and an infrared sensor is selected as the dominant signal source for positioning. When the quality parameter of the second detection signal is lower than the second reliability threshold and the quality parameter of the first detection signal is higher than the first reliability threshold, it is determined to be an ultrasonic reliable working area, and the ultrasonic sensor is selected as the dominant signal source for positioning. When the quality parameters of both the first and second detection signals are higher than their respective reliability thresholds, it is determined to be a reliable working area for dual sensors, and multi-source data fusion is performed.

7. The ultrasonic infrared positioning and sorting method according to claim 6, characterized in that: The step of determining the cross-sensor correction strategy based on the dominant positioning signal source to obtain the target object positioning information includes: When the infrared sensor is used as the primary signal source for positioning, the contour positioning data of the object is obtained based on the infrared sensor. At the same time, reliable distance reference information is extracted from the ultrasonic signal, which includes the arrival time reference point of the ultrasonic signal envelope. Based on the distance reference information, the contour positioning data is calibrated and compensated to generate target object positioning information; When an ultrasonic sensor is used as the primary signal source for positioning, distance positioning data of an object is obtained based on the ultrasonic sensor. At the same time, reliable spatial feature information is extracted from the infrared signal, including the intensity distribution pattern or edge features of the infrared signal. Based on the spatial feature information, the distance positioning data generated by the ultrasonic wave is spatially calibrated to correct the deviation of the ranging point caused by the geometric characteristics of the reflecting surface, and to generate the positioning information of the target object. When performing multi-source data fusion, the measurement data from ultrasonic and infrared sensors are directly fused and calculated to generate the fused target object positioning information.

8. The ultrasonic infrared positioning and sorting method according to claim 1, characterized in that: The step of performing overlapping motion control while the sorting execution mechanism performs a grasping operation on the current object includes: Predict the arrival time and location of the next object to be sorted based on the conveyor belt speed; plan the movement trajectory of the sensor module based on the prediction results; While the sorting robot occupies the workspace, the control sensor module moves ahead of time along the avoidance trajectory to the next detection starting position.

Images