Radar detection device and method for construction equipment with improved radar mal-detection

Abstract

One embodiment of the present invention relates to a radar detection device for construction equipment, wherein the radar detection device for construction equipment may comprise: a radar for sensing objects around the construction equipment; a tilt sensor for sensing the tilt of the construction equipment; a turning angle sensor for sensing the turning angle of the construction equipment; and a detection server that communicates with the radar, the tilt sensor, and the turning angle sensor and determines whether objects are present around the construction equipment.

Description

Construction equipment radar detection device and method for improving radar misrecognition

[0001] The present invention relates to a construction equipment radar detection device and method, and more particularly, to a construction equipment radar detection device and method that improves misrecognition in monitoring obstacles around construction equipment using radar.

[0002] Construction sites are exposed to the risk of safety accidents due to the diverse construction equipment used. While significant efforts have been made to improve the safety of construction equipment over the past several decades, tangible results have been limited.

[0003] This is primarily due to the lack of safety solutions applicable to construction equipment. Ultrasonic sensors, widely used for rear-view detection in vehicles, are vulnerable to vibration, making them difficult to apply to construction equipment. Some companies have installed ultrasonic sensors in construction equipment, but have since withdrawn due to damage caused by vibration.

[0004] LiDAR, a technology gaining traction in the autonomous driving field, is an effective way to detect approaching objects. However, LiDAR is not suitable for construction sites, where dust and muddy water are common.

[0005] Cameras and radar are currently the most widely used safety solutions for construction equipment. Among these, radar has garnered attention as an effective means of detecting approaching objects even in harsh construction sites. Recently, with the growing research on radar in the field of autonomous vehicles, efforts have been made to expand its application to construction equipment.

[0006] Radar is generally used to detect approaching objects, and is increasingly recognized as a crucial safety element, particularly for construction equipment operating in hazardous industrial settings. However, conventional radar has suffered from false detection issues. In specialized environments like construction sites, conventional radar often mistakenly identifies the ground or the construction equipment itself as an approaching object. This false detection reduces radar reliability and can even lead users to abandon radar use.

[0007] Conventional construction equipment has suffered from a self-reflection phenomenon, where the upper part of the equipment detects the lower part of the equipment as it rotates. For example, radar in conventional construction equipment has had a problem detecting the caterpillar as an obstacle. Conventional construction equipment lacks an effective way to eliminate the reflections caused by the equipment itself, which, like ground reflection, has hindered the widespread application of radar in construction equipment. Because radar uses radio waves to detect reflected waves from nearby objects, there is no way to distinguish between reflections caused by the ground and reflections caused by the construction equipment itself.

[0008] Meanwhile, introducing imaging techniques can easily identify components reflected from the ground or from the construction equipment itself. However, techniques that utilize cameras mounted on construction equipment to learn from these reflections and eliminate ground or self-reflection are unsuitable for dusty construction sites and incur significant costs.

[0009] To address this issue, Korean Patent Publication No. 10-2016-0140113 discloses a safety system and method for monitoring the perimeter of construction equipment. This system utilizes impulse radar sensor technology and wireless communication technology to more easily detect workers or objects around the equipment, and to quickly notify and alert the operator of any detected hazards. However, this system also has the drawback of not being able to accurately distinguish between people and obstacles under special conditions.

[0010] Existing construction equipment does not have an effective way to remove the reflection component caused by the construction equipment itself, which, like ground reflection, is an obstacle to expanding the application of radar to construction equipment.

[0011] The technical problem to be achieved by the present invention is to provide a construction equipment radar detection device and method that improves misrecognition when monitoring obstacles around construction equipment using radar.

[0012] The technical problems to be solved by the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned can be clearly understood by a person having ordinary skill in the technical field to which the present invention belongs from the description below.

[0013] In order to achieve the above technical task, one embodiment of the present invention may include a construction equipment radar detection device, which includes: a radar for detecting an object around the construction equipment; a tilt sensor for detecting an inclination of the construction equipment; a turning angle sensor for detecting a turning angle of the construction equipment; and a detection server for communicating with the radar, the tilt sensor, and the turning angle sensor, and determining whether an object exists around the construction equipment.

[0014] In addition, the detection server determines whether there is an object around the construction equipment using the inclination value received through the inclination sensor and the reflection component detected through the radar, and if the detected object is determined to be a wide-ranging object, the object can be determined to be the ground.

[0015] In addition, the measurement values ​​of the inclination sensor and the turning angle sensor are corrected by applying a Kalman filter, and the reflection component of the radar for determining the presence of the object can be selected based on the measured values ​​of the corrected inclination sensor and the turning angle sensor.

[0016] Additionally, the above wide range of objects may be objects that are simultaneously detected within the azimuth range of the radar.

[0017] Additionally, the detection server can determine that the detected object is the ground if the slope value received through the slope sensor is greater than a preset slope reference value.

[0018] In addition, the detection server further includes a processor, and the processor repeatedly learns and accumulates a process of determining the detected object as the ground, and can determine the object as the ground based on the accumulated learning information.

[0019] In addition, the reflection component of the radar for determining the presence or absence of the object may be a reflection component that is secondarily selected by applying a Kalman filter to the reflection component that is firstly selected based on the measured values ​​of the calibrated inclination sensor and the turning angle sensor.

[0020] In order to achieve the above technical task, one embodiment of the present invention may include a construction equipment radar detection device, which includes: a radar for detecting an object around the construction equipment; a turning angle sensor for detecting a turning angle of the construction equipment; and a detection server that communicates with the radar and the turning angle sensor and determines whether an object exists around the construction equipment.

[0021] In addition, the detection server determines whether there is an object around the construction equipment using the turning angle value received through the turning angle sensor and the reflection component detected through the radar. If the detected object is repeatedly detected at the same location, the detection server can determine that the object is a self-reflection of the construction equipment.

[0022] In addition, the measurement value of the above turning angle sensor is corrected by applying a Kalman filter, and the reflection component of the radar for determining the presence or absence of the object can be selected based on the measured value of the corrected turning angle sensor.

[0023] In addition, the detection server further includes a processor, and the processor repeatedly learns and accumulates a process of determining the detected object as a self-reflection of the construction equipment, and can determine the detected object as a self-reflection of the construction equipment based on the accumulated learning information.

[0024] Additionally, the radar outputs a Doppler frequency generated according to the turning motion of the construction equipment, and can recognize a nearby obstacle if the Doppler frequency exhibits a plus or minus Doppler frequency.

[0025] In order to achieve the above technical task, one embodiment of the present invention provides a method for detecting an obstacle using a construction equipment radar detection device that improves radar misrecognition, the method including: a step in which a detection server detects the approach of an object based on output information of a radar; a step in which the detection server determines whether the detected object is a wide-area object; a step in which the detection server determines whether the slope of the ground is greater than or equal to a preset reference value; and a step in which the detection server determines the detected object to be the ground if it is determined that the detected object is a wide-area object and the slope of the ground is greater than or equal to the preset reference value.

[0026] In addition, in the step of detecting the approach of the object, the inclination of the ground is measured using a tilt sensor, the turning angle of the construction equipment is measured using a turning angle sensor, the measured values ​​of the tilt sensor and the turning angle sensor are corrected by applying a Kalman filter, and the output information of the radar for detecting the approach of the object may be selected based on the measured values ​​of the corrected tilt sensor and the turning angle sensor.

[0027] Additionally, in the step of determining whether the above-mentioned wide-area object is a wide-area object, the detection server can determine that an object detected simultaneously within the azimuth range of the radar is a wide-area object.

[0028] According to an embodiment of the present invention, since a caterpillar that can be detected by turning of construction equipment can be determined by its own reflection component, there is an effect of reducing false recognition of a construction equipment radar detection device and increasing reliability.

[0029] In addition, according to an embodiment of the present invention, even when the radar turns, it is possible to detect a nearby obstacle using the Doppler frequency characteristic, thereby having the effect of preventing safety accidents.

[0030] In addition, according to an embodiment of the present invention, since reflections from the ground can be recognized even when construction equipment is located on a sloped terrain, there is an effect of reducing false recognition and increasing reliability of a construction equipment radar detection device.

[0031] In addition, according to an embodiment of the present invention, ground reflection and self-reflection can be detected using radar, a tilt sensor, and a turning angle sensor without installing a high-cost camera, thereby reducing manufacturing costs.

[0032] The effects of the present invention are not limited to the above-described effects, and should be understood to include all effects that can be inferred from the composition of the invention described in the description or claims of the present invention.

[0033] FIG. 1 is a schematic drawing of a construction equipment radar detection device (100) that improves radar misrecognition according to one embodiment of the present invention.

[0034] FIG. 2 is a schematic drawing of a construction equipment (200) equipped with a construction equipment radar detection device (100) illustrated in FIG. 1.

[0035] Figures 3 and 4 are drawings showing the turning state of the construction equipment (200) illustrated in Figure 2.

[0036] FIG. 5 is a graph showing the Doppler characteristics of the radar (10) when the construction equipment (200) illustrated in FIG. 2 rotates.

[0037] FIG. 6 is a schematic diagram illustrating a case in which a construction equipment radar detection device (100) according to one embodiment of the present invention detects a person (300).

[0038] FIG. 7a and FIG. 7b are schematic drawings illustrating ground reflection of construction equipment (200) in one embodiment of the present invention.

[0039] FIG. 8 is a drawing illustrating a case where a construction equipment radar detection device (100) according to one embodiment of the present invention detects the ground.

[0040] FIG. 9 is a schematic diagram illustrating a case where a construction equipment radar detection device (100) according to one embodiment of the present invention detects the ground and a person simultaneously.

[0041] FIG. 10 is a flowchart for explaining a ground detection method of a construction equipment radar detection device (100) according to one embodiment of the present invention.

[0042] FIG. 11 is a flowchart for explaining a self-reflection detection method of a construction equipment radar detection device (100) according to one embodiment of the present invention.

[0043] Hereinafter, the present invention will be described with reference to the attached drawings. However, the present invention can be implemented in various different forms and is therefore not limited to the embodiments described herein. In the drawings, irrelevant parts have been omitted for clarity of description, and similar parts have been designated with similar reference numerals throughout the specification.

[0044] Throughout the specification, when a part is said to be "connected (connected, contacted, or coupled)" to another part, this includes not only cases where it is "directly connected," but also cases where it is "indirectly connected" with another part in between. Furthermore, when a part is said to "include" a component, this does not exclude other components, but rather implies that it may include other components, unless otherwise specifically stated.

[0045] The terminology used herein is merely used to describe specific embodiments and is not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this specification, it should be understood that the terms "comprises" or "has" indicate the presence of a feature, number, step, operation, component, part, or combination thereof described in the specification, but do not exclude in advance the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0046] The artificial intelligence neural network referred to herein may be a neural network trained using supervised learning. Supervised learning is a machine learning method for inferring a function from training data. Among these inferred functions, outputting continuous values ​​is called regression analysis, while predicting and outputting the class of an input vector is called classification. In supervised learning, an artificial neural network is trained with labels for training data. Here, the label can refer to the correct answer (or output value) that the artificial neural network must infer when training data is input to the artificial neural network. The correct answer that the artificial neural network must infer when training data is input can be called a label or labeling data, and labeling the training data for the purpose of artificial neural network learning can be referred to as labeling. In this case, the training data and the corresponding labels constitute a training set, which can be input to the artificial neural network in the form of a training set. Meanwhile, training data represents multiple features, and labeling the training data can mean that the features represented by the training data are labeled. In this case, the training data can represent the features of the input object in vector form. An artificial neural network can infer a function regarding the relationship between the training data and the labeled data using the training data and the labeled data. Then, the parameters of the artificial neural network can be determined (optimized) through the evaluation of the inferred function. Loss functions can be used as an indicator for evaluating the inferred function. The loss function is a score value generated by receiving the correct answer and the prediction as input, and it serves as an indicator to determine how well the neural network predicted the correct answer.The smaller the value of this loss function, the greater the prediction accuracy of the artificial neural network. The loss function can be appropriately selected depending on the specific problem or data set.

[0047] The Recurrent Neural Network (RNN) model mentioned in this specification refers to a neural network structure designed to handle temporal dependencies in sequential data. This model operates by remembering information from previous time points as input data arrives sequentially, combining it with information from the current time point, and processing it. The input layer of an RNN receives data at each time point, and the input data is typically expressed in vector form. Data passing through the input layer is passed to the hidden layer, where the hidden layer calculates a new hidden state by combining the input from the current time point with the hidden state from the previous time point. During this process, a weight matrix is ​​applied between the input data and the hidden state, and a bias is added to this weight matrix, which is processed as an activation function. Examples of activation functions include the hyperbolic tangent function or the Rectified Linear Unit (ReLU), which introduce nonlinearity and enable the model to learn complex patterns. The new hidden state is used to calculate the output value at the current time point. The output layer uses data from the hidden layer to calculate the final output value, which is then passed to the next time point and reflected in the iterative learning process. For example, the cross-entropy loss function can be used as a loss function, which measures the difference between the model's predicted values ​​and the actual values ​​and is used to update the model's weights in a direction that minimizes the error during the learning process.

[0048] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings.

[0049] FIG. 1 is a schematic drawing of a construction equipment radar detection device (100) that improves radar misrecognition according to one embodiment of the present invention.

[0050] As illustrated in FIG. 1, a construction equipment radar detection device (100) for improving radar misrecognition according to one embodiment of the present invention includes a radar (10), a tilt sensor (20), a turning angle sensor (30), and a detection server (40).

[0051] The radar (10) detects obstacles around construction equipment based on RF transmission and reception. The radar (10) can output information such as the direction, distance, and speed of the obstacle. As a possible example, the radar (10) may be an FMCW (Frequency Modulated Continuous Wave) radar. The FMCW radar measures distance by transmitting a continuously modulated RF signal (Radio Frequency) and analyzing how the frequency of the signal changes while being reflected from the target and returning. At this time, the distance and speed of the target can be measured simultaneously by analyzing the difference in power and frequency of the received signal.

[0052] The inclination sensor (20) detects the inclination of the construction equipment. The inclination sensor (20) may be a combination of two or more of a gyro sensor, an acceleration sensor, and a geomagnetic sensor.

[0053] The turn angle sensor (30) detects the turn angle of construction equipment. Construction equipment can be divided into a lower part for driving and an upper part for operating. The upper part of the construction equipment can rotate 360 ​​degrees relative to the lower part. The turn angle sensor (30) measures the degree to which the upper part of the construction equipment turns relative to the lower part, or the turn angle. As a possible example, the turn angle sensor (30) may be an IMU (Inertial Measurement Unit).

[0054] The detection server (40) communicates with the radar (10), the tilt sensor (20), and the turning angle sensor (30). The detection server (40) determines whether there are obstacles around the construction equipment using information received from the radar (10), the tilt sensor (20), and the turning angle sensor (30). The detection server (40) may further include a processor (42).

[0055] A processor (42) may refer to a data processing device having a physically structured circuit to perform a function expressed by, for example, a code or command included in a program. As such, examples of a data processing device may include a microprocessor, a central processing unit (CPU), a processor core, a multiprocessor, an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), a graphics processing unit (GPU), a neural processing unit, and the like, but the scope of the present invention is not limited thereto.

[0056] The processor (42) may be an artificial intelligence processor that executes a pre-learned artificial intelligence model.

[0057] FIG. 2 is a schematic drawing of a construction equipment (200) equipped with a construction equipment radar detection device (100) illustrated in FIG. 1.

[0058] As illustrated in Fig. 2, the construction equipment (200) can be conveniently divided into an upper part (210) and a lower part (220). The upper part (210) has construction tools for work attached thereto, and a space provided for a worker to ride and drive the construction equipment (200). The lower part (220) is a part for driving the construction equipment (200) and includes a track-type driving body called a caterpillar (221).

[0059] A radar (10) is mounted on the upper part (210) of a construction equipment (200). There may be multiple radars (10), and three radars (10) may be provided on the left, right, and rear sides of the construction equipment. Each radar (10) has its own detection area from the location where it is mounted. When the construction equipment (200) is not turning, the three radars (10) transmit RF signals toward the left, right, and rear of the construction equipment (200), and objects detected by the reflected signals can be detected as obstacles. At this time, when the construction equipment is not turning, the radar detection area of ​​the radar (10) does not include the caterpillar (221), so there is no detection error due to the caterpillar (221).

[0060] The tilt sensor (20) and the turning angle sensor (30) are mounted on the upper part (210) of the construction equipment (200). The mounting positions of the tilt sensor (20) and the turning angle sensor (30) are not particularly limited, but may be mounted on the front or rear part of the upper part (210) of the construction equipment (200).

[0061] Figures 3 and 4 are drawings showing the turning state of the construction equipment (200) illustrated in Figure 2.

[0062] As illustrated in FIGS. 3 and 4, the upper part (210) of the construction equipment (200) rotates, and the plurality of radars (10) mounted on the upper part (210) also rotate together. Accordingly, the detection area of ​​the radar (10) changes. At this time, the caterpillar (221) of the lower part (220) of the construction equipment (200) may be included in the detection area of ​​the radar (10). Comparing FIGS. 2 and 3 together, the caterpillar (221) that was not within the detection area of ​​the radar (10) may be included within the detection area of ​​the radar (10) as the construction equipment (200) turns.

[0063] In an embodiment of the present invention, a construction equipment radar detection device (100) may utilize a radar (10) and a turning angle sensor (30) to minimize self-reflection of the construction equipment. The processor (42) may recognize a signal reflected from the caterpillar (221) according to the turning angle as self-reflection of the construction equipment through machine learning. That is, when the upper part (210) of the construction equipment (200) turns, the radar (10) detects the caterpillar (221) of the lower part (220) of the construction equipment (200). At this time, the self-reflection component by the caterpillar (221) is repeatedly generated at the same location. The processor (42) may repeatedly learn and accumulate the same reflection component that is repeatedly generated at a specific angle as a self-reflection component. In addition, the processor (42) may determine the self-reflection component based on the accumulated learning information.

[0064] The measurement data of the radar (10) determined to be a self-reflective component may be masked within the detection area of ​​the object. Specifically, the measurement coordinate values ​​of the radar (10) determined to be a self-reflective component may be ignored so that they no longer fall within the detection area of ​​the object. More specifically, the processor (42) may not perform additional operations related to obstacle detection and / or avoidance on the measurement data of the radar (10) having the masked coordinate values.

[0065] Through this, it is possible to reduce misrecognition of the radar (110) while also increasing the computational efficiency of the processor (42).

[0066] As described above, in the embodiment of the present invention, the construction equipment radar detection device (100) can distinguish self-reflection by detecting a part of the construction equipment using the radar (10) and the turning angle sensor (30). Accordingly, the construction equipment radar detection device (100) can improve the problem of false obstacle recognition, in which a part of the construction equipment (200), such as the caterpillar (221), is judged as an obstacle.

[0067] FIG. 5 is a graph showing the Doppler characteristics of the radar (10) when the construction equipment (200) illustrated in FIG. 2 rotates.

[0068] As illustrated in FIG. 5, when the upper part of the construction equipment (210) continues to turn, the output information of the radar (10) may include a Doppler frequency. When the construction equipment (200) rotates, a Doppler frequency plane (510) for each azimuth of the radar (10) may be generated. At this time, the detection server (40) may recognize an obstacle at a specific azimuth that exhibits a Doppler characteristic that deviates from the generated Doppler frequency plane (510) as a proximity obstacle. An object located in the direction in which the upper part of the construction equipment (210) rotates outputs a positive Doppler frequency, whereas an object located in the opposite direction to the direction in which the upper part of the construction equipment (210) rotates outputs a negative Doppler frequency. That is, when an object at a specific azimuth exhibits a positive or negative Doppler frequency compared to the Doppler frequency plane (310) based on the azimuth-dependent Doppler frequency plane (510), it may be recognized as a proximity obstacle.

[0069] Therefore, the construction equipment radar detection device (100) according to one embodiment of the present invention has the advantage of being able to detect a nearby obstacle based on the Doppler frequency characteristics even when the radar (10) turns.

[0070] FIG. 6 is a schematic diagram illustrating a case in which a construction equipment radar detection device (100) according to one embodiment of the present invention detects a person (300).

[0071] As illustrated in Fig. 6, a person (300) may be positioned around construction equipment (200). The construction equipment radar detection device (100) may determine an obstacle as a proximity obstacle if the distance to the obstacle measured by the radar (10) is shorter than a proximity distance reference value. Here, the proximity distance reference value may be the maximum working range according to the turning motion of the construction equipment (200).

[0072] The construction equipment radar detection device (100) can detect the Doppler frequency generated when a person (300) moves. Accordingly, the construction equipment radar detection device (100) has the advantage of being able to distinguish between stationary objects and moving objects.

[0073] FIG. 7a and FIG. 7b are schematic drawings illustrating ground reflection of construction equipment (200) in one embodiment of the present invention.

[0074] The construction equipment (200) can operate not only on flat terrain but also on sloped terrain. When the construction equipment (200) is on sloped terrain, one or more of the three radars (10) may change direction to face the ground. The radar (10) may transmit a signal toward the ground, and a ground reflection phenomenon occurs, which detects a reflected component from the ground. Accordingly, a construction equipment radar detection device according to the prior art may misrecognize the ground as an obstacle.

[0075] Fig. 7a shows the distance (L1) between the radar (10) and the ground when the construction equipment (200) is located on a sloped terrain (slope). Fig. 7b shows the case when the construction equipment (200) is located on a flat terrain and a person (300) is located at a certain distance (L2) from the radar (10). In Figs. 7a and 7b, the radar (10) can determine that an object is located at the same distance (L1=L2). However, in Fig. 7a, a reflection signal is detected in a wide area, whereas in Fig. 7b, a reflection signal is detected in a narrower area than in Fig. 7a. Accordingly, the construction equipment radar detection device (100) can distinguish between reflection by the ground and reflection by an obstacle (person).

[0076] Meanwhile, in a possible embodiment of the present invention, the construction equipment radar detection device (100) can improve the accuracy of the measurement data of the radar (10) by applying a Kalman filter to the data measured by each sensor.

[0077] The slope data measured by the slope sensor (20) on the slope is used to correct the measurement data of the radar (10). Since the radar (10) measures the straight-line distance to an object, the actual distance between the construction equipment (200) extending along the slope and the object needs to be corrected based on the slope data measured by the slope sensor (20).

[0078] For example, if the slope value measured by the slope sensor (20) on a slope is a (degree), the actual distance to the object can be calculated by multiplying the cosine value of a by the distance measured by the radar (10). At this time, if an error occurs in the slope angle, which is the measurement data of the slope sensor (20), the distance value corrected with the slope angle with an error will also inevitably have an error, no matter how accurate the measurement data of the radar (10).

[0079] The present invention can apply a Kalman filter to the measurement data of the tilt sensor (20) in order to remove noise corresponding to an error among the measurement data obtained from the tilt sensor (20).

[0080] Meanwhile, the turning angle sensor (30) also affects the reliability of the measurement data of the radar (10). Specifically, when the turning angle of the construction equipment (200) is determined based on the measurement value of the turning angle sensor (30), the radar facing the slope is changed. For example, when the construction equipment (200) is located on a slope and does not turn at the same time, the rear radar becomes the primary radar to detect the slope. For example, when the construction equipment (200) is located on a slope and turns 30 degrees to the right at the same time, the right radar, not the rear radar, will face down the slope. Therefore, in this case, the right radar must become the primary radar to detect the slope. That is, the processor (42) selects the primary radar to detect the slope according to the turning angle.

[0081] At this time, if there is an error in the rotation angle, which is the measurement data of the turning angle sensor (30), an error will also occur in the selection of the main radar, and no matter how accurate the measurement data of the radar (10) is, an error is bound to occur in the judgment because the judgment of whether or not there is a slope is made based on unnecessary data.

[0082] In addition, according to the present invention, self-reflection of construction equipment (200) can be distinguished based on the rotation angle measured by the rotation angle sensor (30), but if there is an error in the rotation angle, which is the measurement data of the rotation angle sensor (30), an error also occurs in the self-reflection judgment. For example, if the rotation angle is 30 degrees but an error occurs as 20 degrees, radar (10) measurement data of incorrect coordinates may be judged as self-reflection.

[0083] An embodiment of the present invention can improve the reliability of measurement data of a radar (10) by removing noise from a tilt sensor (20) and / or a turning angle sensor (30).

[0084] As is well known, the Kalman filter is an algorithm that optimally estimates state from noisy data. Primarily used in sensor data processing, it combines state predictions and measurements to remove noise and provide accurate estimates. The Kalman filter operates by estimating future states during the prediction phase and then comparing them with actual measurements during the update phase, revising the state estimates.

[0085] In the case of the tilt sensor (20) and the turning angle sensor (30), the measurement data does not fundamentally change rapidly. Therefore, by applying a Kalman filter to the data of each sensor, values ​​that change rapidly can be corrected.

[0086] As a possible example, in the case of the tilt sensor (20) and the turning angle sensor (30), the state vector to be applied to the Kalman filter may include an angle value and an angular velocity value. In this case, the angle may be a measurement vector. In addition, the state transition matrix may be configured to represent a state vector that changes over time. In addition, the observation matrix may play a role in connecting the state vector and the measurement value of the system, and may select only measurable elements of the state vector.

[0087] In addition, the values ​​to be substituted for the initial state vector, initial error covariance matrix, process noise covariance matrix, measurement noise covariance matrix, sampling cycle time, etc. can be empirically and experimentally determined based on the noise characteristics of the sensor, the complexity of the system, the precision of the system design, etc.

[0088] In a possible embodiment of the present invention, a Kalman filter may be applied to data from radar (10). Since the data from radar (10) also does not change rapidly, noise can be removed by applying a Kalman filter. In addition, applying a Kalman filter to data from radar (10) can reduce false detections due to protrusions when detecting the ground. Since data is input from a wide range of the ground, and data values ​​change significantly when a protrusion exists in the ground, applying a Kalman filter to the measured data from radar (10) can eliminate the influence of the protrusions, thereby reducing errors in detecting the ground.

[0089] At this time, the measurement data of the radar (10) to which the Kalman filter is to be applied may be obtained based on the data of the tilt sensor (20) and / or the turning angle sensor (30) from which noise has been removed by first applying the Kalman filter.

[0090] That is, the measurement data of the radar (10) is first selected by applying a Kalman filter to the measurement data of the tilt sensor (20) and / or the turning angle sensor (30), and the Kalman filter is secondarily applied to the first-selected radar (10) data, and the second-selected radar (10) data can be used to detect an object.

[0091] As a possible example, in the case of a radar sensor (10), the state vector to be applied to the Kalman filter may include coordinate values ​​and power values. In this case, in the case of the radar sensor (10), both the coordinate values ​​and the power values ​​may be measurement vectors. In addition, the state transition matrix may be configured to represent a state vector that changes over time. In addition, the observation matrix may play a role in connecting the state vector and measurement values ​​of the system, and may select only measurable elements of the state vector.

[0092] In addition, the values ​​to be substituted for the initial state vector, initial error covariance matrix, process noise covariance matrix, measurement noise covariance matrix, sampling cycle time, etc. can be empirically and experimentally determined based on the noise characteristics of the sensor, the complexity of the system, the precision of the system design, etc.

[0093] Meanwhile, more detailed information and formulas related to the meaning of each matrix in the Kalman filter, Kalman gain calculation, state update, and error covariance update are already well-known techniques, so their explanation is omitted here.

[0094] Through this configuration, the data reliability of the radar (10) can be greatly improved and the object detection accuracy can also be further improved.

[0095] The processor (42) can repeatedly learn and accumulate ground reflections based on the values ​​of the inclination sensor (120). In addition, the processor (42) can increase the recognition accuracy of the radar (110) by determining the ground reflection component based on the accumulated learning information.

[0096] Meanwhile, the reflection, reflection signal, and reflection component of the radar (10) repeatedly mentioned in this specification are signals received when a signal transmitted from the radar (10) is reflected by an object, and may also be referred to as a measurement signal, measurement data, output signal, output data, output information, etc.

[0097] FIG. 8 is a drawing illustrating a case where a construction equipment radar detection device (100) according to one embodiment of the present invention detects the ground.

[0098] In Fig. 8, since the construction equipment (200) is located in an environment with a high front and a low rear slope, the ground reflection component can be detected from the radar (10) mounted at the rear of the construction equipment (200).

[0099] In an embodiment of the present invention, a construction equipment radar detection device (100) can minimize the influence of ground reflection components by using a radar (10) and a tilt sensor (20).

[0100] Under conditions where the construction equipment (200) is tilted, the reflected signal of the radar (10) can enter a wide area. At this time, the reflected component measured at the slope of the set value according to the value of the tilt sensor (20) can be determined as a reflection by the ground. If the slope measured from the tilt sensor (20) is equal to or greater than a reference value, the detection server (40) can determine that a shorter distance than the distance from the flat ground to the ground is the ground. The processor (42) can continuously and repeatedly learn and accumulate the ground reflection component measured in various ground slope environments. The construction equipment radar detection device (100) can determine the ground reflection component based on the learning information accumulated by the processor (42) and determine this ground reflection component as the ground.

[0101] Meanwhile, the distance to the ground according to the slope value learned by the processor (42) can be set as the ground distance reference value. The ground distance reference value may be a value that changes according to the slope of the ground.

[0102] FIG. 9 is a schematic diagram illustrating a case where a construction equipment radar detection device (100) according to one embodiment of the present invention detects the ground and a person simultaneously.

[0103] As illustrated in Fig. 9, the construction equipment radar detection device (100) recognizes an object located closer than the ground distance reference value as a proximity obstacle when the slope is greater than a reference value. Furthermore, the construction equipment radar detection device (100) can detect the Doppler frequency of the proximity obstacle. Accordingly, the construction equipment radar detection device (100) can distinguish and recognize reflections from the ground and people (300) even when the construction equipment (200) is located on a sloped terrain.

[0104] As described above, in a possible embodiment of the present invention, the processor (42) can determine through learning whether the reflection component of the radar (10) is a ground reflection component.

[0105] For example, the processor (42) can train an artificial intelligence model. In addition, the processor (42) can execute the trained artificial intelligence model to determine whether the reflection component of the radar (10) is a ground reflection component.

[0106] Recurrent Neural Networks (RNNs) can be used to train AI models. As is well known, RNNs are artificial neural networks that process sequential data, such as time series data. They can learn temporal patterns by incorporating information from previous time points into current calculations. At each time step, RNNs calculate the current state based on the input and previous hidden states, which are then used as input for the next time step.

[0107] In a possible embodiment of the present invention, a learning data set for training a neural network that determines ground reflection components may include measurement data of a tilt sensor (20) and measurement data (coordinates, power value, distance) of a radar (10).

[0108] In the preprocessing step, the learning data set can be labeled as the ground or obstacles (people, objects, walls, etc.). For example, the ground and obstacles can be distinguished, judged, and labeled based on whether the measurement data of the radar (10) has different power values ​​at the same coordinates (standing objects / people), different power values ​​at different coordinates (a wide area spanning the radar's azimuth) (ground), the same power value at different coordinates (wall), or whether the construction equipment (200) is located on a slope higher than a specific slope. Here, the degree of variation in coordinate values ​​and the degree of variation in power values ​​for the determination of identity can be used for learning through prior testing.

[0109] In the preprocessing process, noise can be removed from the sensor data constituting the learning data set using a Kalman filter. The correction of measurement data from the tilt sensor (20), the turn angle sensor (30), and the radar (10) using a Kalman filter has already been described above. In particular, the data from the radar (10) to be used for learning can be re-selected by first applying a Kalman filter to the tilt sensor (20) and the turn angle sensor (30), and then secondly applying a Kalman filter to the selected data.

[0110] Through this preprocessing process, neural network training can be performed faster and more efficiently, and model performance can be improved.

[0111] When using a learned model, measurements from a tilt sensor (20), measurements from a turning angle sensor (30), and measurements from a radar (10) can be used as input data. At this time, output data can output whether the object detected by the radar (10) is the ground or an obstacle, and if it is an obstacle, the location of the obstacle.

[0112] As described above, in a possible embodiment of the present invention, the processor (42) can determine through learning whether the reflection component of the radar (10) is a self-reflection component.

[0113] In a possible embodiment of the present invention, a learning data set for training a neural network that determines self-reflection components may be composed of measurement data of a turning angle sensor (30) and measurement data (coordinates, power values, distances) of a radar (10). In a preprocessing step, the learning data set may be labeled as a caterpillar or an obstacle. For example, caterpillars may be distinguished, determined, and labeled based on whether the same reflection component (same power value and coordinates) occurs repeatedly more than a preset number of times at a specific turning angle. Here, the degree of variation in coordinate values, the degree of variation in power values, etc. for determining identity may be used for learning through prior testing.

[0114] In the preprocessing process, noise can be removed from the sensor data constituting the learning data set using a Kalman filter. The correction of measurement data from the tilt sensor (20), the turn angle sensor (30), and the radar (10) using a Kalman filter has already been described above. In particular, the radar (10) data to be used for learning may be re-selected by first applying a Kalman filter to the turn angle sensor (30) and then secondarily applying a Kalman filter to the selected data.

[0115] Through this preprocessing process, neural network training can be performed faster and more efficiently, and model performance can be improved.

[0116] When using a learned model, the measurement values ​​of the turning angle sensor (30) and the measurement values ​​of the radar (10) can be used as input data. At this time, as output data, whether the object detected by the radar (10) is a caterpillar or an obstacle, and if it is an obstacle, the location of the obstacle can be output.

[0117] FIG. 10 is a flowchart for explaining a ground detection method of a construction equipment radar detection device (100) according to one embodiment of the present invention.

[0118] As illustrated in Fig. 10, the detection server (40) can detect the approach of an object based on the output information of the radar (10) (S11). The radar (10) is mounted on the left, right, and rear of the construction equipment (200) and transmits RF signals, and can detect objects based on reflected signals.

[0119] In this step (S11), in order to improve the reliability of the reflected signal (measurement data) of the radar (10), a Kalman filter may be applied to the measurement data of the tilt sensor (20) and the turn angle sensor (30). In addition, a Kalman filter may also be applied to the measurement data of the radar (10). The measurement data of the radar (10) may be first screened by applying a Kalman filter to the measurement data of the tilt sensor (20) and the turn angle sensor (30), and then secondly screened by applying a Kalman filter to the screened data. Since this has been described in detail in the description of the construction equipment radar detection device (100) above, a duplicate description will be omitted here.

[0120] Next, the detection server (40) can determine whether the detected object is a wide-area object (S12). The detection server (40) can determine that objects simultaneously detected within the azimuth range of the radar (10) are wide-area objects. The detection server (40) can determine that objects detected only in a portion of the azimuth range of the radar (10) are obstacles.

[0121] Next, if the detection server (40) determines that the detected object is a wide-area object, it can determine whether the slope of the ground is greater than or equal to a slope reference value (S13). The slope of the ground can be measured by a slope sensor (20). The slope reference value can include values ​​greater than or equal to 10 degrees and less than or equal to 20 degrees, and for example, can be 15 degrees.

[0122] The detection server (40) can determine that the detected object is the ground if the slope of the ground is greater than the slope reference value (S14). In addition, the ground reflection can be learned and accumulated by the processor (42) (S15).

[0123] The detection server (40) can determine that the detected object is an approaching obstacle when the slope of the ground is less than the slope reference value (S16).

[0124] FIG. 11 is a flowchart for explaining a self-reflection detection method of a construction equipment radar detection device (100) according to one embodiment of the present invention.

[0125] As illustrated in Fig. 11, the detection server (40) can detect the approach of an object based on the output information of the radar (10) (S21). The radar (10) is mounted on the left, right, and rear of the construction equipment (200) and transmits RF signals, and can detect objects based on reflected signals.

[0126] In this step (S21), in order to improve the reliability of the reflection signal (measurement data) of the radar (10), a Kalman filter may be applied to the measurement data of the turn angle sensor (30). In addition, a Kalman filter may also be applied to the measurement data of the radar (10). The measurement data of the radar (10) may be first screened by applying a Kalman filter to the measurement data of the turn angle sensor (30), and then secondarily screened by applying a Kalman filter to the screened data. Since this has been described in detail in the description of the construction equipment radar detection device (100) above, a duplicate description will be omitted here. Next, the detection server (40) may determine whether the detected object is repeatedly detected at the same location (S22). The detection server (40) determines whether the detected object is repeatedly detected with the same reflection component at a specific turn angle using the value output from the turn angle sensor (30). Here, “repetitive” may mean more than a preset specific number of times.

[0127] If the detection server (40) determines that the detected object is repeatedly detected at the same location, it can determine that the detected object is a self-reflection of the construction equipment (200) (S23). Then, the self-reflection can be learned and accumulated by the processor (42) (S24).

[0128] The detection server (40) can determine that the detected object is an obstacle if it is determined that the detected object is not repeatedly detected at the same location (S25).

[0129] As described above, the construction equipment radar detection device (100) according to one embodiment of the present invention can efficiently detect ground reflection and self-reflection using the radar (10), the tilt sensor (20), and the turning angle sensor (30) without having an expensive camera, thereby improving obstacle detection performance.

[0130] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will readily appreciate that the present invention can be readily modified into other specific forms without altering the technical spirit or essential characteristics of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single entity may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a combined manner.

[0131] The scope of the present invention is indicated by the claims set forth below, and all changes or modifications derived from the meaning and scope of the claims and their equivalent concepts should be interpreted as being included in the scope of the present invention.

Claims

1. In a construction equipment radar detection device, Radar for detecting objects around the above construction equipment; A tilt sensor for detecting the tilt of the above construction equipment; A turning angle sensor for detecting the turning angle of the above construction equipment; and A detection server that communicates with the radar, the tilt sensor, and the turning angle sensor and determines the presence of an object around the construction equipment; Including, The above detection server, The presence of an object around the construction equipment is determined by using the inclination value received through the inclination sensor and the reflection component detected through the radar. If the detected object is determined to be a wide-ranging object, the object is determined to be the ground. The measurement values ​​of the inclination sensor and the turning angle sensor are corrected by applying a Kalman filter, and the reflection component of the radar for determining the presence of the object is selected based on the measured values ​​of the corrected inclination sensor and the turning angle sensor. Construction equipment radar detection device that improves radar misrecognition.

2. In paragraph 1, The above wide range of objects is characterized by being objects that are simultaneously detected within the azimuth range of the radar. Construction equipment radar detection device that improves radar misrecognition.

3. In paragraph 1, The above detection server, A method characterized in that, if the inclination value received through the inclination sensor is greater than or equal to a preset inclination reference value, the detected object is judged to be the ground. Construction equipment radar detection device that improves radar misrecognition.

4. In paragraph 3, The above detection server further comprises a processor, The above processor repeatedly learns and accumulates the process of determining the detected object as the ground, and is characterized in that it determines the object as the ground based on the accumulated learning information. Construction equipment radar detection device that improves radar misrecognition.

5. In paragraph 1, The radar reflection component for determining the presence or absence of the object is characterized in that it is a reflection component that is secondarily selected by applying a Kalman filter to the reflection component that is firstly selected based on the measured values ​​of the calibrated inclination sensor and the turning angle sensor. Construction equipment radar detection device that improves radar misrecognition.

6. In construction equipment radar detection devices, Radar for detecting objects around the above construction equipment; A turning angle sensor for detecting the turning angle of the above construction equipment; and A detection server that communicates with the radar and the turning angle sensor and determines the presence of objects around the construction equipment; Including, The above detection server, The presence of an object around the construction equipment is determined by using the turning angle value received through the turning angle sensor and the reflection component detected through the radar. If the detected object is repeatedly detected at the same location, the object is determined to be a self-reflection of the construction equipment. The measurement value of the above turning angle sensor is corrected by applying a Kalman filter, and the reflection component of the radar for determining the presence of the object is selected based on the measured value of the corrected turning angle sensor. Construction equipment radar detection device that improves radar misrecognition.

7. In paragraph 6, The above detection server further comprises a processor, A construction equipment radar detection device for improving radar misrecognition, characterized in that the processor repeatedly learns and accumulates a process of determining the detected object as a self-reflection of the construction equipment, and determines the detected object as a self-reflection of the construction equipment based on the accumulated learning information.

8. In paragraph 6, A construction equipment radar detection device with improved radar misrecognition, characterized in that the radar outputs a Doppler frequency generated according to a turning motion of the construction equipment, and recognizes a nearby obstacle when the Doppler frequency exhibits a plus or minus Doppler frequency.

9. In a method for detecting obstacles using a construction equipment radar detection device that improves radar misrecognition, A step in which the detection server detects the approach of an object based on the output information of the radar; A step in which the above detection server determines whether the detected object is a wide-range object; A step in which the above detection server determines whether the slope of the ground is greater than a preset reference value; If the detected object is a wide-area object and the slope of the ground is determined to be greater than a preset reference value, the step of the detection server determining the detected object as the ground Including, In the step of detecting the approach of the above object, A method for detecting an obstacle using a construction equipment radar detection device with improved radar misrecognition, characterized in that the slope of the ground is measured using a slope sensor, the turning angle of the construction equipment is measured using a turning angle sensor, the measurement values ​​of the slope sensor and the turning angle sensor are corrected by applying a Kalman filter, and the output information of the radar for detecting the approach of the object is selected based on the measured values ​​of the corrected slope sensor and the turning angle sensor.

10. In paragraph 9, At the step of determining whether the above-mentioned wide-ranging object is The above detection server determines that objects detected simultaneously within the azimuth range of the radar are the wide-range objects. An obstacle detection method using a construction equipment radar detection device having improved radar misrecognition characteristics.

Images