67 results about "Machine perception" patented technology

The invention discloses an industrial automobile crane positioning and mapping method and device based on multi-sensing fusion, and relates to the field of machine sensing, and the method comprises the following steps: initializing a pose and offset of an inertial measurement unit by using an input laser point cloud and a pre-integration result of the inertial measurement unit; correcting laser point cloud motion distortion by using an inertial measurement unit, extracting a dynamic object point cloud, and synchronously extracting laser features and visual features with depth; calculating a key frame pose in real time by minimizing a visual re-projection error and an inertial measurement unit pre-integration error, and performing frame-to-map matching by taking the key frame pose as an initial value to further optimize the pose; a visual word bag model is utilized to quickly retrieve a closed-loop candidate frame, after laser passes geometric matching to verify a closed loop, a laser radar factor, a visual factor, an inertial measurement unit factor and a closed-loop factor are fused, a global factor graph is constructed and optimized, and a pose and point cloud map is updated. According to the method, the environment sensing capability of the industrial automobile crane can be improved, and the sensing accuracy and robustness are improved.

The present invention provides a shared control obstacle avoidance method based on brain-machine perception fusion, which includes an environmental perception module, a perception fusion module, and a path planning module. The environmental perception module uses a brain-computer interface and a lidar to perceive the environment and obtain the environmental information where the robot is located. The perception fusion module fuses the perception of the brain-computer interface and the lidar according to weights to obtain the environmental information required for the robot path planning. The path planning control module completes the intelligent obstacle avoidance of the robot. The beneficial effects of the present invention: The shared control obstacle avoidance method based on brain-machine perception fusion fuses the biological perception of humans and the machine perception of robots in the environment to complete multi-modal environmental perception, effectively improving the ability of humans and robots to jointly perceive and make decisions, and at the same time avoiding the cumbersome control of the brain-computer interface in various scenarios.

A loading system is constructed to load a hauling machine by a loading machine. The system includes a fiducial marker affixed to the hauling machine. A set of sensors of diverse sensory modalities is deployed on the loading machine that generate respective signals indicative of the fiducial marker in the corresponding sensor modalities. A processor on the loading machine determines a machine orientation in space of the hauling machine through machine perception applied to the signals. Loading the hauling machine by the loading machine is guided according to the machine orientation.

Disclosed is a system to influence detectability of a target object by machine perception. The system receives input data that includes a representation of the target object and an objective corresponding to reduce detectability or induce a designated detection of that object. The system generates a plurality of candidate adversarial patterns based on the objective for the object. For each candidate adversarial pattern, the system simulates an integration of the target object in a virtual environment across a plurality of sensor modalities. The system evaluates each simulated pattern against a plurality of sensor modalities and calculates a performance metric for each candidate. The system determines whether the metric reaches a predetermined threshold level for the objective. If not, it iteratively adjusts the candidate to converge on an optimized output adversarial pattern. If so, it confirms the pattern as optimized and compiles it for output.

The present disclosure relates to Communicational and Conversational Artificial Intelligence, Machine Perception, Perceptual-User-Interface, and a professional training method. A chatbot may comprise at least one skills module. The chatbot engages with trainee(s) on communicational training on a subject matter provided by the skills module. A trainer may create, remove, or update a skills module with interaction skills and training materials through an onboarding module. A trainee can upload recorded interactions to a skills module for evaluation or for role playing an interaction without a trainer or partner. An administrator may monitor a trainee's performance, and correlate with the organization's metrics. Based on the evaluation, the trainer or chatbot may provide the trainee with feedback and recommended improvement plans. The chatbot may be implemented in an Internet-of-Things or any device. The subject matters may extend to cover different industries / markets. The trainer may be a professional or individual.

The disclosed semiconductor device package may include a compute chip configured to perform contextual artificial intelligence and machine perception operations. The disclosed semiconductor device package may additionally include a sensor positioned above the compute chip in the semiconductor device package. The disclosed semiconductor device package may also include one or more electrical connections configured to facilitate communication between the compute chip and the sensor, between the compute chip and a printed circuit board, and between the sensor and the printed circuit board. Various other methods, systems, and computer-readable media are also disclosed.

A real-time machine perception system and vehicles with real-time perception systems can be adapted for navigation in conflict zone environments and include an optical sensor and an optical processing component coupled to the optical sensor. The optical processing component can be configured to solve a non-linear control problem comprising solving a multi-vehicle merging or multi-vehicle synchronization problem. The optical processing component can perform optical processing on a signal from the optical sensor according to a deep learning algorithm having weights determined by solving the non-linear control problem.

A method and device for performing a perception task are disclosed. The method and device incorporate a dense regression model. The dense regression model advantageously incorporates a distortion-free convolution technique that is designed to accommodate and appropriately handle the varying levels of distortion in omnidirectional images across different regions. In addition to distortion-free convolution, the dense regression model further utilizes a transformer that incorporates an spherical self-attention that use distortion-free image embedding to compute an appearance attention and uses spherical distance to compute a positional attention.

A controller of a machine may identify that an implement and a linkage of the machine are in a first position. The controller may determine, based on identifying that the implement and the linkage are in the first position, that the implement and the linkage are not blocking a first perception zone that is associated with a first set of one or more sensors of a perception sensor system of the machine. The controller may cause, based on determining that the implement and the linkage are not blocking the first perception zone, first perception data captured by the first set of one or more sensors to be used in association with an automatic control operation.

The application discloses a low-light target detection method based on a double-layer optimization theory, which significantly improves the enhancement quality and detection accuracy. The method first adopts a double-layer learning architecture to bridge the enhancement and detection tasks, and organically combines the low-light image enhancement task and the target detection task. Secondly, the enhancement network parameters are taken as the main optimization target, and the detection network parameters are taken as the constraint condition. The enhancement network generates images that are conducive to machine perception. Finally, in order to improve the training efficiency, a method based on the implicit function theory is proposed to approximately solve the corresponding gradient. The training cost is effectively reduced, and the training speed is improved. The application significantly improves the target detection accuracy in weak light scenes, and can be applied to various target detection tasks in weak light scenes, such as night pedestrian recognition, automatic driving, security monitoring and the like.

The invention provides an accumulated drift error correction method for point cloud pose motion based on PTZ attitude priori, and belongs to the technical field of machine vision and machine perception. Through multi-modal data acquisition and preprocessing, initial pose constraint construction based on PTZ prior, semantic enhancement loopback detection, semantic consistency registration and residual construction, global graph optimization and drift correction and map reconstruction, accurate correction of point cloud pose accumulated drift is realized. The method has the capabilities of improving the loopback detection robustness, realizing adaptive drift correction and remarkably improving the global precision in a complex scene.

The invention discloses a token stream guide variable code rate image compression method oriented to unification of perception and understanding. The method comprises the following steps of: 1, acquiring a two-dimensional image, and processing the two-dimensional image to acquire a one-dimensional token sequence; 2, processing the one-dimensional token sequence through a variable token mask to generate a binary compressed bit stream; 3, decoding the compressed bit stream to obtain an unmasked token sequence, and carrying out dynamic prediction on the unmasked token sequence to recover a complete token sequence; and 4, realizing human perception through the receiving end I after the complete token sequence is recovered, realizing machine perception through the receiving end II after the complete token sequence is recovered, and finally realizing image compression. The method has the capability that a single model supports the continuous variable bit rate, supports a large language model to directly perform semantic understanding based on the compressed code stream, and has the characteristics of low calculation complexity and low delay.

The invention provides a quality marking method and system for a machine vision system, a medium and equipment, and the method comprises the steps: determining a reference / distortion image pair; under a preset temperature parameter, a machine subject is adopted to execute preset downstream tasks on the reference / distortion image pair, and a machine perception quality score corresponding to each preset downstream task is determined; dimension division is carried out on the machine perception quality score corresponding to each preset downstream task, and a score corresponding to each dimension is determined; and according to the score corresponding to each dimension, an average opinion score is determined, and the average opinion score represents the preference comprehensive representation of the machine subject to the same distorted image. According to the method and the device, quality marking is carried out on the machine-oriented vision system, the establishment of a machine-oriented image quality evaluation data set is facilitated, the consistency and the reliability of data are ensured, and the development of a machine-oriented image quality evaluation technology is driven.

The present application relates to the technical field of machine perception, and specifically proposes a recessed target station recognition method and system with multiple detector cooperation. The method comprises: acquiring point cloud data accumulated by a laser radar in a world coordinate system and a region of interest of a calibration work area, preprocessing the point cloud data to obtain preprocessed point cloud data; running at least two types of heterogeneous detectors in parallel on the preprocessed point cloud data, and each detector independently outputs a candidate oriented bounding box of a recessed target station; calculating a consistency index, a confidence index and a prior matching degree index for each detector output candidate oriented bounding box, and performing forward normalization on the consistency index and the confidence index, and performing reverse normalization on the prior matching degree index; weighted sum of the three types of normalized indexes to obtain the comprehensive score of each detector, and the final station oriented bounding box is determined by optimization or fusion according to the comprehensive score, and the center, orientation, size and depth of the station opening are output.

A system and method are disclosed herein for developing a machine perception model in the omnidirectional image domain. The system and method utilize the knowledge distillation process to transfer and adapt knowledge from the perspective projection image domain to the omnidirectional image domain. A teacher model is pre-trained to perform the machine perception task in the perspective projection image. A student model is trained by adapting the pre-existing knowledge of the teacher model from the perspective projection image domain to the omnidirectional image domain. By way of this training, the student model learns to perform the same machine perception task, except in the omnidirectional image domain, using limited or no suitably labeled training data in the omnidirectional image domain.

A stakeout system is constructed to stakeout a worksite according to a site plan. The system includes a set of fiducial markers, each being associated with a corresponding construction operation performed at the worksite. A set of sensors having diverse sensory modalities generate respective signals from which information associated with the construction operation is conveyed. A processor is constructed to ascertain from the signals a location of a work machine at the worksite, as well as the corresponding construction operation performed thereat. The work machine is guided in performing the construction operation according to the location of the work machine in the site plan.

The invention belongs to the technical field of flexible sensors, and particularly provides an ultra-sensitive flexible temperature sensor based on a fluff structure and a preparation method thereof, which are used for improving the signal response strength and response speed of the temperature sensor. The ultra-sensitive flexible temperature sensor comprises a flexible electronic skin and a villus sensing structure, wherein the villus sensing structure is vertically arranged on the flexible electronic skin; the bionic fluff three-dimensional structure design is adopted, the thermoelectric material is arranged on the bionic fluff structure to form the fluff sensing structure, the temperature of the external environment (airflow) can be rapidly sensed through the slender fluff structure, the sensitivity and precision of the device are greatly improved, meanwhile, the structural complexity is reduced, the size of the sensor is effectively reduced, and the cost is reduced. Therefore, the method is suitable for more application scenes, such as machine perception and man-machine interaction.

The utility model relates to a bridge girder erection machine perception early warning system, and belongs to the technical field of bridge girder erection machine monitoring, and the early warning system comprises a sensor assembly which is installed on a bridge girder erection machine and is used for detecting the working state of the bridge girder erection machine and sending state data; the controller is used for receiving the state data and sending an alarm signal when the state data is abnormal; the alarm responds to the alarm signal and executes an alarm action; the face recognizer is mounted on the bridge girder erection machine and is used for acquiring a face image, determining whether an operator is an operator or not according to the face image and sending a correct signal when the operator is the operator; after the controller receives the correct signal, the timer is controlled to time, and the timing duration is fed back; the unlocking structure is used for sending an unlocking signal after the controller determines that the timing duration reaches a set duration threshold value, and the unlocking structure responds to the unlocking signal and unlocks the operation room; the duration threshold value is the average duration of historical comprehensive inspection of the bridge erecting machine. The method has the beneficial effect of reducing potential safety hazards.

This application provides a differentiated compression method, apparatus, device, and medium for 3DGS in autonomous driving, including: establishing a mapping relationship between Gaussian primitives and macroscopic semantic attributes from the original observation images and 3DGS scene model of the autonomous driving scene, and determining the comprehensive semantic weight of each Gaussian primitive; constructing a global optimization objective function with perception task accuracy as an explicit constraint, and determining the perception sensitivity of Gaussian primitives to the task based on the global optimization objective function; dividing the Gaussian primitives into three physical energy levels based on the comprehensive semantic weight and perception sensitivity, and performing differentiated encoding. Through three progressive steps of "semantic weight construction - perception task accuracy optimization - differentiated encoding", a complete 3DGS compression technology chain for machine perception is constructed, which improves machine perception accuracy, realizes dynamic intelligent scheduling of bit resources, and significantly improves storage and bandwidth utilization efficiency.

A loading system (105) constructed to load a hauling machine (136) by a loading machine (138). The system (105) includes a fiducial marker (134) affixed to the hauling machine (136). A set (152) of sensors of diverse sensory modalities is deployed on the loading machine (138) that generate respective signals indicative of the fiducial marker (134) in the corresponding sensor modalities. A processor (156) on the loading machine (138) determines a machine orientation (207) in space of the hauling machine (136) through machine perception applied to the signals. Loading the hauling machine (136) by the loading machine (138) is guided according to the machine orientation (207).

The application relates to the technical field of low-altitude aircraft navigation and control, and discloses a low-altitude aircraft cooperative obstacle avoidance and route dynamic adjustment system, which comprises a multi-machine perception networking module, a global obstacle identification module, a cooperative decision planning module, a route dynamic adjustment module and an execution feedback module, and each module is connected through low-altitude self-networking to establish bidirectional real-time communication connection. Through the multi-machine perception networking module, the low-altitude self-networking is realized, millisecond-level interactive sharing of aircraft perception data is realized, a global perception data pool is constructed, and the limitation of the sensing range of a single aircraft is broken through; combined with the basic geographic information of the low-altitude geographic information database, the sensing deviation is corrected through a multi-sensor data fusion algorithm, static obstacles, dynamic obstacles and temporary no-fly areas are accurately identified, the motion trajectory, speed and heading of the dynamic obstacles are predicted, a comprehensive and accurate global obstacle analysis report is generated, and reliable data support is provided for obstacle avoidance decision.

The present disclosure relates to an image encoding / decoding method and apparatus for using optimized quantization according to an image, a method for transmitting a bitstream, and a bitstream storage medium. An image data encoding method according to one embodiment of the present disclosure comprises the steps of: acquiring image use information for image data from a bitstream; and performing inverse quantization to reconstruct the image data based on image usage information capable of indicating human perception and / or machine perception of the image data.

Example aspects of the present disclosure relate to an example computer-implemented method for data annotation for training machine perception models. The example method can include (a) receiving source sensor data descriptive of an object, the source sensor data having a source reference frame of at least three dimensions, wherein the source sensor data includes annotated data associated with the object; (b) receiving target sensor data descriptive of the object, the target sensor data having a target reference frame of at least two dimensions; (c) providing an input to a machine-learned boundary recognition model, wherein the input includes the target sensor data and a projection of the source sensor data into the target reference frame; and (d) determining, using the machine-learned boundary recognition model, a bounded portion of the target sensor data, wherein the bounded portion indicates a subset of the target sensor data descriptive of the object.

This invention relates to the field of noise monitoring and intelligent analysis technology, and discloses a method and system for intelligent noise source tracing based on human-machine perception alignment. The method includes acquiring acoustic signals and public perception data from the acoustic environment of a social living area; performing time warping on both using a cross-similarity matrix to obtain a perception feature set; analyzing the perception feature set to obtain a deviation distribution and quantifying it into a global deviation correction coefficient; using this coefficient to calibrate the acoustic signal to obtain a signal flow; when the coefficient exceeds a threshold, reconstructing the spatial gain distribution of the microphone array based on the signal flow to locate the noise; combining the location coordinates with image data to identify the type of abnormal event; if it belongs to multi-source interference, predicting the noise intensity trend and generating a real-time parameter optimization scheme; adjusting the detection threshold accordingly and combining the signal flow to pinpoint the final noise source. This method can solve the problems of misalignment between early warning strategies and public perception, as well as the low reliability of noise source tracing.

The invention relates to a machine visual angle video quality evaluation system based on human and machine dual-system perception, and belongs to the field of video quality evaluation. The system comprises a human perception evaluation subsystem for obtaining a video collected by external equipment, performing multi-dimensional quality scoring and distortion type identification on the video, and obtaining a video quality evaluation result; generating a specific score and overall quality evaluation of human perception multi-dimensional evaluation, and outputting a distortion type report; the machine perception evaluation subsystem is used for executing multiple types of tasks through a large model and a specific model according to machine task suitability to obtain human perception evaluation and machine perception evaluation results of the video output by the large model and machine perception scores in a specific task scene; and the comprehensive evaluation result generation subsystem dynamically adjusts the weights of human perception and machine perception according to the application scene, and fuses the weights to generate a comprehensive evaluation result. According to the system disclosed by the invention, a dual-system collaborative evaluation framework is constructed, so that the comprehensive and accurate evaluation of the video quality is realized.

A stakeout system is constructed to stakeout a worksite according to a site plan. The system includes a set of fiducial markers, each being associated with a corresponding construction operation performed at the worksite. A set of sensors having diverse sensory modalities generate respective signals from which information associated with the construction operation is conveyed. A processor is constructed to ascertain from the signals a location of a work machine at the worksite, as well as the corresponding construction operation performed thereat. The work machine is guided in performing the construction operation according to the location of the work machine in the site plan.

The application discloses a roadside parking high-low fusion multi-modal recognition method and system, and belongs to the technical field of roadside parking management, and specifically comprises the following steps: collecting radar and visual data in real time through a curb machine perception terminal, and generating a blind area trigger signal when detecting data conflict; based on the trigger time, extracting a video sequence before the vehicle enters the blind area, and extracting a vehicle feature vector with space-time correlation; calling a high-level monitoring system video stream, and realizing vehicle re-identification and continuous tracking through a feature matching algorithm; after the vehicle is parked, compliance is determined, if the vehicle is compliant, the curb machine is re-enabled for verification and data fusion to generate a standardized event record, and if the vehicle is abnormal, whole-process space-time information is recorded. Through the synchronous monitoring of high-low sensors, multi-source data cross-validation is realized at the key nodes of vehicle state recognition, and sensor resources are dynamically dispatched based on the verification result, so that the accuracy and reliability of parking management are improved, and the resource consumption of the system is reduced.

The present invention relates to a method and system for acoustic machine perception for an aircraft. In an example, a method is described. The method comprises: causing one or more sensors arranged on the aircraft to acquire, within a time window, first data associated with a first object within a surrounding environment of the aircraft, wherein the one or more sensors comprise one or more of a light detection and ranging (LIDAR) sensor, a radar sensor, and a camera; causing a microphone array arranged on the aircraft to acquire, within approximately the same time window as the first data is acquired, first acoustic data associated with the first object; and training a machine learning model by using the first acoustic data as input values for the machine learning model and using an azimuth, a range, a height, and a type of the first object identified from the first data as ground truth output labels for the machine learning model.

In various examples, image signal processing (ISP) using infrared (IR) image-based control for visible wavelength color data channel processing is presented. Image color data channel processing stages of an ISP pipeline are configured to exploit IR information represented in an IR-based control channel so that IR information may be used to enhance the image processing for human visualization and machine perception functions. In some embodiments, a pseudo-radiometric (PR) channel may be formed based on the color data and IR data channels and propagated through the ISP pipeline. An IR-based control channel may carry control data computed based on luma data, IR channel data, PR channel data, image statistics, or combinations thereof. Image modulators may be derived from the control channel, and the image modulator used by one or more of the image color data channel processing stages of the ISP pipeline to adjust the visible wavelength color data channels.

The invention discloses a generative asynchronous voice anonymization method and a related device, and the method comprises the steps: firstly obtaining a voice signal of a speaker, then carrying out the decoupling of the voice signal, obtaining a speaker representation used for recognizing the identity of the speaker, decomposing the speaker representation into D-dimensional feature vectors, and carrying out the recognition of the identity of the speaker through the D-dimensional feature vectors; as the anonymized voice comprises contents in two aspects of machine perception anonymization and human subjective perception anonymization, the feature vector used for identifying machine perception is edited, the identity of a machine recognition speaker can be confused, and finally, the voice signal is reconstructed based on the edited speaker representation, so that the recognition accuracy of the speaker is improved. The generated anonymized voice realizes asynchronous voice anonymization, that is, while human subjective perception is reserved, the identification accuracy of the identity of a machine dialogue person is reduced, so that a good anonymization effect is achieved from the perspective of machine perception, and scene requirements of sound uniqueness, expressive force and specialty of the anonymized voice are met at the same time.