1. A system (100) for detecting moving objects in dynamic scenes using an adaptive machine learning framework, the system (100) comprising: an input unit (102) adapted to receive input data comprising video frames captured from a sensing source; a pre-processing unit (104) adapted to generate stabilized and noise-reduced frames using de-warping, temporal filtering, and motion stabilization techniques; a decision unit (118) adapted to generate output data comprising detection results, alerts, or actions based on analyzed contextual information; and a processing unit (120) operatively coupled with the input unit (102), and the pre-processing unit (104), characterized in that the processing unit (120) is configured to: receive, from the input unit (102), the input data comprising video frames; enable the pre-processing unit (104) to generate stabilized and noise-reduced frames; activate a background modeling unit (106) and a feature extraction unit (108) to isolate candidate moving regions and extract temporal-spatial features; initiate an event filtering unit (110) and an inference unit (112) to filter the extracted features and detect and track moving objects; update, via an adaptive learning unit (114), the machine learning model based on detected scene variations; and utilize a contextual analysis unit (116) and the decision unit (118) to analyze contextual information and generate output data comprising detection results, alerts, or actions.

2. The system (100) as claimed in claim 1, wherein the sensing source associated with the input unit (102) comprises a surveillance camera, a drone-mounted camera, a robotic vision system, or a vehicular imaging device.

3. The system (100) as claimed in claim 1, wherein the pre-processing unit (104) is adapted to perform illumination normalization and shadow suppression to enhance frame quality.

4. The system (100) as claimed in claim 1, wherein the background modeling unit (106) is adapted to perform background modeling and motion segmentation to isolate candidate moving regions.

5. The system (100) as claimed in claim 1, wherein the feature extraction unit (108) is adapted to extract temporal-spatial features from the candidate moving regions based on motion patterns and frame variations.

6. The system (100) as claimed in claim 1, wherein the event filtering unit (110) is adapted to remove redundant and irrelevant information from extracted features.

7. The system (100) as claimed in claim 1, wherein the inference unit (112) comprises the machine learning model adapted for object detection and tracking.

8. The system (100) as claimed in claim 1, wherein the adaptive learning unit (114) is adapted to update the machine learning model based on incoming data and detected scene variations.

9. The system (100) as claimed in claim 1, wherein the contextual analysis unit (116) is adapted to analyze contextual information associated with detected moving objects to determine scene dynamics and object interactions.

10. A method (300) for detecting moving objects in dynamic scenes using an adaptive machine learning framework, the method (300) is characterized by steps of: receiving input data comprising video frames from an input unit (102); enabling a pre-processing unit (104) to generate stabilized and noise-reduced frames; activating a background modeling unit (106) and a feature extraction unit (108) to isolate candidate moving regions and extract temporal-spatial features; initiating an event filtering unit (110) and an inference unit (112) to filter the extracted features and detect and track moving objects; updating, via an adaptive learning unit (114), a machine learning model based on detected scene variations; and utilizing a contextual analysis unit (116) and a decision unit (118) to analyze contextual information and generate output data comprising detection results, alerts, or actions. Date: May 13, 2026 Place: Noida Nainsi Rastogi Patent Agent (IN/PA-2372) Agent for the Applicant

Description:BACKGROUND
Field of Invention
[001] Embodiments of the present invention generally relate to real-time object detection and particularly to a system and method for detecting moving objects in dynamic scenes.
Description of Related Art
[002] Dynamic scene environments present substantial challenges in accurate identification of moving objects. Variations in illumination, presence of shadows, weather disturbances, and background motion such as tree movement or water surface fluctuations reduce reliability of conventional detection approaches. Non-stationary camera setups mounted on drones, robots, or vehicles introduce additional distortions and motion artifacts, that complicate separation between foreground objects and background regions. High computational demand restricts practical use in systems that require immediate response and consistent accuracy.
[003] Existing approaches rely on combinations of edge-based processing systems and centralized video management platforms, where local devices perform preliminary detection and transmit metadata for further analysis. Hybrid architectures exist. Further, edge devices execute real-time detection tasks while cloud infrastructure supports storage, analytics, and model updates. Deep learning techniques with automated feature extraction provide improved accuracy compared to traditional rule-based methods. Such solutions find use in surveillance systems, smart city deployments, and autonomous navigation frameworks due to their capability to process large-scale visual data.
[004] However, present solutions exhibit several limitations. Many systems depend on static models that require manual updates when environmental conditions change, that reduces adaptability. High false alarm rates persist in scenarios with complex or dynamic backgrounds. Computational requirements remain significant, that restricts deployment on resource-constrained devices. Lack of contextual understanding further limits robustness, as most methods rely only on motion or shape cues without deeper scene interpretation. These shortcomings reduce efficiency, scalability, and reliability in real-world dynamic environments.
[005] There is thus a need for an improved and advanced system and method for detecting moving objects in dynamic scenes that can administer the aforementioned limitations in a more efficient manner.
SUMMARY
[006] Embodiments in accordance with the present invention provide a system for detecting moving objects in dynamic scenes. The system comprising an input unit adapted to receive input data comprising video frames captured from a sensing source. The system further comprising a pre-processing unit adapted to generate stabilized and noise-reduced frames using de-warping, temporal filtering, and motion stabilization techniques. The system further comprising a decision unit adapted to generate output data comprising detection results, alerts, or actions based on analyzed contextual information. The system further comprising a processing unit operatively coupled with the input unit and the pre-processing unit. The processing unit is configured to receive, from the input unit, the input data comprising video frames, enable the pre-processing unit to generate stabilized and noise-reduced frames, activate the background modeling unit and the feature extraction unit to isolate candidate moving regions and extract temporal-spatial features, initiate the event filtering unit and the inference unit to filter the extracted features and detect and track moving objects, update, via the adaptive learning unit, the machine learning model based on detected scene variations, and utilize the contextual analysis unit and the decision unit to analyze contextual information and generate output data comprising detection results, alerts, or actions.
[007] Embodiments in accordance with the present invention further provide a method for detecting moving objects in dynamic scenes. The method comprising steps of: receiving, input data comprising video frames from an input unit; enabling a pre-processing unit to generate stabilized and noise-reduced frames; activating a background modeling unit and a feature extraction unit to isolate candidate moving regions and extract temporal-spatial features; initiating an event filtering unit and an inference unit to filter the extracted features and detect and track moving objects; updating, via an adaptive learning unit, a machine learning model based on detected scene variations; and utilizing a contextual analysis unit and a decision unit to analyze contextual information and generate output data comprising detection results, alerts, or actions.
[008] Embodiments of the present invention may provide a number of advantages depending on their particular configuration. First, embodiments of the present application may provide a system for detecting moving objects in dynamic scenes.
[009] Next, embodiments of the present application may provide a system for detecting moving objects in dynamic scenes that provides enhanced adaptability to dynamic environmental conditions including illumination variation, weather disturbance, and background motion without requirement of frequent manual intervention.
[0010] Next, embodiments of the present application may provide a system for detecting moving objects in dynamic scenes that provides reduction in false detection events through improved discrimination between actual moving objects and dynamic background elements.
[0011] Next, embodiments of the present application may provide a system for detecting moving objects in dynamic scenes that provides efficient real-time performance with optimized computational resource utilization suitable for deployment on edge-based platforms.
[0012] Next, embodiments of the present application may provide a system for detecting moving objects in dynamic scenes that provides improved detection reliability through incorporation of contextual scene understanding and motion pattern evaluation.
[0013] Next, embodiments of the present application may provide a system for detecting moving objects in dynamic scenes that provides scalable operation across multiple application domains including surveillance, autonomous systems, and urban monitoring environments.
[0014] These and other advantages will be apparent from the present application of the embodiments described herein.
[0015] The preceding is a simplified summary to provide an understanding of some embodiments of the present invention. This summary is neither an extensive nor exhaustive overview of the present invention and its various embodiments. The summary presents selected concepts of the embodiments of the present invention in a simplified form as an introduction to the more detailed description presented below. As will be appreciated, other embodiments of the present invention are possible utilizing, alone or in combination, one or more of the features set forth above or described in detail below.
BRIEF DESCRIPTION OF THE DRAWINGS
[0016] The above and still further features and advantages of embodiments of the present invention will become apparent upon consideration of the following detailed description of embodiments thereof, especially when taken in conjunction with the accompanying drawings, and wherein:
[0017] FIG. 1 illustrates a block diagram of system for detecting moving objects in dynamic scenes, according to an embodiment of the present invention;
[0018] FIG. 2 illustrates components of a processing unit of the system for detecting moving objects in dynamic scenes, according to an embodiment of the present invention; and
[0019] FIG. 3 depicts a flowchart of a method for detecting moving objects in dynamic scenes, according to an embodiment of the present invention.
[0020] The headings used herein are for organizational purposes only and are not meant to be used to limit the scope of the description or the claims. As used throughout this application, the word "may" is used in a permissive sense (i.e., meaning having the potential to), rather than the mandatory sense (i.e., meaning must). Similarly, the words “include”, “including”, and “includes” mean including but not limited to. To facilitate understanding, like reference numerals have been used, where possible, to designate like elements common to the figures. Optional portions of the figures may be illustrated using dashed or dotted lines, unless the context of usage indicates otherwise.
DETAILED DESCRIPTION
[0021] The following description includes the preferred best mode of one embodiment of the present invention. It will be clear from this description of the invention that the invention is not limited to these illustrated embodiments but that the invention also includes a variety of modifications and embodiments thereto. Therefore, the present description should be seen as illustrative and not limiting. While the invention is susceptible to various modifications and alternative constructions, it should be understood, that there is no intention to limit the invention to the specific form disclosed, but, on the contrary, the invention is to cover all modifications, alternative constructions, and equivalents falling within the scope of the invention as defined in the claims.
[0022] In any embodiment described herein, the open-ended terms "comprising", "comprises”, and the like (which are synonymous with "including", "having” and "characterized by") may be replaced by the respective partially closed phrases "consisting essentially of", “consists essentially of", and the like or the respective closed phrases "consisting of", "consists of”, the like.
[0023] As used herein, the singular forms “a”, “an”, and “the” designate both the singular and the plural, unless expressly stated to designate the singular only.
[0024] FIG. 1 illustrates a block diagram of a system 100 for detecting moving objects in dynamic scenes using an adaptive machine learning framework, according to an embodiment of the present invention. In an embodiment of the present invention, the system 100 may be adapted to receive input data comprising sequential video frames from a sensing environment and process the input data to generate stabilized and noise-reduced frames. The system 100 may be further adapted to compensate for motion associated with non-stationary platforms including drones, robotic systems, and vehicular systems to distinguish between background motion and object motion.
[0025] The system 100 may be adapted to perform background modeling and motion segmentation to isolate candidate moving regions and extract temporal-spatial features representing motion patterns. The system 100 may be further adapted to filter redundant and low-significance data to reduce computational load prior to performing detection and tracking of moving objects using the machine learning model. The machine learning model may be adapted to update dynamically based on incoming data and detected scene variations without requirement of complete retraining.
[0026] The system 100 may be adapted to analyze contextual information including object interaction, motion behavior, and environmental conditions to generate output data comprising detection results, alerts, or actions. The system 100 may be further adapted to operate in an edge-based configuration for real-time processing and may transmit metadata to a remote system adapted for model refinement and parameter optimization. Updated parameters may be integrated into the system 100 to enable continuous adaptive performance across dynamic environments.
[0027] The system 100 may be robust, scalable, and adaptive for real-time object detection that intelligently processes incoming visual data, dynamically updates detection models, and generates reliable outputs suitable for monitoring, alert generation, and automated response with minimal latency.
[0028] According to the embodiments of the present invention, the system 100 may incorporate non-limiting hardware components to enhance the processing speed and efficiency such as the system 100 may comprise an input unit 102, a pre-processing unit 104, a background modeling unit 106, a feature extraction unit 108, an event filtering unit 110, an inference unit 112, an adaptive learning unit 114, a contextual analysis unit 116, a decision unit 118, and a processing unit 120. In an embodiment of the present invention, the hardware components of the system 100 may be integrated with computer-executable instructions for overcoming the challenges and the limitations of the existing systems.
[0029] In an embodiment of the present invention, the input unit 102 may be adapted to receive input data comprising video frames captured from a sensing source. The sensing sources may include, but are not limited to, surveillance cameras, drone-mounted imaging devices, robotic vision systems, vehicular imaging systems, visual data acquisition devices, and so forth. The input data may comprise sequential image frames representative of dynamic environments including indoor, outdoor, urban, or industrial settings. Embodiments of the present application are intended to include or otherwise cover any type of sensing source and input data acquisition mechanism implemented using known, related art, and/or later developed technologies.
[0030] In an embodiment of the present invention, the input unit 102 may be, but not limited to, surveillance camera, a drone-mounted camera, a robotic vision system, or a vehicular imaging device, a sensor data acquisition interface, a video streaming interface, and so forth. Embodiments of the present invention are intended to include or otherwise cover any type of the input unit 102, including known, related art, and/or later developed technologies.
[0031] In an embodiment of the present invention, the input unit 102 may transmit the received input data to the pre-processing unit 104. The pre-processing unit 104 may be configured to process the input data to generate stabilized and noise-reduced frames suitable for further analysis. The pre-processing may include, but not limited to, de-warping, temporal filtering, motion stabilization, illumination normalization, shadow suppression, frame enhancement operations, and so forth. Embodiments of the present application are intended to include or otherwise cover any type of preprocessing techniques implemented using known, related art, and/or later developed technologies.
[0032] In an embodiment of the present invention, the pre-processing unit 104 may be adapted to compensate for non-stationary camera motion associated with mobile platforms including drones, robotic systems, and vehicular imaging devices. The pre-processing unit 104 may be adapted to perform motion compensation, frame alignment, and geometric correction to stabilize input video frames.
[0033] The pre-processing unit 104 may be further adapted to distinguish between camera-induced motion and object motion by analyzing global motion patterns across consecutive frames. The processing unit 120 may be adapted to utilize motion estimation techniques to isolate background displacement caused by camera movement. This configuration may enable accurate separation of foreground objects from dynamic backgrounds in scenarios involving moving cameras.
[0034] In an embodiment of the present invention, the pre-processing unit 104 may be, but not limited to, a frame stabilization engine, a noise reduction module, a temporal filtering engine, a geometric correction engine, an illumination normalization unit, a shadow suppression unit, and so forth. Embodiments of the present invention are intended to include or otherwise cover any type of the pre-processing unit 104, including known, related art, and/or later developed technologies.
[0035] In an embodiment of the present invention, processed frames may be transmitted to the background modeling unit 106 operatively coupled with the pre-processing unit 104. The background modeling unit 106 may be adapted to perform background modeling and motion segmentation to isolate candidate moving regions from dynamic background elements. The background modeling may include adaptive background subtraction techniques capable of accommodating environmental variations such as moving vegetation, water surfaces, and illumination changes, and other dynamic scene components. Embodiments of the present application are intended to include or otherwise cover any type of background modeling techniques implemented using known, related art, and/or later developed technologies.
[0036] In an embodiment of the present invention, the background modeling unit 106 may be, but not limited to, an adaptive background subtraction engine, a motion segmentation unit, a probabilistic background model, a statistical modeling engine, a dynamic scene modeling unit, and so forth. Embodiments of the present invention are intended to include or otherwise cover any type of the background modeling unit 106, including known, related art, and/or later developed technologies.
[0037] In an embodiment of the present invention, candidate moving regions may be provided to the feature extraction unit 108. The feature extraction unit 108 may be adapted to extract temporal-spatial features associated with moving regions based on motion and frame variations. The feature extraction unit 108 may be adapted to perform optical flow computation and motion vector analysis for extracting temporal-spatial features. The feature extraction unit 108 may be adapted to compute optical flow, motion vectors, and other relevant descriptors that characterize object movement across consecutive frames. Embodiments of the present application are intended to include or otherwise cover any type of feature extraction techniques implemented using known, related art, and/or later developed technologies.
[0038] In an embodiment of the present invention, the feature extraction unit 108 may be, but not limited to, an optical flow computation engine, a motion vector extraction unit, a feature descriptor generator, a temporal-spatial analysis engine, a pattern recognition unit, and so forth. Embodiments of the present invention are intended to include or otherwise cover any type of the feature extraction unit 108, including known, related art, and/or later developed technologies.
[0039] In an embodiment of the present invention, the extracted features may be transmitted to the event filtering unit 110. The event filtering unit 110 may be adapted to remove redundant and irrelevant information. The event filtering unit 110 may eliminate static regions, noise-induced motion patterns, and low-significance features to reduce computational load overhead and enhance processing efficiency. Embodiments of the present application are intended to include or otherwise cover any type of filtering techniques implemented using known, related art, and/or later developed technologies.
[0040] In an embodiment of the present invention, the event filtering unit 110 may be adapted to implement a lightweight event filtering mechanism configured for computational optimization. The event filtering unit 110 may be adapted to evaluate motion significance, spatial continuity, and temporal persistence of extracted features to identify relevant motion events. The event filtering unit 110 may be further adapted to discard low-intensity motion patterns, noise-induced variations, and redundant frame regions that do not contribute to meaningful object detection.
[0041] The event filtering unit 110 may be adapted to reduce data dimensionality prior to processing by the inference unit 112, thereby minimizing computational load on the system 100. The processing unit 120 may be adapted to dynamically adjust filtering thresholds based on system load conditions and environmental complexity. This configuration may enable efficient utilization of computational resources while maintaining detection accuracy in dynamic scenes.
[0042] In an embodiment of the present invention, the event filtering unit 110 may be, but not limited to, a motion filtering engine, a noise suppression unit, a redundancy elimination unit, a threshold-based filtering unit, a significance evaluation engine, and so forth. Embodiments of the present invention are intended to include or otherwise cover any type of the event filtering unit 110, including known, related art, and/or later developed technologies.
[0043] In an embodiment of the present invention, the filtered feature data may be provided to the inference unit 112 operatively coupled with the event filtering unit 110. The inference unit 112 may be adapted to detect and track moving objects using a machine learning model. The machine learning model may include deep learning architectures adapted for object detection and tracking in dynamic environments. Embodiments of the present application are intended to include or otherwise cover any type of the machine learning models implemented using known, related art, and/or later developed technologies.
[0044] In an embodiment of the present invention, the inference unit 112 may be, but not limited to, a deep learning inference engine, a neural network-based detection unit, a convolutional neural network processor, a transformer-based detection engine, a tracking algorithm unit, and so forth. Embodiments of the present invention are intended to include or otherwise cover any type of the inference unit 112, including known, related art, and/or later developed technologies.
[0045] In an embodiment of the present invention, the inference unit 112 may be further operatively coupled with the adaptive learning unit 114. The adaptive learning unit 114 may be configured to update parameters of the machine learning model based on incoming data and detected scene variations. The adaptive learning may include online learning techniques, incremental learning techniques, domain adaptation techniques, and so forth, to maintain detection accuracy without requirement of complete retraining. Embodiments of the present application are intended to include or otherwise cover any type of adaptive learning techniques implemented using known, related art, and/or later developed technologies.
[0046] In an embodiment of the present invention, the adaptive learning unit 114 may be, but not limited to, an online learning engine, an incremental learning module, a domain adaptation unit, a parameter update engine, a reinforcement learning component, and so forth. Embodiments of the present invention are intended to include or otherwise cover any type of the adaptive learning unit 114, including known, related art, and/or later developed technologies.
[0047] In an embodiment of the present invention, the detected objects and associated data may be transmitted to the contextual analysis unit 116. The contextual analysis unit 116 may be adapted to evaluate motion behavior, object interaction, and environmental conditions for improving detection accuracy. The contextual analysis unit 116 may be adapted to analyze contextual information associated with detected moving objects to determine scene dynamics and object interactions. The contextual analysis unit 116 may evaluate motion behavior, object interaction, and environmental conditions to determine scene dynamics and improve detection reliability. Embodiments of the present application are intended to include or otherwise cover any type of contextual analysis techniques implemented using known, related art, and/or later developed technologies.
[0048] In an embodiment of the present invention, the contextual analysis unit 116 may be adapted to perform context-sensitive scene interpretation based on interaction patterns among detected objects and environmental elements. The contextual analysis unit 116 may be adapted to analyze spatial relationships, motion trajectories, and temporal correlations to identify interaction events such as grouping, collision, or coordinated movement.
[0049] The contextual analysis unit 116 may be further adapted to incorporate environmental context including scene layout, background dynamics, and motion density. The processing unit 120 may be adapted to utilize contextual insights to refine detection outputs and reduce false detection events. This approach may enhance reliability of the system 100 in complex dynamic scenes.
[0050] In an embodiment of the present invention, the contextual analysis unit 116 may be, but not limited to, a scene understanding engine, an interaction analysis unit, a behavioral modeling engine, a spatial-temporal reasoning unit, an environmental context evaluation unit, and so forth. Embodiments of the present invention are intended to include or otherwise cover any type of the contextual analysis unit 116, including known, related art, and/or later developed technologies.
[0051] In an embodiment of the present invention, contextual information may be provided to the decision unit 118 configured to generate output data comprising detection results, alerts, or actions based on analyzed contextual information. The decision unit 118 may determine and generate appropriate responses based on analyzed information and may facilitate communication with external systems for monitoring or control purposes. Embodiments of the present application are intended to include or otherwise cover any type of decision-making and output generation mechanisms implemented using known, related art, and/or later developed technologies.
[0052] In an embodiment of the present invention, the system 100 may be adapted to operate in a hybrid edge–cloud architecture to enhance adaptive learning and scalability. The decision unit 118 may be further adapted to generate metadata and telemetry data derived from detected moving objects and contextual analysis. The metadata may include object identifiers, motion vectors, temporal markers, environmental descriptors, and confidence scores associated with detection events. The generated metadata may be transmitted through a communication interface to a remote cloud system adapted for large-scale analytics and model management.
[0053] The cloud system may be adapted to aggregate metadata received from multiple instances of the system 100 deployed across distributed environments. The cloud system may be further adapted to perform model retraining, parameter optimization, and pattern generalization using aggregated data. Updated model parameters may be generated based on evolving environmental conditions including illumination variations, seasonal changes, and motion dynamics. The updated model parameters may be transmitted back to the system 100 and may be applied through the adaptive learning unit 114 and the processing unit 120.
[0054] The processing unit 120 may be adapted to integrate updated model parameters without requiring complete retraining at the system level. This architecture may enable continuous improvement of detection accuracy while maintaining real-time performance at the edge. The system 100 may be adapted to maintain synchronization between edge inference operations and cloud-driven model updates to ensure consistent detection behavior across deployments.
[0055] In an embodiment of the present invention, the decision unit 118 may be, but not limited to, an alert generation engine, an action decision module, a rule-based decision system, a response triggering unit, an output generation engine, and so forth. Embodiments of the present invention are intended to include or otherwise cover any type of the decision unit 118, including known, related art, and/or later developed technologies.
[0056] In an embodiment of the present invention, the processing unit 120 may be operatively coupled with the input unit 102, the pre-processing unit 104, the background modeling unit 106, the feature extraction unit 108, the event filtering unit 110, the inference unit 112, the adaptive learning unit 114, the contextual analysis unit 116, and the decision unit 118. The processing unit 120 may receive the input data via the input unit 102, enable the preprocessing unit 104 to generate stabilized and noise-reduced frames, activate the background modeling unit 106 and the feature extraction unit 108, initiate the event filtering unit 110 and the inference unit 112 to filter the extracted features and detect and track moving objects, update the machine learning model through adaptive learning unit 114, and utilize the contextual analysis unit 116 and the decision unit 118 to analyze contextual information and generate output data comprising detection results, alerts, or actions. The processing unit 120 may ensure synchronized data flow and efficient execution across all functional units to enable reliable and adaptive detection of moving objects in dynamic scenes.
[0057] In an embodiment of the present invention, the processing unit 120 may be adapted to implement a multi-stage processing pipeline comprising sequential execution of the pre-processing unit 104, the background modeling unit 106, the feature extraction unit 108, the event filtering unit 110, the inference unit 112, the adaptive learning unit 114, the contextual analysis unit 116, and the decision unit 118.
[0058] The processing unit 120 may be adapted to enforce sequential dependency between stages such that output generated by a preceding unit may be utilized as input for a subsequent unit. The processing unit 120 may be further adapted to manage data propagation across stages to maintain temporal consistency and processing integrity. This structured pipeline may enhance robustness in dynamic environments by ensuring coordinated execution of all functional units.
[0059] In an embodiment of the present invention, the processing unit 120 may be adapted to operate under real-time constraints with optimized utilization of computational resources. The processing unit 120 may be adapted to dynamically adjust processing parameters based on system load, input data complexity, and environmental conditions. The processing unit 120 may be adapted to prioritize latency-sensitive operations including object detection and tracking while maintaining accuracy. Resource allocation strategies may be applied through the processing unit 120 to ensure efficient execution across all units. This configuration may enable deployment on resource-constrained edge devices without degradation in performance.
[0060] In an embodiment of the present invention, the processing unit 120 may be, but not limited to, a central processing unit, a graphics processing unit, an edge processing engine, a distributed computing unit, a hardware accelerator, a programmable logic controller, and so forth. Embodiments of the present invention are intended to include or otherwise cover any type of the processing unit 120, including known, related art, and/or later developed technologies. The processing unit 120 may be explained further in detail in conjunction with FIG. 2.
[0061] FIG. 2 illustrates components of the processing unit 120 of the system 100, according to an embodiment of the present invention. The processing unit 120 may comprise a data flow management module 200, a task scheduling module 202, a model management module 204, and a decision coordination module 206. The modules may collectively enable controlled data handling, efficient task execution, adaptive model management, and coordinated decision generation for accurate and real-time object detection.
[0062] In an embodiment of the present invention, the data flow management module 200 may be configured to manage transmission of data across the system 100. The data flow management module 200 may receive input data from the input unit 102 and may be further configured to route the input data to the pre-processing unit 104. Further, the data flow management module 200 may be configured to transmit processed data from the pre-processing unit 104 to the background modeling unit 106 and the feature extraction unit 108. The data flow management module 200 may be further configured to regulate data transfer between the event filtering unit 110, the inference unit 112, the adaptive learning unit 114, the contextual analysis unit 116, and the decision unit 118. The data flow management module 200 may be configured to maintain sequencing of data and reduce latency during transmission.
[0063] In an embodiment of the present invention, the task scheduling module 202 may be configured to control execution of tasks associated with the system 100. The task scheduling module 202 may be configured to assign execution priority to operations performed by the pre-processing unit 104, the background modeling unit 106, the feature extraction unit 108, the event filtering unit 110, and the inference unit 112. Further, the task scheduling module 202 may be configured to allocate computational resources based on processing requirements and system 100 load conditions. In an exemplary scenario, if the task scheduling module 202 determines that the inference unit 112 requires higher computational priority for real-time detection, the task scheduling module 202 may allocate additional resources to the inference unit 112. In another scenario, the task scheduling module 202 may enable parallel execution of feature extraction and background modeling operations to improve processing efficiency.
[0064] In an embodiment of the present invention, the model management module 204 may be configured to manage the machine learning models associated with the inference unit 112 and the adaptive learning unit 114. The model management module 204 may be configured to store model parameters in a memory associated with the processing unit 120. Further, the model management module 204 may be configured to retrieve and apply appropriate model parameters for detection and tracking operations. The model management module 204 may be configured to receive updated parameters from the adaptive learning unit 114 and incorporate the updated parameters into the machine learning model. In an exemplary scenario, if the model management module 204 determines variation in scene conditions, the model management module 204 may be configured to select an updated model corresponding to the detected variation.
[0065] In an embodiment of the present invention, the decision coordination module 206 may be configured to coordinate generation of output data based on processed information. The decision coordination module 206 may be configured to receive contextual information from the contextual analysis unit 116. Further, the decision coordination module 206 may be configured to control operation of the decision unit 118 for generation of detection results, alerts, or actions. In an embodiment of the present application, the decision coordination module 206 may be configured to apply predefined rules or logic to ensure consistency in output generation. In an exemplary scenario, if the decision coordination module 206 determines that a detected object satisfies a predefined condition, the decision coordination module 206 may be configured to enable the decision unit 118 to generate an alert. In another scenario, the decision coordination module 206 may be configured to transmit output data to external systems for monitoring or control purposes.
[0066] FIG. 3 depicts a flowchart of a method 300 for detecting the moving objects in the dynamic scenes, according to an embodiment of the present invention.
[0067] At step 302, the system 100 may receive the input data comprising video frames from the input unit 102.
[0068] At step 304, the system 100 may enable the pre-processing unit 104 to generate stabilized and noise-reduced frames.
[0069] At step 306, the system 100 may activate the background modeling unit 106 and the feature extraction unit 108 to isolate candidate moving regions and extract temporal-spatial features.
[0070] At step 308, the system 100 may initiate the event filtering unit 110 and the inference unit 112 to filter the extracted features and detect and track the moving objects.
[0071] At step 310, the system 100 may update, via the adaptive learning unit 114, the machine learning model based on detected scene variations.
[0072] At step 312, the system 100 may utilize the contextual analysis unit 116 and the decision unit 118 to analyze contextual information and generate output data comprising detection results, alerts, or actions.
[0073] While the invention has been described in connection with what is presently considered to be the most practical and various embodiments, it is to be understood that the invention is not to be limited to the disclosed embodiments, but on the contrary, is intended to cover various modifications and equivalent arrangements included within the scope of the appended claims.
[0074] This written description uses examples to disclose the invention, including the best mode, and also to enable any person skilled in the art to practice the invention, including making and using any devices or systems and performing any incorporated methods. The patentable scope of the invention is defined in the claims, and may include other examples that occur to those skilled in the art. Such other examples are intended to be within the scope of the claims if they have structural elements that do not differ from the literal language of the claims, or if they include equivalent structural elements within substantial differences from the literal languages of the claims. , Claims:CLAIMS
I/We Claim:
1. A system (100) for detecting moving objects in dynamic scenes using an adaptive machine learning framework, the system (100) comprising:
an input unit (102) adapted to receive input data comprising video frames captured from a sensing source;
a pre-processing unit (104) adapted to generate stabilized and noise-reduced frames using de-warping, temporal filtering, and motion stabilization techniques;
a decision unit (118) adapted to generate output data comprising detection results, alerts, or actions based on analyzed contextual information; and
a processing unit (120) operatively coupled with the input unit (102), and the pre-processing unit (104), characterized in that the processing unit (120) is configured to:
receive, from the input unit (102), the input data comprising video frames;
enable the pre-processing unit (104) to generate stabilized and noise-reduced frames;
activate a background modeling unit (106) and a feature extraction unit (108) to isolate candidate moving regions and extract temporal-spatial features;
initiate an event filtering unit (110) and an inference unit (112) to filter the extracted features and detect and track moving objects;
update, via an adaptive learning unit (114), the machine learning model based on detected scene variations; and
utilize a contextual analysis unit (116) and the decision unit (118) to analyze contextual information and generate output data comprising detection results, alerts, or actions.
2. The system (100) as claimed in claim 1, wherein the sensing source associated with the input unit (102) comprises a surveillance camera, a drone-mounted camera, a robotic vision system, or a vehicular imaging device.
3. The system (100) as claimed in claim 1, wherein the pre-processing unit (104) is adapted to perform illumination normalization and shadow suppression to enhance frame quality.
4. The system (100) as claimed in claim 1, wherein the background modeling unit (106) is adapted to perform background modeling and motion segmentation to isolate candidate moving regions.
5. The system (100) as claimed in claim 1, wherein the feature extraction unit (108) is adapted to extract temporal-spatial features from the candidate moving regions based on motion patterns and frame variations.
6. The system (100) as claimed in claim 1, wherein the event filtering unit (110) is adapted to remove redundant and irrelevant information from extracted features.
7. The system (100) as claimed in claim 1, wherein the inference unit (112) comprises the machine learning model adapted for object detection and tracking.
8. The system (100) as claimed in claim 1, wherein the adaptive learning unit (114) is adapted to update the machine learning model based on incoming data and detected scene variations.
9. The system (100) as claimed in claim 1, wherein the contextual analysis unit (116) is adapted to analyze contextual information associated with detected moving objects to determine scene dynamics and object interactions.
10. A method (300) for detecting moving objects in dynamic scenes using an adaptive machine learning framework, the method (300) is characterized by steps of:
receiving input data comprising video frames from an input unit (102);
enabling a pre-processing unit (104) to generate stabilized and noise-reduced frames;
activating a background modeling unit (106) and a feature extraction unit (108) to isolate candidate moving regions and extract temporal-spatial features;
initiating an event filtering unit (110) and an inference unit (112) to filter the extracted features and detect and track moving objects;
updating, via an adaptive learning unit (114), a machine learning model based on detected scene variations; and
utilizing a contextual analysis unit (116) and a decision unit (118) to analyze contextual information and generate output data comprising detection results, alerts, or actions.

Date: May 13, 2026
Place: Noida

Nainsi Rastogi
Patent Agent (IN/PA-2372)
Agent for the Applicant