Intelligent camera system with integrated artificial intelligence for real-time object classification and adaptive function control

Abstract

An intelligent camera system with integrated artificial intelligence for real-time object classification and adaptive function control, consisting of: an image sensor unit configured to capture successive image sequences of an observed environment; an optical assembly operationally coupled to the image sensor unit to project incident light onto a sensor surface; at least one processing unit that is arranged in a camera housing and electrically connected to the image sensor unit; a non-volatile memory that is operationally connected to the processing unit and configured to store executable instructions, trained object classification parameters, and control decision data; an image preprocessing unit, which is executed by the processing unit and is configured to normalize, filter, and align successive image frames in time; an artificial intelligence inference unit executed by the processing unit, configured to perform object classification in real time directly on the preprocessed image sequences using the stored trained parameters; a decision control unit executed by the processing unit, configured to generate adaptive control outputs based on classified object attributes such as object category, spatial position, motion properties, and confidence level; and A control interface unit electrically coupled to the decision control unit and configured to transmit adaptive control outputs to one or more functional components of a machine or structure, wherein all image acquisition, classification, decision making and control signal generation takes place locally within the camera housing, wherein the image sensor unit comprises a solid-state image sensor configured to operate at a variable frame rate, which is dynamically adjusted by the processing unit based on the detected scene complexity and object motion characteristics to maintain classification accuracy in real time while reducing the computational load.

Description

Technical field of the invention

[0001] The present invention relates generally to intelligent imaging and control systems and in particular to an intelligent camera system equipped with integrated hardware for processing artificial intelligence and configured to perform object classification in real time and autonomously control one or more functional processes of an associated machine or structure on the basis of classified visual information, without relying on external computing resources. Background of the invention

[0002] Conventional camera systems in industrial automation, surveillance, transportation, robotics, and infrastructure monitoring are primarily passive imaging devices that capture visual data and transmit it to remote computing systems for analysis and decision-making. These systems typically rely on central servers or cloud-based processing platforms to perform object detection, classification, and scene interpretation. As a result, they have significant limitations, including high communication latency, bandwidth dependency, reduced operational reliability in network-constrained environments, and delayed response times, making them unsuitable for real-time control applications.

[0003] Common image-based control solutions often use discrete image sensors coupled with separate processing units or programmable logic controllers (PLCs). The captured images are processed using predefined rule-based procedures or limited feature extraction methods. Such approaches are not adaptable to dynamic environments and cannot learn or improve their classification accuracy over time. Furthermore, traditional systems typically require manual reconfiguration or software updates to adjust system behavior. This makes them inflexible for applications that require autonomous responses to changing operating conditions.

[0004] Recent advances in artificial intelligence have enabled the use of deep learning models for object detection and classification. However, most existing implementations rely on external graphics processors or cloud infrastructure due to the high computational demands. This architecture introduces security risks related to data transmission, increased power consumption, and dependence on external networks. Furthermore, these systems are not optimized for direct integration with mechanical or electromechanical controls, limiting their ability to provide adaptive, closed-loop control based on visual intelligence.

[0005] Therefore, there is a need for a self-contained intelligent camera system capable of classifying objects in real time using integrated artificial intelligence hardware and software, while simultaneously interacting directly with and adaptively controlling functional components of a machine or structure – autonomously, with low latency and energy efficiency.

[0006] Conventional camera systems in industrial automation, transportation, security technology, retail analytics, robotics, and infrastructure monitoring traditionally function as passive image capture devices whose primary function is limited to acquiring visual data for subsequent analysis. In such systems, image sensors are typically coupled with simple signal processing electronics and transmit raw or only lightly processed image streams to external computing units for interpretation. This reliance on external processing architectures leads to inherent latency due to data transmission delays, rendering these systems unsuitable for applications requiring immediate decision-making or real-time control based on visual input.Furthermore, the need for continuous high-bandwidth data transmission places significant demands on the communication infrastructure, which becomes a critical limitation in distributed or resource-constrained environments.

[0007] Most current image processing-based systems utilize centralized server architectures or cloud computing platforms for object detection and classification. While these platforms offer high computing power, they depend on a stable network connection. In real-world applications, network instability, packet loss, or limited bandwidth can significantly impair system performance or even lead to total system failure. Furthermore, transmitting visual data to remote servers raises security and privacy concerns, particularly in sensitive environments such as manufacturing plants, critical infrastructure, healthcare facilities, and surveillance systems. These architectures also increase operating costs due to the continuous use of cloud services and the maintenance required for the infrastructure.

[0008] Conventional image processing solutions for industrial inspection and automation often employ rule-based image processing techniques such as edge detection, thresholding, and template matching. While these methods can be effective in controlled environments with consistent lighting and unchanged objects, they are highly sensitive to changes in environmental conditions. Modifications to lighting, object orientation, surface texture, or background often result in a significant degradation in performance. Consequently, such systems require extensive calibration and frequent manual reconfiguration to maintain an acceptable level of accuracy. This lack of adaptability limits their suitability for dynamic environments where viewing conditions change unpredictably.

[0009] Newer approaches integrate artificial intelligence and deep learning techniques to improve the accuracy and robustness of object classification. However, most existing implementations run these computationally intensive processes on external GPUs or dedicated inference servers. This separation between the image acquisition hardware and the AI ​​processing layer leads to additional latency and increased system complexity. The need to transfer image data from the camera to the processing server and then feed control decisions back to the operating devices prevents true real-time response, especially in safety-critical or high-speed applications. Furthermore, centralized inference architectures become bottlenecks as the number of cameras deployed increases, leading to scaling issues and reduced overall system responsiveness.

[0010] Edge-computing camera solutions emerged as an attempt to address some of these limitations by performing limited data processing directly at or near the camera. However, many existing edge cameras offer only limited intelligence, supporting basic analytics such as motion detection or predefined object counting. These systems typically lack sufficient computing resources to run complex object classification models in real time. Therefore, they either compromise on model complexity, resulting in lower classification accuracy, or offload more complex processing tasks to external systems. This hybrid approach cannot completely eliminate latency, network dependency, and architectural complexity.

[0011] Another significant drawback of existing intelligent camera systems is their limited integration with functional control mechanisms. In many applications, the output of visual analyses is restricted to warnings, notifications, or data logs, requiring human intervention or separate control systems to initiate operational responses. The lack of direct, autonomous control interfaces prevents these systems from adaptively controlling machines or structures based on visual intelligence. Consequently, the potential of image-based artificial intelligence to enable closed-loop control, particularly in industrial machinery, autonomous infrastructure, and smart environments, remains untapped.

[0012] Power consumption and thermal management pose additional challenges for existing solutions. Cameras designed for continuous video streaming and external processing often operate with high power consumption to ensure data transmission and sensor operation. When combined with external processing hardware, the overall energy demand of the system increases significantly, limiting its use in battery-powered, remote, or mobile applications. Attempts to integrate processing functions into cameras frequently result in thermal issues that restrict continuous operation or necessitate active cooling mechanisms. This increases system size, cost, and mechanical complexity.

[0013] Existing intelligent camera systems also exhibit limited adaptability over time. Many systems rely on static, offline-trained models that lack mechanisms for context-aware adaptation or incremental improvement. Updating classification models often requires system downtime, manual intervention, or a complete firmware replacement. This rigidity is particularly problematic in environments where object properties, operational contexts, or security requirements change. As a result, system performance gradually degrades, and maintaining accuracy necessitates repeated development efforts and operational interruptions.

[0014] From a hardware integration perspective, conventional camera systems are often designed as standalone components with standardized communication interfaces, with insufficient consideration given to their close integration with machine or structural elements. This design approach necessitates additional control units, programmable logic controllers (PLCs), or middleware layers to convert image analysis results into usable control signals. The resulting system architectures are complex, costly, and prone to synchronization errors, further impairing real-time capability. Furthermore, the physical separation of sensor, processing, and actuator components increases susceptibility to electromagnetic interference, mechanical vibrations, and environmental influences.

[0015] In safety-critical applications such as automated manufacturing, intelligent transportation systems, and access-controlled infrastructures, existing solutions reach their limits when it comes to meeting stringent reliability and deterministic requirements. Delays or misclassifications due to network interruptions, processing bottlenecks, or environmental influences can lead to unsafe conditions or operational disruptions. Current systems typically rely on redundant hardware or manual intervention mechanisms to minimize these risks. However, this increases system complexity and reduces overall efficiency.

[0016] Despite advances in image sensors and AI methods, existing camera-based systems remain limited by architectural separation, network dependency, limited integrated intelligence, insufficient control integration, and a lack of adaptability. These drawbacks prevent conventional solutions from enabling truly autonomous, real-time object classification and adaptive function control in a single, self-contained device. Therefore, there remains a need for an integrated intelligent camera system that combines powerful integrated AI, robust real-time processing, and direct control capabilities in a unified hardware architecture, capable of operating independently and reliably across diverse application environments. Summary of the invention

[0017] The present invention describes an intelligent camera system comprising an image sensor unit, integrated hardware for processing artificial intelligence, non-volatile memory, and a control interface. This system autonomously classifies objects within its field of view and dynamically controls one or more functional elements of an associated machine or structure. It performs image acquisition, preprocessing, feature extraction, and object classification in real time using locally stored and executed machine learning models. Based on the classification results and context parameters, the system generates control signals that adaptively adjust the operating states of connected mechanical, electrical, or electromechanical components.The invention further describes a device structure in which the intelligent camera system is physically integrated into or mounted on a machine or structure to enable direct, visual, intelligent control.

[0018] The main objective of the present invention is to provide an intelligent camera system with fully integrated AI processing functions that enables real-time object classification directly during image acquisition. This eliminates dependence on external computing resources and reduces latency in data transmission and remote processing. The invention aims to implement autonomous visual intelligence in a self-contained device that operates reliably in dynamic and resource-constrained environments.

[0019] A further objective of the invention is the adaptive functional control of a machine or structure based on classified visual information. Control decisions are generated locally by the intelligent camera system and applied directly to the functional components without human intervention. Through the close integration of visual perception and control output, the invention aims to enable closed-loop automation that reacts immediately to changes in the observed environment.

[0020] A further objective of the invention is to provide a hardware-based camera system that integrates image sensors, artificial intelligence, storage, and control interfaces in a robust housing and is suitable for industrial, outdoor, or mobile applications. The invention aims to ensure deterministic performance, improved reliability, and increased insensitivity to environmental influences such as vibrations, temperature fluctuations, and electromagnetic interference.

[0021] A further objective of the invention is to improve the accuracy and robustness of object classification under varying operating conditions, including changes in lighting, object orientation, background complexity, and motion dynamics. The invention aims to achieve this through the local execution of advanced artificial intelligence models while maintaining consistent real-time performance without compromising energy efficiency.

[0022] A further objective of the invention is to reduce system complexity and deployment costs by eliminating external servers, cloud infrastructure, or intermediate control units that are typically required for image-based decision-making. By consolidating sensor technology, processing, and control into a single device, the invention simplifies the system architecture and improves scalability in large installations.

[0023] A further objective of the invention is to support adaptive behavior and long-term operational relevance through the reliable updating of object classification parameters and the control logic within the intelligent camera system without interrupting ongoing operation. The invention thus aims to maintain accuracy and functionality under changing environmental conditions and application requirements.

[0024] A further objective of the invention is to provide an intelligent camera system with optimized power consumption and thermal management, making it suitable for continuous operation in energy-sensitive applications, such as battery-powered or remotely controlled systems. The invention aims to dynamically adapt the processing load to the operational requirements in order to ensure sustained performance and a long service life for the device.

[0025] A further objective of the invention is to improve operational reliability and autonomy in applications where an immediate response to visual events is crucial, for example, in automated machines, intelligent infrastructure, and controlled-access environments. The invention aims to minimize errors related to network latency, communication losses, or bottlenecks in central processing by ensuring that all critical functions are executed locally within the camera system. BRIEF DESCRIPTION OF THE IMAGE

[0026] These and other features, aspects and advantages of the present invention will be better understood if the following detailed description is read with reference to the accompanying drawing, in which the same symbols represent the same parts: Fig. Figure 1 shows a block diagram of an intelligent camera system with integrated artificial intelligence for real-time object classification and adaptive function control.

[0027] Furthermore, those skilled in the art will recognize that the elements in the drawing are simplified and not necessarily drawn to scale. For example, the flowcharts illustrate the process by highlighting the main steps to facilitate understanding of the present disclosure. With regard to the construction of the device, one or more components may be represented in the drawing by conventional symbols. The drawing may show only those specific details relevant to understanding the embodiments of the present disclosure, so as not to clutter the drawing with details that are already apparent to those skilled in the art from the description contained herein. Detailed description of the invention

[0028] To facilitate understanding of the principles of the invention, reference is made below to the embodiment shown in the drawing, which is described using specific terms. It is understood, however, that this does not limit the scope of protection of the invention. Rather, modifications and further developments of the depicted system, as well as further applications of the inventive principles shown therein, are conceivable, insofar as they would normally occur to a person skilled in the art in the field of the invention.

[0029] It will be clear to those skilled in the art that the foregoing general description and the following detailed description are exemplary and explanatory of the invention and are not to be understood as a limitation of it.

[0030] References to “an aspect”, “another aspect”, or similar phrases in this description mean that a particular feature, structure, or property described in connection with the embodiment is included in at least one embodiment of the present disclosure. Therefore, phrases such as “in one embodiment”, “in another embodiment”, and similar expressions in this description may, but do not necessarily, all refer to the same embodiment.

[0031] The terms "includes," "comprehensive," or similar expressions denote non-exclusive inclusion. Thus, a procedure or method containing a list of steps does not only include those steps but may also include further steps not explicitly listed or inherent in the procedure or method. Likewise, the statement "includes..." for one or more devices, subsystems, elements, structures, or components, without further limitations, does not preclude the existence of other devices, subsystems, elements, structures, or components.

[0032] Unless otherwise defined, all technical and scientific terms used herein have the same meanings generally known to those skilled in the art in the field to which this invention belongs. The systems, methods, and examples described herein serve only for illustration and are not to be understood as limiting.

[0033] Embodiments of the present disclosure are described in detail below with reference to the attached drawing.

[0034] Fig.Figure 1 shows the block diagram of an intelligent camera system with integrated artificial intelligence for real-time object classification and adaptive function control. The system 100 comprises: an image sensor unit (102) for capturing successive image sequences of the observed environment; an optical unit connected to the image sensor unit to project incident light onto a sensor surface; at least one processing unit (104) located in a camera housing and electrically connected to the image sensor unit; a non-volatile memory (106) connected to the processing unit that stores executable instructions, trained object classification parameters, and control data; and an image preprocessing unit (108) executed by the processing unit and configured to normalize, filter, and temporally align the successive image frames.an AI inference unit (110) that is executed by the processing unit and is configured to perform object classification in real time directly on the preprocessed image frames using the stored trained parameters; a decision control unit (112) that is executed by the processing unit and is configured to generate adaptive control outputs based on classified object attributes such as object category, spatial position, motion properties, and confidence level;and a control interface unit (114) that is electrically coupled to the decision control unit and is configured to transmit the adaptive control outputs to one or more functional components of a machine or structure, wherein all image acquisition, classification, decision making and control signal generation takes place locally within the camera housing without dependence on external computer systems.

[0035] In one embodiment, the image sensor unit (102) comprises a solid-state image sensor configured to operate at a variable frame rate, which is dynamically adjusted by the processing unit based on the detected scene complexity and object motion characteristics to maintain classification accuracy in real time while reducing the computational load.

[0036] In one embodiment, the image preprocessing unit (106) is further configured to perform illumination compensation, noise reduction and spatial normalization on successive image sequences in order to minimize performance losses due to fluctuations in ambient illumination and environmental conditions.

[0037] In one embodiment, the processing unit (104) comprises a heterogeneous processing architecture with a general-purpose processing element and a neural computing element optimized for the parallel execution of object classification operations stored in non-volatile memory.

[0038] In one embodiment, the artificial intelligence inference unit (110) is configured to perform multiple classification passes on temporally adjacent image frames and to aggregate the classification outputs to improve classification stability in the case of partial occlusion or rapid object movement.

[0039] In one embodiment, the decision control unit (112) is further configured to compare classified object attributes with predefined operating thresholds and context state data stored in non-volatile memory to determine whether an adaptive control response is required.

[0040] In one embodiment, the control interface unit (114) comprises at least one electrical output interface configured to directly control electromechanical components of the machine or structure without intermediate control hardware.

[0041] In one embodiment, the control interface unit (114) is further configured to generate variable control signals whose size, duration or timing is dynamically adjusted based on real-time changes in the position and trajectory of the classified object.

[0042] In one embodiment, the non-volatile memory (106) additionally stores historical classification data and previous control results, wherein the decision control unit is configured to access the historical classification data to refine subsequent control decisions.

[0043] In one embodiment, the processing unit (108) is further configured to execute energy management logic that selectively activates or deactivates parts of the artificial intelligence inference unit based on the detected object presence in the field of view.

[0044] The intelligent camera system is made possible by the detailed description of each functional component as a physically realizable hardware unit that works together in a coordinated manner within a camera housing. The image sensor unit is implemented as a semiconductor image sensor that generates successive digital images. The optical unit consists of standard refractive or diffractive optical elements that focus the incident light onto a sensor area with a controlled field of view and controlled resolution. The processing unit comprises one or more embedded processors or integrated computing circuits that can execute stored instructions and perform parallel arithmetic operations required for image preprocessing, artificial intelligence inference, and the generation of control decisions.The non-volatile memory is implemented using semiconductor memory chips that store executable instructions, trained classification parameters, intermediate results, and control outputs so they can be directly retrieved by the processing unit. The image preprocessing unit is implemented using executable instructions that instruct the processing unit to perform deterministic operations such as image normalization, noise filtering, resolution adjustment, and temporal alignment to ensure consistency between successive images. The AI ​​inference unit utilizes stored training parameters and inference instructions, enabling the processing unit to perform classification calculations directly on preprocessed image data in real time and without external computing resources.The decision control unit is implemented as control logic executed by the processing unit to evaluate classified object attributes such as category, spatial position, motion behavior, and confidence metrics, and to generate corresponding adaptive control signals. The control interface unit consists of electrical interface circuitry and communication logic configured to transmit generated control outputs to functional components of an associated machine or structure via wired or wireless signals. Dynamic frame rate adjustment of the image sensor unit is enabled by feedback signals generated by the processing unit based on the detected scene complexity and object motion characteristics. This allows the processing unit to adjust the sensor timing parameters to achieve an optimal balance between classification accuracy and computational load.

[0045] The intelligent camera system with integrated artificial intelligence for real-time object classification and adaptive function control operates with a tightly integrated sequence of hardware processing stages. Together, these implement real-time image recognition and control technology entirely within the camera housing. During operation, the image sensor unit continuously captures successive image sequences of the environment using the optical unit, which focuses the incoming light onto the sensor surface. The image sensor unit operates at a dynamically adjustable frame rate, and the processing unit continuously evaluates scene activity metrics from previous images to determine an optimal frame rate that ensures a balanced relationship between temporal resolution and computational efficiency.

[0046] Each captured image is transferred directly to the image preprocessing unit of the processing unit. During preprocessing, intensity values ​​are normalized at the pixel level to standardize the entire image. Spatial filtering is then performed to suppress sensor noise and environmental artifacts. Illumination compensation reduces sensitivity to uneven lighting, glare, or shadows, while geometric normalization ensures consistent spatial scaling and alignment between successive images. Temporal alignment logic assigns a corresponding time to each image, enabling precise tracking of object movements and their tracing across multiple images.

[0047] After preprocessing, the image data is forwarded to the AI ​​inference unit running on the processing unit. This unit accesses trained object classification parameters stored in non-volatile memory and uses these parameters to extract hierarchical visual features from the preprocessed image sequences. Feature extraction occurs in several steps, from low-level spatial features to higher-level semantic representations that encode object shape, texture, and context. The inference unit evaluates these representations to assign object categories, spatial boundary information, motion vectors, and confidence scores to one or more detected objects in each image sequence.

[0048] To improve robustness under conditions such as partial occlusion, motion blur, or rapid object movement, the inference unit classifies temporally adjacent image sequences and aggregates the results over a sliding time window. This aggregation process stabilizes the classification results by correlating consistent object features across different sequences and suppressing transient misclassifications. The aggregated classification results are temporarily stored in non-temporary memory for further decision processing.

[0049] The decision control unit, executed by the processing unit, receives the classification results and evaluates them based on predefined operating thresholds and contextual data stored in permanent memory. This evaluation includes determining whether detected objects lie within defined operating zones, assessing object trajectories relative to protected or functional areas, and analyzing classification accuracy to ensure reliability. The decision control unit also accesses historical classification data and previous control results to optimize current decision-making, thus enabling adaptive behavior that takes into account recurring patterns or changing environmental conditions.

[0050] As soon as it is determined that a control response is required, the decision control unit generates adaptive control outputs that define control parameters such as activation time, output quantity, and response time. These control outputs are transmitted to the control interface, which converts the decision parameters into electrical control signals suitable for directly controlling the functional components of the associated machine or structure. The control interface supports real-time modulation of control signals, enabling the system to continuously adapt machine behavior to changes in object position, movement, or classification state.

[0051] The control interface provides status feedback signals indicating the execution and result of the transmitted control actions. These feedback signals are received by the processing unit and stored in non-volatile memory, enabling validation of the control effect within the closed control loop. The decision control unit uses this feedback to adjust subsequent control outputs, thus forming a continuous, adaptive control loop driven by real-time visual intelligence.

[0052] To ensure continuous real-time operation, the processing unit performs image preprocessing and AI inference in a sequential sequence. While one image is being classified, a subsequent image is preprocessed in parallel. This guarantees uninterrupted throughput and minimal latency. The processing unit's power management logic dynamically activates or deactivates parts of the inference and preprocessing operations based on detected objects and scene complexity. This optimizes power consumption without compromising responsiveness.

[0053] The system also supports the secure updating of object classification parameters in permanent memory while the camera remains in operation. Updated parameters are validated and activated without interrupting ongoing image acquisition or control.

[0054] This allows the system to adapt to new object categories or changed operational requirements. If the classification reliability falls below predefined reliability thresholds, the decision control unit modifies or disables the control outputs to prevent unintended machine actions. If a critical object classification corresponding to a safety state is detected, the decision control unit generates fail-safe control outputs that put the machine or plant into a predefined safe operating state.

[0055] Through this integrated technical execution sequence, the intelligent camera system achieves fully automatic, real-time object classification and adaptive function control within a single, self-contained device. This process ensures low latency, high classification reliability, and direct control integration, enabling the system to function as an intelligent visual control node for machines and plants in a wide variety of operating environments.

[0056] The intelligent camera system with integrated AI for real-time object classification and adaptive function control consists of a housing for mounting on or integration into a machine or component. This housing contains an image sensor unit, one or more processing units, permanent memory, a voltage regulation unit, and a control interface. The image sensor unit comprises an optical lens array and a semiconductor image sensor that captures images of the monitored environment at a predefined frame rate and resolution for real-time analysis.

[0057] The captured image frames are internally transferred to a preprocessing unit, which is executed by at least one onboard processing unit. There, the image data is normalized, resized, filtered, and time-synchronized to remove noise, compensate for lighting variations, and align successive frames. The onboard processing unit comprises a dedicated AI accelerator or a heterogeneous processing architecture with a central processing unit and a neural processing unit optimized for executing trained object classification models. Non-volatile memory stores model parameters, inference instructions, object classification thresholds, and adaptive control rules.

[0058] After preprocessing, the image data undergoes feature extraction and classification, both performed entirely within the camera system. Classification identifies object categories, spatial positions, motion characteristics, and confidence levels of the detected objects in real time. The classification results are evaluated by decision logic executed on the processing unit, correlating them with predefined operating contexts, environmental conditions, and historical state data stored in memory.

[0059] Based on this analysis, the system generates adaptive control signals via the control interface. The control interface comprises electrical and communication interfaces configured to directly control or regulate one or more functional components of a machine or structure. These include, for example, motors, actuators, alarms, lighting systems, access control mechanisms, and safety interlocks. The control signals are dynamically adapted to changes in the object classification, thus enabling adaptive control in a closed-loop system without external intervention.

[0060] The system also supports continuous learning or periodic model updates, allowing updated classification parameters or inference models to be safely loaded into permanent memory without interrupting system operation. The energy management logic ensures efficient energy use by dynamically scaling processing activity based on scene complexity and operational requirements.

[0061] The drawing and the preceding description illustrate embodiments. Those skilled in the art will recognize that one or more of the described elements can be combined to form a single functional element. Alternatively, certain elements can be divided into several functional elements. Elements of one embodiment can be added to another. For example, the process flows described here can be modified and are not limited to the manner described herein. Furthermore, the actions of a flowchart need not be performed in the sequence shown; nor do all actions necessarily need to be carried out. Actions that do not depend on other actions can be performed in parallel with the other actions. The scope of protection of the embodiments is in no way limited by these specific examples. Numerous variations, whether explicitly stated in the description or not, such as...Differences in structure, dimensions, and materials are possible. The scope of protection of the embodiments is at least as comprehensive as described by the following claims.

[0062] The advantages, other benefits, and problem solutions have been described above with reference to specific embodiments. However, the advantages, benefits, problem solutions, and any components that can effect or enhance an advantage, benefit, or solution are not to be construed as critical, necessary, or essential features or components of the claims. REFERENCES 100 An intelligent camera system with integrated artificial intelligence for real-time object classification and adaptive function control 102 Image sensor unit 104 At least one processing unit 106 Non-transient memory 108 Image preprocessing unit 110 Units for Artificial Intelligence for Inference 112 Decision Control Unit 114 Control interface unit

Claims

[1] An intelligent camera system with integrated artificial intelligence for real-time object classification and adaptive function control, consisting of: an image sensor unit configured to capture successive image sequences of an observed environment; an optical assembly operationally coupled to the image sensor unit to project incident light onto a sensor surface; at least one processing unit that is arranged in a camera housing and electrically connected to the image sensor unit; a non-volatile memory that is operationally connected to the processing unit and configured to store executable instructions, trained object classification parameters, and control decision data; an image preprocessing unit, which is executed by the processing unit and is configured to normalize, filter, and align successive image frames in time; an artificial intelligence inference unit executed by the processing unit, configured to perform object classification in real time directly on the preprocessed image sequences using the stored trained parameters; a decision control unit executed by the processing unit, configured to generate adaptive control outputs based on classified object attributes such as object category, spatial position, motion properties, and confidence level; and A control interface unit that is electrically coupled to the decision control unit and configured to transmit the adaptive control outputs to one or more functional components of a machine or structure, wherein all image acquisition, classification, decision making and control signal generation takes place locally within the camera housing, wherein the image sensor unit comprises a solid-state image sensor configured to operate at a variable frame rate, which is dynamically adjusted by the processing unit based on the detected scene complexity and object motion characteristics to maintain classification accuracy in real time while reducing the computational load. [2] System according to claim 1, wherein the processing unit comprises a heterogeneous processing architecture including a universal processing element and a neural computing element optimized for the parallel execution of object classification operations stored in non-volatile memory. [3] System according to claim 1, wherein the artificial intelligence inference unit is configured to perform multiple classification passes on temporally adjacent image frames and aggregate the classification outputs to improve classification stability in the case of partial occlusion or rapid object movement. [4] System according to claim 1, wherein the decision control unit is further configured to compare classified object attributes with predefined operating thresholds and context state data stored in non-volatile memory to determine whether an adaptive control response is required. [5] System according to claim 1, wherein the control interface unit comprises at least one electrical output interface configured to directly actuate electromechanical components of the machine or structure without intermediate control hardware. [6] System according to claim 1, wherein the control interface unit is further configured to generate variable control signals whose size, duration or timing is dynamically adjusted based on real-time changes in the position and motion trajectory of the classified object. [7] System according to claim 1, wherein the non-volatile memory further stores historical classification data and previous control results and wherein the decision control unit is configured to access the historical classification data to refine subsequent control decisions. [8] System according to claim 1, wherein the processing unit is further configured to execute energy management logic that selectively activates or deactivates parts of the artificial intelligence inference unit based on the detected object presence in the field of view.

Images