Machine-Code NPU Processing for Fast, Low-Power Object Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Movable apparatuses require artificial intelligence systems that can process information quickly, recognize distant objects, and operate with low power to support tasks such as autonomous flight and object detection.

Innovation Solution

An electronic device equipped with a neural processing unit (NPU) and signal generator to process artificial neural network models for object detection and tracking, and control camera operations to enhance detection accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If the processing speed of the artificial intelligence system is increased to support quick decisions in movable apparatus, then the response time for object detection is improved, but the power consumption increases

Engineering Contradiction:

Improveprocessing speedVSAvoidpower consumption

Core Design Contradiction:

SpeedVSUse of energy by moving object

Solution Approach 1:

The artificial intelligence system is divided into multiple processing elements (PEs) that operate in parallel. Each PE handles specific portions of the neural network computations, allowing the system to achieve high processing speeds through parallelization while managing power consumption by activating only necessary processing units

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system dynamically adjusts its processing mode based on operational requirements. The NPU can switch between different processing states and activate only the necessary number of processing elements based on the current task complexity, enabling quick decisions when needed while conserving power during normal operation

Inventive Principle:

Principle #15Dynamics

2Length of stationary object

If the artificial intelligence system is configured to recognize objects at great distances, then the detection range is improved, but the measurement precision and power consumption are adversely affected

Engineering Contradiction:

Improvedetection rangeVSAvoiddetection accuracy

Core Design Contradiction:

Length of stationary objectVSMeasurement precision

Solution Approach 1:

The system employs multiple cameras positioned at different locations and angles to capture images of distant objects. By processing images from multiple perspectives and combining the data, the system achieves both extended detection range and maintained measurement precision through multi-dimensional observation

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The neural processing unit acts as an intermediary that processes and fuses data from multiple cameras. It applies neural network algorithms to enhance the quality of distant object images, compensate for image degradation, and improve detection accuracy while maintaining the ability to detect objects at great distances

Inventive Principle:

Principle #24Intermediary (Mediator)

3Duration of action of moving object

If the artificial intelligence system operates continuously to maintain long flight duration, then the operational time is improved, but the power consumption increases

Engineering Contradiction:

Improveflight durationVSAvoidpower consumption

Core Design Contradiction:

Duration of action of moving objectVSUse of energy by moving object

Solution Approach 1:

The artificial intelligence system operates periodically rather than continuously. The NPU processes images at intervals and can enter low-power states between processing cycles. This periodic operation allows the system to maintain long flight duration by consuming power only when object detection is required, rather than continuously

Inventive Principle:

Principle #19Periodic action

Solution Approach 2:

The system autonomously manages its own power consumption by monitoring operational requirements and adjusting its activity level. The NPU can determine when object detection is necessary and activate processing only during those periods, enabling long flight duration through self-regulated power management

Inventive Principle:

Principle #25Self-service

4Measurement precision

If multiple cameras are used to improve object detection accuracy, then the measurement precision is improved, but the device complexity increases

Engineering Contradiction:

Improvedetection accuracyVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

Multiple cameras are merged into a unified processing system where the NPU receives and processes images from all cameras simultaneously. The neural network algorithms combine the data from multiple sources to improve detection accuracy while presenting a unified, manageable interface that reduces the perceived complexity of operating multiple cameras

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS12462421B2Apparatus including NPU operated based on machine codes corresponding to ANN model

Publication Date: 2025.11.04 DEEPX CO LTD

AI summary

An electronic device mounted on a fixed or a movable apparatus is provided. The electronic device may comprise an image signal processor (ISP) for at least one camera; a neural processing unit (NPU), including a plurality of processing elements (PEs), configured to: process an operation of an artificial neural network model trained to detect or track at least one object, based on an input feature map generated from at least one image, which is acquired via the ISP from the at least one camera, and output an inference result; and a signal generator generating a signal applicable to the at least one camera or the ISP.