─ Core Technology ─
Neural networks and deep learning
Data Collection and Preparation
Deep learning first focuses on data. We collects sufficient data and executes filtering, cleaning, unified format, missing value processing, transform and other processes of decontamination and storage to obtain useful data set which can make the subsequent machine learning output more ideal.
Learning method and network model selection
According to various requirements and applications, we evaluate the deep learning methods and neural network models such as using Convolutional Neural Networks (CNN) for Supervised Learning to allow machine learn to recognize images, or Generative Adversarial Network (GAN) for unsupervised learning which can repair or synthesize pictures.
Analysis, evaluation and parameter adjustment
For machines that have completed deep learning training, we use test data to confirm whether the learning goals are achieved. However, in response to changes in business or environment, it is necessary to collect complete information again and re-learn and train, and constantly correct the thresholds or parameters in the hidden layer according to the results of each round of training. So that the AI model learned and trained can achieve the best results.
Core Functions
Image recognition
The automatic defect classification system detects defects in the image and judges the defect types automatically. The deep vision processing engine breaks through the bottleneck of traditional image processing technology through deep learning technology and provides high-precision defect classification results.
Automatic take-off and return
Through artificial intelligence, the integration of various sensing signals and the core UTM platform enable the drone to have the ability to automate the take-off and return to the field, and reduce the flight safety incidents caused by human operation errors.
UAV autonomous protection
Through the latest artificial intelligence technology, we integrates various sensing signals and the Counter Unmanned Aerial System (C-UAS) platform. Using a full-scale process of real-time detection, analysis, prediction and processing to identify the interception time of the UAV that can achieve the intrusion target. Thereby it is automatically proceeding the target Capture operations and assist personnel or carriers to fully realize the potential and expected benefits of operations.