Object detection is a computer vision and image processing technique for detecting objects in a visual range in the forms of images or video streams. It’s used for locating the precise position, recognition, and monitoring of objects while labeling them with accurate localization. The method is used for people detection, facial recognition, patterns, objects, extraction texts, and vehicle identification.

Problem

Development of a framework to detecting vehicles and people using a network of a camera monitoring system for monitoring purposes and provide an analytical view of the detected objects as a dashboard with timebound KPIs and logistical statistics. The framework is intended to equip separate mechanisms apart from Computer Vision powered detections such as segregating the detection patterns and converting them to understandable statistical insights in terms of count of detected objects, percentile amounts, detections in correlation to the time of detection.

Solution

Our science team developed an automatic workflow to capture images using the camera network in specific intervals where motions are detected. These images are then run through the image analysis process for the detection of objects, peoples, and vehicles before saving them to cloud storage with faces blurred for reference purposes. Additional processes were separately developed and integrated into the main workflow to pass the analytics information to the Dashboard environment, data cleaning/structuring. The workflow can run in both serverless and hosted solutions depending on the preference of the end-user with configuration changes. Our team has provided the solution by systematically following the below 5 main stages.

• Development of the Object Detection Methodology.
• Automation of Framework Functions.
• Configuration of cloud Storage.
• Research and Development.
• Dashboard Development and Visualizations.