Introduction

Vehicle Detection Systems are enabling law enforcement firms to automatically detect vehicles from images and videos. This enhances the capabilities and expediate the entire investigative process. Modern traffic surveillance systems have enabled us to gather videos and images of roads in different day-to-day traffic conditions.


This feature was launched in Version 8.0.239


VIDIZMO provides automatic vehicle detection systems using the videos and images from these surveillance systems and detect vehicles along with their color, direction, and type. VIDIZMO categories a vehicle into 4 different categories:

  1. Car
  2. Truck
  3. Bus
  4. Bike

VIDIZMO, with its robust content processing capabilities, has built an AI indexer automatically detects vehicles within videos and images.


Concept

Automatic Vehicle Detection employs machine learning algorithms to detect vehicles in videos that frequently include cross crime scenes mainly from CCTV footages. This Automatic Vehicle Detection technique enables VIDIZMO indexer to read vehicles quickly and automatically, without the need for human intervention.


VIDIZMO Indexer takes videos and images as input and processes them to detect the vehicles within them. The time taken in processing depends on multiple factors such as video quality, duration and resolution. Once the processing is completed, bounding boxes can be seen over the detected vehicles from studio space. User can also filter the results on the basis of color and direction of vehicles.

To understand redaction in detail, please refer to this article: Understanding Redaction using Studio Space. In order to redact the automatically detected license plates, please refer to this article:  How to Redact PII from Digital Evidences 

Moreover, VIDIZMO Indexer also gives the flexibility to choose from three different models, which are "Small", "Medium", and "Large". It also allows the user to set a confidence threshold, a tracking frame based on which the indexer will detect license plates so that the user is able to get most out of the VIDIZMO Indexer capability according to the use case.

Please refer to the following article to understand what model size, confidence threshold, and tracking frames are: How to configure VIDIZMO Indexer for object detection  

 

Color

After vehicles are detected the ROI (detected images) are sent to the Machine Learning color classifier to assign respective color to each object. For every tracked vehicle, the color is continuously detected over multiple images/frames, and the color which appears in most of the images/frames of that tracked vehicle is assigned to it. At the moment VIDIZMO is able to detect and assign the following colors to vehicles:

  1. White
  2. Black
  3. Yellow
  4. Green
  5. Blue
  6. Red


Direction

For Direction there must be a change of 100 pixels in the position of an object over "x" or "y" coordinates. The change of position is calculated from the first and last frames in which a particular vehicle is detected. In case the difference is less than 100 pixels the accuracy of detected direction could be affected. At the moment VIDIZMO is able to detect the following directions of vehicles:

  • North
  • East
  • South
  • West
  • North-East
  • North-West
  • South-East
  • South-West



Detection

VIDIZMO Indexer's Automatic Vehicle Detection capability offers numerous advantages that are the basis for real-world scenarios. The majority of its advantages are related to road safety and traffic analysis. Some of the advantages VIDIZMO indexer offers for Automatic Vehicle Detection are listed below:


Performance

Human input is not required for precise and fast Vehicle detection. As a result, it promotes cost-effective governance and shortens wait times. VIDIZMO Indexer detects all vehicles with maximum accuracy in a short span of time.


Night and Daytime lightening condition

VIDIZMO vehicle detection accuracy reaches above 80% in daytime. Poor image/video quality and low light conditions can affect this accuracy.


Weather Condition

VIDIZMO Indexer has been built keeping in mind the importance of detecting vehicles in every weather condition including whether if its sunny, cloudy or rainy. Vehicles in different weather conditions are efficiently detected. As a result, it promotes higher detection accuracy and less manual work.



Contributions were made by Nabeel Ali & Mustafa Zulfiqar.


Read More

Understanding Automatic License Plate Recognition (ALPR)