AI Products / Services > AI Software

ZIA™ Plate

header image

Overview

Low-load and high-precision license plate recognition software utilizing AI technology, which supports not only cameras on general roads but also security management and improvement of customer satisfaction in various fields such as parking lots, factories, offices, hospitals, hotels, public facilities, financial institutions and retail stores.

Example of application: License plate detection

(Numbers are masked for privacy protection.)

Features

Highly precise license plate recognition
Being equipped with a recognition engine utilizing AI technology makes it possible to read vehicle numbers with high precision.
Recognition Precision
95% or more
Recognition Speed
38ms (*Measured by DMP's AI FPGA module ZIA™ C3)
Operable on edge devices
1/100 lower load than software OCR technology.
Provision in source code makes it possible to implement in customer's target edge devices such as CPUs, GPUs, or DMP's AI processor ZIA™ DV720)
Applicable to various applications
ZIA™ Plate supports security management and customer satisfaction in various fields such as factories, offices, hospitals, hotels, public facilities, financial institutions, and retail stores in addition to roadside cameras.
Possible to recognize special plates and various countries plates
Possible to recognize special license plates and various countries license plates
By utilizing AI, it can recognize license plates with designs such as Mt. Fuji or Olympics. It also supports overseas license plates recognition by learning images.

Specification

License plate color tone (density)

  • The density of a license plate can be detected. Frame color detection is not supported.
  • Size (large format / medium format) detection is not supported.

Correspondence to land transportation offices

  • Corresponds to the names of the regions started in 2020 such as Shiretoko, Tomakomai, and Hirosaki.
  • One-character display (older version) is not supported.

Vehicle classification number

  • Compatible with numbers and A, B, C, E, F, H, K, L, M, P, T, X and Y

Vehicle model classification

  • Compatible with hiraganas except お, し, へ, ゐ, ゑ and ん and A, B, C, E, F, H, K, L, M, P, T, X and Y

Precision

Angle conditions
Depression angle: within ±30°, Rotation angle(Horizontal angle): within ±30°, Approach angle: within ±30°
Land transportation offices: 97%,Vehicle classification number: 98%, Vehicle model classification: 97%, Vehicle number: 99%

Recommended Usage Environment

  Operation Environment
Embedded
On-premises / Cloud
  • Ubuntu, Amazon Linux2 etc. (x86_64)
Stand-alone PC
  • Windows 10 (x64_64) or Ubuntu (x86_64)

Implementation Steps

1. Software Evaluation
  • Select evaluation environment (camera type, software operating device)
  • Prepare input image data (considering the number of pixels, shooting angle, external shooting environment conditions)
  • Prepare trained model (model supplied with ZIA™ Plate or prepared by customer)
2. Re-Training of Trained Model
  • Extract unrecognized image conditions
  • Collect images (training data) under the conditions to be strengthened and re-train
3. Software Implementation
Field Test
  • Design and examine overall software operation system
    1. All edge processing
    2. Partial edge processing (edge & cloud environment)
  • Design and implement screen display (pre-processing / post-processing)
  • Perform field test in the actual usage environment
4. Operation / Maintenance
  • Possible to perform timely maintenance by re-learning the trained model regarding the occurrence of external factors. For example,
    1. Specification change of license plate display due to law revision
    2. Change of camera device
    3. Change of shooting angle, usage environment conditions

Related News

expand_less