GA/T 1399.2-2017Video and image analysis system for public security—Part 2: Technical specifications for analysis and description of video and image content (English PDF)

3.1 Terms and definitions

GA/T 1399.1-2017, GA/T 1400.1-2017 and GB/T 30147-2013 and the following terms and definitions Applies to this document.

3.1.1

True positive

The target or event that should be detected appears in the video image, and the video image analysis system outputs the correct detection result.

3.1.2

Falsenegative

The target or event that should be detected appears in the video image, but the video image analysis system does not output the correct detection result.

3.1.3

False positive

The target or event that should be detected does not appear in the video image, but the video image analysis system outputs the detection result.

5.6 Video summary

5.6.1 It should be supported to remove the redundant part of the video that is not of interest, keep the concerned part, and generate a new video file.

5.6.2 Should support the reconstruction of the target video images at different time points in the same video screen to generate new video files.

5.6.3 The format of the newly generated video file should conform to the format specified in GB/T 28181-2016.

6.1 Video quality

The quality of the input video image should reach level 4 or above described in 5.4.3 of GB 50198-2011.

6.2 Target detection and feature extraction

6.2.1 Moving target detection

For the detection of moving objects in the video that are larger than (16×16) pixels and speed greater than 1 pixel/frame, the following requirements should be met.

a) The detection rate should not be less than 95%;

b) The false detection rate should not be higher than 15%.

6.2.2 Target color detection

For the detection color of the target area larger than (32×32) pixels in the video image collected under good lighting conditions, it should meet the following Claim.

a) The recognition accuracy of pink, red, orange, yellow, green, blue, and purple colors should not be less than 80%;

b) The recognition accuracy of black, white, gray and brown colors should not be less than 70%.

6.2.3 Pedestrian detection

For pedestrian detection larger than (32×64) pixels in the video image, the following requirements should be met.

a) The detection rate should not be less than 90%;

b) The false detection rate should not be higher than 10%.

6.2.4 Vehicle inspection

The detection of vehicle targets larger than (64×64) pixels in the video image should meet the following requirements.

a) The detection rate should not be less than 90%;

b) The false detection rate should not be higher than 10%.

6.3 Target quantity analysis

The performance indicators of the target quantity analysis should comply with the relevant regulations in GB/T 30147-2013 in 7.8 and 7.9.

6.4 Target recognition

6.4.1 Vehicle number plate recognition

To identify the vehicle number plate with a width greater than 100 pixels in the video image, the following requirements should be met.

a) During the day. the recognition accuracy of the vehicle number plate number in the front direction is not less than 95%, and the vehicle number plate number recognition in the rear direction is accurate the rate is not less than 80%;

b) At night. the recognition accuracy of the vehicle number plate number in the front direction is not less than 90%, and the vehicle number plate number recognition in the rear direction is accurate the rate is not less than 70%.

The color recognition, type, and identification performance requirements of the number plate of the vehicle shall comply with the relevant regulations of GA/T 833-2016.

6.4.2 Recognition of basic characteristics of vehicles

6.4.2.1 Body color recognition

Car body color recognition for vehicle targets larger than (256×256) pixels in vehicle video images collected under good lighting conditions no, the recognition accuracy of white, gray, yellow, pink, purple, green, blue, red, brown, black and other body colors should not be less than 70%.

6.4.2.2 Vehicle identification

When vehicle type recognition is performed on a vehicle target that is larger than (256×256) pixels in the video image and recognized as a car, the characteristics of the vehicle image are complete Under the premise, the recognition accuracy rate of the vehicle type (car, trailer, motorcycle, etc.) should not be less than 90%, and the vehicle sub-type (bus, car, bread, etc.) the recognition accuracy rate of vehicles, off-road buses, heavy and medium trucks, and other vehicles, etc.) should not be less than 80%.

6.4.2.3 Other identification

The performance requirements for the recognition of other basic features of the vehicle shall comply with the relevant regulations of GA/T 833-2016.

6.5 Behavior analysis

The performance indicators of behavior analysis and judgment should meet the relevant regulations in GB/T 30147-2013 in 7.2~7.7.

6.6 Video summary processing capability

The processing capacity of video abstracts should not be lower than the performance requirements of 4.4.1.5 in GA/T 1154.2-2014.

7.1 Characteristic attribute description

7.1.1 Refer to A.1 in Appendix A for the description requirements of characteristic attributes of video image analysis rules.

7.1.2 The feature attribute of the output result mainly refers to the feature attribute of the video image tag. Refer to the attachment for the description of the characteristics of the video image tag Record A.2 in A.

7.2 XMLSchema

7.2.1 See B.1 in Appendix B for the XMLSchema description of video image analysis rules.

7.2.2 For the XMLSchema description of video image tags, see B.2.in Appendix B.