CN-115861925-B - Intelligent visual monitoring and alarming method and system

CN115861925BCN 115861925 BCN115861925 BCN 115861925BCN-115861925-B

Abstract

The invention discloses an intelligent visual monitoring alarm method and system, which are characterized in that target detection frames are cached in a target frame cache dictionary structure, the latest obtained target detection frames are firstly subjected to calculation of an intersection ratio with the target detection frames in the target frame cache dictionary structure, whether alarm is inhibited or not is determined according to a set inhibition threshold, the target detection frames with which the intersection ratio does not reach the inhibition threshold are added into a target detection frame set T which needs to be alarmed, all target detection frames in the target detection frame set T are alarmed, repeated false alarms are inhibited, the condition that an intelligent visual monitoring system continuously misreports the target of a video picture background is reduced, and the technical problem that the traditional intelligent visual monitoring alarm method is easy to continuously generate repeated false alarms on the same target and seriously influences user experience is solved.

Inventors

LIAN JUNJIAN
ZHANG LIWEN
JIN ZIJIE
GAO CHUANG
DONG HUAPENG

Assignees

天翼数字生活科技有限公司

Dates

Publication Date: 20260505
Application Date: 20221128

Claims (6)

1. An intelligent visual monitoring and alarming method is characterized by comprising the following steps: S1, acquiring a plurality of target detection frames in a picture acquired by a camera; S2, classifying each target detection frame according to the target detection frame label, and searching a buffer target frame set corresponding to each target detection frame in a target frame buffer dictionary structure by taking the target detection frame type as a dictionary key, wherein the value in the target frame buffer dictionary structure is a (y, z) structure, y is a buffer target frame, z is a period count, and z is a constant m with an initial value being greater than 0; s3, extracting any target detection frame x in any type; s4, traversing a cache target frame set corresponding to the extracted target detection frame x in the target frame cache dictionary structure, and calculating the intersection ratio of the extracted target detection frame x and a cache target frame y in the cache target frame set; S5, judging whether an intersection ratio larger than a suppression threshold exists, if yes, suppressing the alarm of a target detection frame x, assigning the target detection frame x to a cache target frame y in a cache target frame set, resetting a life cycle count z to m+1, ending traversing the cache target frame set, and if not, marking the target detection frame x as needing to be alarmed, and adding (g, x) into a target detection frame set T needing to be alarmed; S6, judging whether all kinds of target detection frames are extracted, if not, extracting any target detection frame x in the next kind, returning to the step S4, and if so, executing the step S7; S7, alarming is carried out on all target detection frames in the target detection frame set T.
2. The intelligent visual monitoring and warning method according to claim 1, further comprising, after step S7: s8, subtracting 1 from the life cycle count of all values in the target frame cache dictionary structure, deleting the value with the life cycle count of 0, and adding (x, z) into the target frame cache dictionary structure according to the target detection frame set T.
3. The intelligent visual monitoring and alarming method according to claim 1, wherein step S1 specifically comprises: Acquiring a picture acquired by a camera; and carrying out target detection on the picture by adopting YOLOX algorithm, and extracting a plurality of target detection frames.
4. An intelligent visual monitoring alarm system, comprising: the target frame acquisition module is used for acquiring a plurality of target detection frames in the pictures acquired by the camera; The classifying module is used for classifying each target detection frame according to the target detection frame label, searching a buffer target frame set corresponding to each target detection frame in a target frame buffer dictionary structure by taking the target detection frame type as a dictionary key, wherein the value in the target frame buffer dictionary structure is a (y, z) structure, y is a buffer target frame, z is a life cycle count, and z is a constant m with an initial value being greater than 0; The extraction module is used for extracting any target detection frame x in any type; The cross-over ratio calculation module is used for traversing a cache target frame set corresponding to the extracted target detection frame x in the target frame cache dictionary structure and calculating the cross-over ratio of the extracted target detection frame x and a cache target frame y in the cache target frame set; The first judging module is used for judging whether the cross ratio larger than the inhibition threshold exists or not, if yes, the target detection frame x is inhibited from alarming, the target detection frame x is assigned to a cache target frame y in the cache target frame set, the life cycle count z is reset to m+1, the cache target frame set is traversed, if not, the target detection frame x is marked as needing alarming, and (g, x) is added into the target detection frame set T needing alarming; the second judging module is used for judging whether all kinds of target detection frames are extracted, if not, any target detection frame x in the next kind is extracted, the cross-over ratio calculating module is executed in a return mode, and if yes, the alarming module is executed; And the alarm module is used for alarming all the target detection frames in the target detection frame set T.
5. The intelligent visual monitoring alarm system of claim 4 further comprising: And the dictionary updating module is used for subtracting 1 from the life cycle count of all the values in the target frame cache dictionary structure, deleting the value with the life cycle count of 0, and adding (x, z) into the target frame cache dictionary structure according to the target detection frame set T.
6. The intelligent visual monitoring alarm system of claim 4 wherein the target frame acquisition module is specifically configured to: Acquiring a picture acquired by a camera; and carrying out target detection on the picture by adopting YOLOX algorithm, and extracting a plurality of target detection frames.

Description

Intelligent visual monitoring and alarming method and system Technical Field The invention relates to the technical field of machine vision monitoring, in particular to an intelligent vision monitoring alarm method and system. Background The intelligent vision monitoring is a computer vision method, and under the condition that human intervention is not needed, the positioning, the identification and the tracking of the target in the dynamic scene are realized by automatically analyzing the image sequence shot by the camera, and the behavior of the target is analyzed and judged on the basis, so that the daily management can be completed, and the response can be timely made when the abnormal situation occurs. The existing intelligent visual monitoring alarm method has the problem that false alarms are caused by misjudging objects in a picture as targets, and the continuous video picture background acquired in a certain time is basically unchanged due to the characteristic that the monitoring camera is kept at the same position for a long time, if false alarms occur, the same targets are continuously and repeatedly alarmed, and user experience is seriously affected. Disclosure of Invention The invention provides an intelligent visual monitoring alarm method and system, which are used for solving the technical problems that the traditional intelligent visual monitoring alarm method is easy to generate continuous repeated false alarms on the same target and seriously affects the user experience. In view of the foregoing, a first aspect of the present invention provides an intelligent visual monitoring and alarming method, including: S1, acquiring a plurality of target detection frames in a picture acquired by a camera; S2, classifying each target detection frame according to the target detection frame label, and searching a buffer target frame set corresponding to each target detection frame in a target frame buffer dictionary structure by taking the target detection frame type as a dictionary key, wherein the value in the target frame buffer dictionary structure is a (y, z) structure, y is a buffer target frame, z is a period count, and z is a constant m with an initial value being greater than 0; s3, extracting any target detection frame x in any type; s4, traversing a cache target frame set corresponding to the extracted target detection frame x in the target frame cache dictionary structure, and calculating the intersection ratio of the extracted target detection frame x and a cache target frame y in the cache target frame set; S5, judging whether an intersection ratio larger than a suppression threshold exists, if yes, suppressing the alarm of a target detection frame x, giving the target detection frame x to a cache target frame y in a cache target frame set, resetting a life cycle count z to m+1, ending traversing the cache target frame set, if no, marking the target detection frame x as needing to be alarmed, and adding (g, x) into a target detection frame set T needing to be alarmed; S6, judging whether all kinds of target detection frames are extracted, if not, extracting any target detection frame x in the next kind, returning to the step S4, and if so, executing the step S7; S7, alarming is carried out on all target detection frames in the target detection frame set T. Optionally, step S7 further includes: s8, subtracting 1 from the life cycle count of all values in the target frame cache dictionary structure, deleting the value with the life cycle count of 0, and adding (x, z) into the target frame cache dictionary structure according to the target detection frame set T. Optionally, step S1 specifically includes: Acquiring a picture acquired by a camera; and carrying out target detection on the picture by adopting YOLOX algorithm, and extracting a plurality of target detection frames. The second aspect of the present invention provides an intelligent visual monitoring and warning system, comprising: the target frame acquisition module is used for acquiring a plurality of target detection frames in the pictures acquired by the camera; The classifying module is used for classifying each target detection frame according to the target detection frame label, searching a buffer target frame set corresponding to each target detection frame in a target frame buffer dictionary structure by taking the target detection frame type as a dictionary key, wherein the value in the target frame buffer dictionary structure is a (y, z) structure, y is a buffer target frame, z is a life cycle count, and z is a constant m with an initial value being greater than 0; The extraction module is used for extracting any target detection frame x in any type; The cross-over ratio calculation module is used for traversing a cache target frame set corresponding to the extracted target detection frame x in the target frame cache dictionary structure and calculating the cross-over ratio of the extracted target detection frame x and