A method for detecting illegal parking of vehicles

Abstract

The invention provides a method for detecting vehicles illegally parked and occupying lanes, which specifically includes: acquiring ROI binary images, acquiring images to be detected, vehicle detection, judging whether vehicles are detected in the ROI area and counting them, and counting the number of vehicles in the interval Steps such as proportion of times, sending out alarms, etc. This vehicle illegal parking detection method detects the vehicles in the video in real time, and uses a deep learning model to complete the vehicle detection. It is not affected by bad weather, light conditions, lens shake, foreign objects, etc., which greatly improves the detection accuracy. rate, while ensuring real-time and reliability, and can realize automatic alarm, high degree of intelligence.

Description

technical field [0001] The invention relates to the technical field of road monitoring, in particular to a detection method for vehicles illegally parking and occupying roads. Background technique [0002] With the development of modern technology and the improvement of people's living standards, cars are showing a rapid growth trend, but the demand for parking spaces brought about by the growth of more and more cars is also gradually increasing. The lack of parking spaces leads to frequent violations of vehicles, so vehicles Parking and occupancy detection is becoming increasingly important. The existing vehicle illegal parking detection technology mainly uses GPS to locate the area where the vehicle is currently located, or uses the reading equipment set up in the no-parking section or area and the vehicle electronic identification installed on the car windshield to read the car identity The information reaches the detection of illegal parking, or the traffic police super...

AI Technical Summary

Problems solved by technology

Among them, the simple image processing method represented by the background difference method has low accuracy and is susceptible to external influences such as bad weather, lighting conditions, lens shake, and foreign objects; while SIFT (scale-invariant feature transform matching algorithm) Although the representative feature extraction algorithm is not easily affected by radiation transformation and illumination, its algorithm is complex and time-consuming, so the real-time performance is not good. At the same time, there are also situations where feature points cannot be accurately extracted for smooth edges, which is likely to cause certain errors. sentence

Images