Kmeans-based efficient vehicle detection method

Abstract

The invention discloses a kmeans-based efficient vehicle detection method, which comprises the steps of off-line learning and real-time detection. In the off-line learning step, a sample preprocessed whitening matrix obtained in the off-line learning process is utilized for whitening operation of samples in real-time detection, the influence caused by inconsistent weights of different components is avoided, K decision-making trees are trained by using training samples, and the decision-making trees are used for real-time detection, so as to obtain a detection target. In the real-time detection step, simple features are used for selecting a candidate frame which is possibly a vehicle, kmeans feature extraction is conducted, and the K decision-making trees are used for prediction classification, so as to obtain a detection target. Compared with the prior art, the kmeans-based efficient vehicle detection method has the advantages of improving the detection precision, enhancing the real-time performance, reducing the probability of false detection and missing detection, and the like.

Description

technical field [0001] The invention relates to vehicle detection technology, in particular to a kmeans-based efficient vehicle detection method, which is a vehicle detection method for vehicle candidate re-verification. Background technique [0002] In recent years, the growth rate of vehicles has been much higher than that of roads and other traffic facilities. Frequent traffic accidents, increasing casualties, and huge property losses require vehicles not only to have good safety but also to have certain intelligence. With this, the concept of Intelligent Vehicle (Intelligent Vehicle) came into being. Vehicle detection is an important part in the field of intelligent vehicle research, and it is extremely critical for the safe driving of intelligent vehicles. At present, there are many types of vehicle detector products produced at home and abroad, with different technical principles and implementation methods, such as coil detection, video detection, microwave detection,...

AI Technical Summary

Problems solved by technology

A robust image feature can improve the classification prediction results of the classifier very well. However, the design of vehicle features in the traditional image field requires rich experience of image experts, and the update is slow.

Images