Research Of The Vehicle Target Detection Algorithm Based On Visual Saliency Analysis

Visual target detection technology is through the detection and extraction of the target in the image sequence, to obtain the parameters of the target, it is an important branch in the field of image processing and computer vision, which has wide application. To deal with the target detection problem in complex background, this paper proposed a significant features target detection+classification algorithms. This paper mainly completed the following work:Aiming at the vehicle target of complex scene image detection problem, it proposed to determine a target candidate region by the method of saliency analysis for acceleration detection. It included:using a bottom-up strategy and depending on the underlying visual features calculated characteristic saliency map to search the interesting area. By space domain and time domain strengthening characteristic saliency map, it merged similar strong response area to screen interesting area. It can reduce the number of false object. The detection target experiment in real scene image shows that the saliency analysis of space domain and time domain method can be quickly and accurately determine the target candidate region.Secondly, this paper uses unsupervised learning method to get the middle semantic features, such as visual bag of words. Using the visual word frequency statistics histogram as image features, SVM method is used to test the vehicle target classification. According to the experimental results, vehicle classification detection algorithm is analyzed and evaluated. The method proposed in this paper can satisfy certain requirements of the vehicle detection.