Font Size: a A A

Researchs Of Target Detection Based On Visual Attention Mechanism And Its Application

Posted on:2018-10-12Degree:MasterType:Thesis
Country:ChinaCandidate:D J WangFull Text:PDF
GTID:2348330515484641Subject:Mechanical engineering
Abstract/Summary:PDF Full Text Request
At present,the mobile robot target detection as a multidisciplinary integrated complex technology,has now penetrated into t he military,industrial,life and other fields,and also has been widely concerned by researchers at home and abroad.The target detection of mobile robots is based on image processing,analysis and understanding,With the rapid increase of image imaging t echnology,the problem of image information redundancy is brought about by obtaining ultra-high resolution images,which increases the difficulty of image post-processing(such as target detection).In contrast,the human visual system has a very prominent information filtering ability.It can quickly select the important information to respond from the massive information,this selective and active physiological and psychological activities known as the visual attention mechanism.In this paper,the principle and method of visual attention mechanism were studied and used to detect the target automatically.The main research contents and results are as follows:(1)In this paper,we studied the biochemical theory of visual attention mechanism and analyzed the advantages and disadvantages of the existing significant detection algorithms,and then proposed to improve the IT model based on the special environmental target detection.The improved algorithm analyzed the significance of the underlying characteristics and calculated the most favorable features of the target region of interest,which made the extraction of the target region more accurate and complete,and made up the deficiency of the original IT algorithm.Because the images collected in a special environment contain noise,an integrative denoising method based on multiple median filter and curvelet transform was put forword to perform noise reduction pretreatment.So that the noises can be well removed and details can be well protected.Based on the noise reduction pretreatment,we uses the improved IT algorithm to extract the region of interest and carry out the target detection,the correct detection rate of target detection is 89%.It is proved that the special environmental target detection schem e based on the visual attention mechanism is feasible.(2)In this paper,we aim to learn the regional saliency estimator by the random forest algorithm of artificial intelligence from a set of training examples to calculate the saliency value of each regi on and obtain the saliency map,which makes the target detection process more intelligent.In the study of the regional characteristics,we selects the regional contrast feature,the regional background feature and the regional property feature which are favorable to the target region detection of the specific application scenario(shopping item inspection).Experiments showed that the speed of this algorithm was a little slower than several other algorithms,but the detection results was far better.This method greatly improved the accuracy of subsequent target segmentation and detection.At the same time,the correct detection rate of target detection is 92%.It is proved that the visual attention mechanism can be used for shopping items inspection.
Keywords/Search Tags:target detection, visual attention mechanism, image denoise, feature extraction, special environment target detection, shopping goods detection
PDF Full Text Request
Related items