Research On The Detection Algorithm Of Daily Tools For Households Based On RGB-D Data

With the economic development and technological progress,the robot began to enter people’s lives,and can complete the path planning,cleaning and other daily operations.Due to the particularity of the family environment,the rapid and accurate cognition of the robot to the daily tools is still a difficult problem in the field of machine intelligence research and the practical process of robot.Based on the relevant research results at home and abroad,this paper discusses the color depth data to carry out the household daily tool significance detection method to meet the real-time requirement of robot service.Firstly,considering the characteristics of the natural map and the local aggregation in the space,the model divides the large-scale layer of the original image to extract the overall structural characteristics of each region.By small-scale segmentation to preserve the details of the target and ensure the small The characteristics of the region consistency;the use of pixel color correlation and contrast,measure the degree of independence of the characteristics of the selected exact seed derived from the scale of the significant map,the different scales of significant results fusion based on the color of the initial significant The Compared with the traditional method,this method can overcome some background interference,generate a low noise significant map.Then,in the feature comparison section of the significance test,the traditional model often uses the image boundary region as the hypothetical background seed,and uses the color contrast to measure the significant level of each region,but it is difficult to deal with the lower background confidence and the single image area.In order to solve this problem,the initial significant graph is used as a known a priori to correct the background of the hypothesis,so as to obtain a more accurate background seed set,and further introduce the image depth information to construct a more expressive color depth feature to enhance the area Interval distinguishing feature.Combined with the improved representation of the seed set and the color depth of the combination of the characteristics of the way,plus the significant target in the image of the existence of the state.Compared to most of the current algorithm,effectively improve the accuracy of detection accuracy.