Research On Multi-target Recognition Technology Of The Mobilephone Film’s Image Based On Features Combination

Multi-target recognition technology is the core technology of the automatic detection and recognition system of industrial production line, the research involves many fields of image processing, pattern recognition, automatic control and artificial intelligence. This paper regards the mobile phone film of the industrial production line as the research object to study the mobile film’s image recognition technology for multi-object aim at the problem of wrong or omissive identification because of the complex background and partial occlusion,the main research contents are as follows:(1) The hardware design of the system. First,analysis of the principle of selecting hardware in visual detection application; then,in order to obtain high quality images and combine with the rate accuracy requirements of the mobile phone film to determine the hardware’selection of the system including light source,lens,camera;(2) Research on image preprocessing algorithm. Get the two value image of mobile phone film image by preprocessing. Pretreatment process mainly includes the gray processing by weighted average method,image filtering by using ten median filter,image enhancement by histogram equalization enhance,image segmentation with Otsu threshold segmentation algorithm; finally get the clear the two value image of mobile phone film;(3) Study on the feature extraction technology and multi-target recognition algorithm.First,make the connected domain identification for the two value image by the 8 connected domain identification method,and make the position correction of the mobile phone film image by taking advantage of rotation and translationof moment invariants,and obtain some undetermined target; Secondly,extract area features of the undetermined targets and exclude non-real target according to known initial threshold of the area of the single target; Then, cluster to the true target using K means algorithm to get some kind of real targets; Finally,select a target from each category of objectives and to extract the 5*5 ratio feature to match with the standard template,in order to determine whether the target is a mobile phone film;(4) The recognition algorithm research of occluded objects.Aiming at the partially occluded situation of the mobile phone film in the process of recognition, choose the improved distance as the recognition feature to identify mobile phone film. Use the template of the two value image to slide pixel by pixel in the connected domain after marking, calculate the improved distance in each corresponding position. Then the Hausdorff distance value and the set threshold value will be compared, if it is less than the set threshold, the image in the position is the target image.The experimental results show that the multi-target recognition algorithm based on the feature combination proposed in the paper is a good solution to the problem of multi-target recognition in the size zoom and rotation angle. And for the possible situation of partial occlusion, the improved distance is very useful for the recognition of the partially occluded mobile phone film.