Nowadays the automatic iection plays an irrefutable role in many industries. Usually iection of various units of steel companies in Iran is done by human. In pletizing plant of Mobarakeh Steel Complex the crude pellets are carried to furnace via big rectangular palets, for preheating, firing and drying. Due to high and great temperature variation, oxidizing, sudden palets collision, and because of hitting the pellets one the grate-bars, in over time the palets are damaged. The most significant issues are: curvness of palet car body, side-wall fall out, and grate-bars damages. These damages lead to products loss and inereasment in energy and cost.Therefore regular iection is neccessary. By continuosly reporting these damages to the pletizing plant and by repair or replacing the damaged palet, losses can be prevented. In these thesis the design and implementation of automatic palet iection system has been proposed. By installating three qued ultrasonic sensors above the inversed palet in return line and compairing these distances, palet car body curvnese was detected. Also for side wall fall out was recognized by two microswitchs. For finding the ratio of grate-bar damages, the video camera was installed in front of furnace entrance, turning inverse palets to horizontal position. At first step after converting the recoreded video to frame strings, the best frame of each palet was found by compairing the frames to the base frame. Then the palet area was extracted from background by detecting three middle lines between each grate-bar raws. After that the image palet contrast enhanced and the space between every two-adjanced grate bars is segmented by morphology and labeling approaches. At the next step 58 features (e.g. geometric, histogram, gray level co-occurrence and the wavelet coefficient matices) was extracted. Considering the complexcity and velocity by utilization of SVM-RFE for feature selection, 4 of the most important feature respect to stability and 0in 0in 0pt" dir=ltr Keywords Automatic Iection, Image processing, Image Enhancement, Feature Selection, SVM ltr"
