Abstarct: baxsed on the development of human knowledge, the use of electronic, robotic tools and intelligent systems are developed in various industries. One of the important applications of intelligent systems is recognizing defects of products in mass production. Automation of defect detection process in the production of disposable containers, due to the public consumption and mass production and the products variation, is a fundamental action to improve the quality and reduce manpower in order to eliminate errors caused by itself. Automation has an important effect on the profits amount of production unit. The purpose of inspecting levels is to recognize and classify sections of surface and industrial products that do not comply with certain standards and do not have perfect quality and need to separate in production line. In this thesis, we present two methods for defect detection in disposable containers. In the first method we perform angle correction by checking geometrical information and using Canny edge detector. Then we detect lines of the containers using the Hough transform and inspect the inside lines for noise. Using the discrete wavelet transform and thresholding, we recognize defected containers from the proper. The accuracy reached to 88% in this method. To improve the accuracy of performance, we used features of wavelet packet and multi-laxyer perceptron neural network in order to train and find defected containers. The results were very favorable and the accuracy of about 97% is reached, that shows well performance of the final designed system. Keywords:baxsed on the development of human knowledge, the use of electronic, robotic tools and intelligent systems are developed in various industries. One of the important applications of intelligent systems is recognizing defects of products in mass production. Automation of defect detection process in the production of disposable containers, due to the public consumption and mass production and the products variation, is a fundamental action to improve the quality and reduce manpower in order to eliminate errors caused by itself. Automation has an important effect on the profits amount of production unit. The purpose of inspecting levels is to recognize and classify sections of surface and industrial products that do not comply with certain standards and do not have perfect quality and need to separate in production line. In this thesis, we present two methods for defect detection in disposable containers. In the first method we perform angle correction by checking geometrical information and using Canny edge detector. Then we detect lines of the containers using the Hough transform and inspect the inside lines for noise. Using the discrete wavelet transform and thresholding, we recognize defected containers from the proper. The accuracy reached to 88% in this method. To improve the accuracy of performance, we used features of wavelet packet and multi-laxyer perceptron neural network in order to train and find defected containers. The results were very favorable and the accuracy of about 97% is reached, that shows well performance of the final designed system.