Humanoid soccer robot is one of popular developed robot. RoboCup is a competitive competiton of humanoid robot soccer. The rule of RoboCup changed by the time, the previous orange ball changed by white ball which is same color as the field line and the goal. Accordingly, in this research designed a white ball detection system for humanoid soccer robot based on pattern recognition. Histogram of Oriented Gradient (HOG) and Support Vector Machine (SVM) method are used in this research for feature extraction and classification. The result of this research is a system that be able to detect ball in humanoid soccer robot environment. The system tested by sliding window parameter testing, the distance of ball to robot testing, different light intensity testing, and other object testing. The conclusions of this research are: optimal detection is obtained by using 8x8 win_stride parameter size and 1,2 scale0 parameter value, maximum distance of detection with 32×32 window detector is 180 cm and with 64×64 window detector is 140 cm, the response of system in different light intensity is good enough, and the success rate of system against other obstacle object with 32×32 window detector is 68% and with 64×64 window detector is 99%.