Smart fire-extinguishing device and smart fire-extinguishing system including same

A smart fire extinguishing device comprising:

• a chemical storage configured to have therein a fire extinguishing agent;
• a fuse unit exhibiting to have an explosive force to spurt the fire extinguishing agent in the chemical storage to an outside;
• and
• a fire detector configured to detect temperature, flame, and smoke, determine whether a fire has occurred by comparing a result of detection of the temperature, flame, and smoke with a preset fire outbreak reference value, and spurt the fire extinguishing agent by operating the fuse unit according to a result of determination of whether the fire has occurred,
• wherein the fire detector includes:
• a sensor module configured to detect the temperature, flame, and smoke and flame;
• and
• a processor configured to determine whether the fire has occurred by comparing detection information of the sensor module with the preset fire outbreak reference value,
• wherein the processor includes:
• an image acquisition unit to acquire a video image from a video data and divide the video image into a plurality of blocks;
• a video processing unit to detect at least one flame area and at least one smoke area from the video image divided into the plurality of blocks;
• and
• a detection result output unit to calculate areas of the at least one flame area and the at least one smoke area and/or a number of the at least one flame area and the at least one smoke area, compare the calculated areas and/or the number of the at least one flame area and the at least one smoke area with the preset fire outbreak reference value, and output the result of the determination of whether the fire has occurred as a result of the comparison,
• wherein the video processing unit is configured to detect the at least one smoke area by:
• setting the plurality of blocks of the video image that has a reference value of brightness, motion accumulation, and diffusion of the smoke that is equal to or greater than a predetermined reference value as a plurality of regions of interest,
• learning a plurality of random forests for the plurality of regions of interest, respectively, using a random forest classification technique, and
• generating a cumulative probability histogram by accumulating classes generated as a result of the learning.