System and method for detecting product defects across manufacturing process

A fault detection system for detecting faults in products, the fault detection system comprising:

• a processor;
• and
• a memory, communicatively coupled to the processor, wherein the memory stores processor-executable instructions, which on execution, cause the processor to:
• generate a plurality of wavelet coefficients corresponding to one or more transformed images of one or more products, wherein each of the plurality of wavelet coefficients correspond to at least one characteristic of each of the one or more products;
• determine a set of invariant features, corresponding to each of the one or more products, from the plurality of wavelet coefficients by comparing values of the plurality of wavelet coefficients with corresponding predetermined threshold coefficient values;
• generate a dynamic set of invariant features by grouping the invariant features into a set of groups based on type of the one or more products, wherein each group in the set of groups comprises wavelet coefficients distinct to the type of the one or more products;
• quantize the dynamic set of invariant features based on a predetermined quantization threshold and associating a representative coefficient signature for each group in the dynamic set of invariant features;
• and
• detect the faults in the one or more products based on comparison of coefficient signatures associated with each of the one or more products with the representative coefficient signature of each group in the dynamic set of invariant features, wherein the faults are detected by identifying mismatch of coefficient signatures, associated with each of the one or more products, with the representative coefficient signature of each group in the dynamic set of invariant features.

A non-transitory computer readable medium including instructions stored thereon that when processed by at least one processor cause a fault detection system to perform operations comprising:

• generating a plurality of wavelet coefficients corresponding to one or more transformed images of the one or more products, wherein each of the plurality of wavelet coefficients correspond to at least one characteristic of each of the one or more products;
• determining a set of invariant features, corresponding to each of the one or more products, from the plurality of wavelet coefficients by comparing values of the plurality of wavelet coefficients with corresponding predetermined threshold coefficient values;
• generating a dynamic set of invariant features by grouping the invariant features into a set of groups based on type of the one or more products, wherein each group in the set of groups comprises wavelet coefficients distinct to the type of the one or more products;
• quantizing the dynamic set of invariant features based on a predetermined quantization threshold and associating a representative coefficient signature for each group in the dynamic set of invariant features;
• and
• detecting the faults in the one or more products based on comparison of coefficient signatures associated with each of the one or more products with the representative coefficient signature of each group in the dynamic set of invariant features, wherein the faults are detected by identifying mismatch of coefficient signatures, associated with each of the one or more products, with the representative coefficient signature of each group in the dynamic set of invariant features.